Datasets:

TaylorAI

/

pubmed_commercial

like 12

PMC005xxxxxx/PMC5001701.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756421410.1371/journal.pone.0161771PONE-D-16-09893Research ArticleBiology and life sciencesGeneticsGene expressionGene regulationMicroRNAsBiology and life sciencesBiochemistryNucleic acidsRNANon-coding RNAMicroRNAsBiology and Life SciencesCell BiologyCellular TypesAnimal CellsConnective Tissue CellsFibroblastsBiology and Life SciencesAnatomyBiological TissueConnective TissueConnective Tissue CellsFibroblastsMedicine and Health SciencesAnatomyBiological TissueConnective TissueConnective Tissue CellsFibroblastsMedicine and Health SciencesPathology and Laboratory MedicinePathogenesisBiology and Life SciencesImmunologyImmune ResponseInflammationMedicine and Health SciencesImmunologyImmune ResponseInflammationMedicine and Health SciencesDiagnostic MedicineSigns and SymptomsInflammationMedicine and Health SciencesPathology and Laboratory MedicineSigns and SymptomsInflammationMedicine and Health SciencesSurgical and Invasive Medical ProceduresTransplantationOrgan TransplantationLung TransplantationMedicine and Health SciencesSurgical and Invasive Medical ProceduresRespiratory System ProceduresLung TransplantationBiology and Life SciencesCell BiologyCellular TypesAnimal CellsEpithelial CellsBiology and Life SciencesAnatomyBiological TissueEpitheliumEpithelial CellsMedicine and Health SciencesAnatomyBiological TissueEpitheliumEpithelial CellsMedicine and Health SciencesPulmonologyBronchiolitisMedicine and Health SciencesOncologyCancers and NeoplasmsCarcinomasAdenocarcinomasAdenocarcinoma of the LungMedicine and Health SciencesOncologyCancers and NeoplasmsLung and Intrathoracic TumorsAdenocarcinoma of the LungIdentification of miRNAs Potentially Involved in Bronchiolitis Obliterans Syndrome: A Computational Study Computational Analysis of miRNA Involvement in BOShttp://orcid.org/0000-0002-7512-5356Di Carlo Stefano 1*Rossi Elena 2Politano Gianfranco 1Inghilleri Simona 4Morbini Patrizia 2Calabrese Fiorella 5Benso Alfredo 1Savino Alessandro 1Cova Emanuela 4Zampieri Davide 6Meloni Federica 341 Control and Computer Engineer Department, Politecnico di Torino, Torino, Italy2 Department of Molecular Medicine, University of Pavia, Pavia, Italy3 Department of Internal Medicine, University of Pavia, Pavia, Italy4 Department of Respiratory Diseases IRCCS Policlinico S. Matteo, Pavia, Italy5 Department of Cardiovascular and Thoracic Sciences, University of Padova, Padova, Italy6 Thoracic Surgery Unit, Department Cardiothoracic and Vascular Sciences, University of Padova, Medical School, Padova, ItalyMallick Bibekanand EditorNational Institute of Technology Rourkela, INDIACompeting Interests: The authors have declared that no competing interests exist. Conceived and designed the experiments: SDC ER FM. Performed the experiments: ER SI EC. Analyzed the data: SDC AB GP AS ER SI PM FC. Contributed reagents/materials/analysis tools: PM FC FM SDC DZ. Wrote the paper: SDC ER FM PM GP. * E-mail: stefano.dicarlo@polito.it26 8 2016 2016 11 8 e016177110 3 2016 11 8 2016 © 2016 Di Carlo et al2016Di Carlo et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.The pathogenesis of Bronchiolitis Obliterans Syndrome (BOS), the main clinical phenotype of chronic lung allograft dysfunction, is poorly understood. Recent studies suggest that epigenetic regulation of microRNAs might play a role in its development. In this paper we present the application of a complex computational pipeline to perform enrichment analysis of miRNAs in pathways applied to the study of BOS. The analysis considered the full set of miRNAs annotated in miRBase (version 21), and applied a sequence of filtering approaches and statistical analyses to reduce this set and to score the candidate miRNAs according to their potential involvement in BOS development. Dysregulation of two of the selected candidate miRNAs–miR-34a and miR-21 –was clearly shown in in-situ hybridization (ISH) on five explanted human BOS lungs and on a rat model of acute and chronic lung rejection, thus definitely identifying miR-34a and miR-21 as pathogenic factors in BOS and confirming the effectiveness of the computational pipeline. This work was supported by Progetto Cariplo RIF. 2013-0943—BANDO 2013—Ricerca Medica. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll source code files and relevant data are available on GitHub (https://github.com/sysbio-polito/BOS-miRNA-Enrichment).Data Availability All source code files and relevant data are available on GitHub (https://github.com/sysbio-polito/BOS-miRNA-Enrichment). ==== Body Introduction Since James Hardy performed the first human lung transplantation (LTx) in 1963, LTx has become an accepted therapeutic option for carefully selected patients with an end-stage lung disease. The number of LTx has increased steadily over the last two decades and currently more than 3,000 lung transplantation procedures are performed yearly worldwide. LTx mainly intends to improve survival. However, long-term survival is hampered by the development of chronic lung allograft dysfunction (occurring in up to 50% of patients in the 5th post-transplant year), whose main clinical manifestation is represented by Bronchiolitis Obliterans Syndrome (BOS) [1]. As a consequence, long-term survival after LTx remains worse than that achieved for other solid organ transplantations (median overall adult survival, 5.6 years according to the latest ISHLT figures https://www.ishlt.org/registries/quarterlyDataReport.asp). Histologically, BOS is characterized by Bronchiolitis Obliterans (BO), i.e. patchy submucosal fibrosis involving the respiratory bronchioles, resulting in near-total or total occlusion of the airway lumen. BOS is the final outcome of an array of injuries to the airway epithelium and extracellular matrix including alloimmune-specific reactions, autoimmune responses and non-specific inflammatory insults (infections and gastro-esophageal reflux). The exact pathogenesis of BOS is still largely unknown. It is believed that an insult to the epithelium leads to an inflammatory response that might be associated with a shift from tolerance to immune activation, which in turn induces severe neutrophilic airway inflammation [2]. Activated neutrophils can further damage the epithelium by releasing reactive oxygen species, chemokines, alarmins and metalloproteinases (MMPs). This inflammatory phase is ultimately followed by a fibro-proliferative phase leading to transition from epithelial to mesenchymal cells, proliferation of myofibrocytes, and collagen deposition. Even if the pathogenesis of BOS following LTx has not been clarified completely, recent studies suggest that epigenetic regulations involving microRNAs may play an important role in its development [3][4][5]. MicroRNAs (miRNAs) are the most studied class of small non-coding RNAs, containing approximately 20 nucleotides. They regulate post transcriptional gene expression by binding to target messenger RNAs (mRNAs) and inducing either inhibition of translation or mRNA degradation [6][7][8][9]. It is thought that miRNAs regulate approximately 30% of the human protein-coding genome [6]. They control the expression of genes involved in several key biological processes such as cell development, stem cell proliferation, division and differentiation, regulation of immunity, apoptosis, cell signaling and metabolism, etc. [6][8][9][10][11]. Moreover, aberrant expression of miRNAs has been associated with several pathological processes, such as cancer [9][12][13], metabolic disorders [14], autoimmune diseases [11] and acute cellular rejection following kidney transplantation [15]. Clarifying the role of miRNAs in disease pathogenesis is therefore crucial. There are two approaches to the study of the role of miRNAs in diseases. The traditional "omics" approach assesses a wide panel of factors on plasma/tissue samples and identifies those significantly (2-fold to 10-fold) dysregulated in disease samples with respect to controls. In the other, computational algorithms are used to perform a preliminary exploratory analysis to narrow down number of factors to test in the subsequent wet-lab validation phase. These approaches, however, do not always provide information on the specific cellular site of expression, which is far more relevant than absolute expression levels. In this paper, we implemented a pure computational pipeline to perform enrichment analysis of miRNAs in pathways involved in BOS pathogenesis. This enabled us to identify a limited panel of candidate miRNAs with potential involvement in BOS development. Unlike in other approaches, such as GSEA [16] and miTEA [17], we did not rely on expression profiles data to perform our analysis, but instead we integrated and processed information from a large set of different repositories, thus performing a pure exploratory analysis to pinpoint which specific wet-lab experiments would be most useful. Preliminary validation of the selection of miRNAs was attained both through literature analysis during the computational phase; and by subsequent in situ hybridization experiments on miR-34a and miR-21, two of the resulting candidate miRNAs. Interestingly, hybridization experiments clearly showed, for the first time, dysregulation of these two miRNAs at cellular level in BOS tissue samples. This provided important insights into their role in the pathogenesis of BOS, opening perspectives for further mechanistic and therapeutic studies. Materials and Methods Fig 1 shows the experimental workflow followed in this study. Green boxes identify the computational pipeline implemented to identify the candidate miRNAs involved in BOS, whereas blue boxes identify wet-lab activities performed for validation of the results. The computational pipeline implements a set of data fusion engines able to automatically collect and integrate omics information from a set of public repositories (orange boxes in Fig 1). Collected data are then analyzed through a set of statistical tests to score the analyzed miRNAs. The source code and the related data implementing the computation pipeline described in Fig 1 are publicly available through the BOS-miRNA-Enrichment repository on GitHub at https://github.com/sysbio-polito/BOS-miRNA-Enrichment. 10.1371/journal.pone.0161771.g001Fig 1 Computational workflow followed to identify relevant miRNAs (green and orange blocks). Two selected miRNAs (i.e., miR-34a and miR-21) obtained from the computational analysis were then validated in wet lab experiments (blue boxes). Identification of the set of relevant pathways The proposed computational workflow starts from the identification of a set of relevant pathways behind key biological processes involved in the development of BOS. Several pathway repositories collect curated information on a large number of biological networks. These can be explored and analyzed for high-level systemic analysis [18, 19]. In this paper we resorted to the list of networks available in the KEGG database (http://www.genome.jp/kegg/pathway.html). KEGG is unique in its focus on and coverage of yeast, mouse, and human metabolic and signaling pathways, with a set of more than 200 networks for Homo sapiens [20]. To support the pathways identification process, we developed the automatic literature mining workflow, reported in Fig 2, whose code is available in the “1-literature-mining” folder of the BOS-miRNA-Enrichment repository on GitHub. 10.1371/journal.pone.0161771.g002Fig 2 Pathway selection process. The NCBI E-Utilities Application Programming Interface (API) and KEGG API were used to mine the available literature and to identify a preliminary set of relevant pathways connected to BOS pathogenesis. The workflow starts with a search of the available literature using NCBI E-utilities [21], with the three queries (Q1-Q3) reported in Fig 2. Each query contained associations of key terms related to BOS enabling the retrieval of a collection of relevant publications. This set of publications was further refined by including custom studies not directly identified by the three queries but known to be relevant to BOS. We used the NCBI E-utilities again on the new list to identify related genes annotated in the selected publications. Finally, starting from the list of genes identified, we used KEGG API to search for KEGG pathways annotated for the identified genes. The automatic literature mining process identified a list of 51 pathways (see the “Results” folder of the BOS-miRNA-Enrichment repository on GitHub) that were then manually reviewed and filtered. This step was important to remove false positive pathways that might have been introduced by the automatic literature mining process or to add pathways eluded in the automatic search. This resulted in the list of 39 pathways reported in the S1 Table. Each pathway was assigned to one out of three classes, depending on their assumed relevance in the pathogenesis of BOS. Class A identified highly relevant pathways behind very specific biological processes related to the pathogenesis of BOS. Class B identified pathways relevant in more general processes compared to class A, but still important in the development of the disease. Finally, class C pathways are very general pathways with reduced specificity in the development of BOS. Assignment to the different classes and review of the pathways list was carried out by three researchers expert in BOS pathogenesis: a clinician (FM), a pathologist (PM) and a genetist (ER). This information was necessary to weigh the contribution of any given pathway during computational analysis, in order to avoid bias due to low specificity of pathways within the targeted pathology. Pathway extension to include transcription factors Starting from the identified set of relevant pathways for BOS development, our computational workflow was used to search for possible interactions between miRNAs and genes involved in the different pathways. While direct interactions between miRNAs and their target genes can be retrieved from publicly available repositories, as explained in the sections that follow, it is well known that miRNA-mediated post-transcriptional regulation of selected genes can also act indirectly by translational repression or mRNA degradation of their transcription factors (TFs) [10][22][23][24]. Considering these indirect interactions is therefore very important when analyzing miRNA-pathway interactions. However, pathway repositories, such as KEGG, usually lack systematic information regarding the TF of each gene involved in a given regulatory network. To take into account TF-mediated miRNA regulations, we developed a php script able to process the KEGG Markup Language (KGML) description of a pathway and systematically integrate TF information for each described gene (see Fig 1). TF information is retrieved by connecting to the RESTful APIs provided by TargetMine [25]. For a given gene, the “Upstream Transcription Factors Template Query” (Targetmine) enables the retrieval of all upstream regulatory genes (TFs) from the AMADEUS [26][27] and ORegAnno [28] libraries of transcription factors-target gene relations. By applying this computational step, each pathway considered was extended to systematically include the TFs for all the genes involved. The source code used to process each pathway is available in the “2-Pathway-Processing” folder of the BOS-miRNA-Enrichment repository on GitHub. Identification of miRNA Our computational selection of candidate miRNAs involved in BOS development started from the full list of miRNAs annotated in miRBase (version 21 published on June 2014) [29]. MiRBase is a searchable database of published miRNA sequences. Version 21 of miRBase was exported in its SQL format and a list of 1870 miRNA precursor sequences for Homo sapiens species representing our initial list of miRNAs was extracted. In order to perform the miRNA target enrichment analysis, we first retrieved miRNA target information for each miRNA contained in our initial list. miRNA target information was retrieved by automatically mining the TargetHUB repository with a php script (see Fig 1). TargetHUB provides a programmer-friendly interface to aggregate multiple repositories of miRNA target genes with a uniform set of APIs [30]. The TargetHUB interface allows users to query information from four different databases: miRTarBase [31], TargetScan [32], PicTar [33], and miRanda [34]. TargetHUB uses miRBase Version 18 nomenclature. Because the nomenclature for miRNAs is not completely standard across different versions of miRBase, names of those miRNAs whose names in miRBase (version 21) match miRBase (version 18) remained unchanged. miRNAs that had no matching identifier were manually mapped resorting to the name change logs available in miRBase. In our study, we analyzed each miRNA considering two different sets of miRNA targets: Targets obtained by querying TargetHUB with results collected from the miRTarBase database referred to as Validated miRNA Targets (VMTs). MiRTarBase is known to collect information related to experimentally validated miRNA targets [35]. Targets obtained by querying TargetHUB for results present in at least one of the five available repositories. These are referred to as Computational miRNA Targets (CMTs) since they contain a large set of computationally-inferred information. VMTs were the main focus of this research and were used as the main source of information to identify the set of miRNAs to be tested in the laboratory. Nevertheless, CMTs were also of interest because they enlarged the scope of the analysis and allowed the formulation of exploratory predictions on miRNAs, to be tested in future experiments. In the remainder of this paper we will use the general term miRNA targets to identify both VMTs and CMTs. The set of scripts used to perform this computational step is available in the “3-Mirna-Processing” folder of the BOS-miRNA-Enrichment repository on GitHub. miRNAs enrichment analysis To identify relevant miRNAs in BOS development we performed enrichment analysis of miRNA gene targets in the selected pathways for each of the 1870 considered miRNAs. This analysis allowed us to rank miRNAs based on their overall enrichment in the selected pathways. This information was used to skim the initial list of miRNAs, identifying a reduced panel with potential higher involvement in the pathogenesis of BOS. Making use of a combination of php and R scripts, we constructed a 2x2 contingency matrix for each miRNA-pathway pair, as shown in Fig 3. Np+ represents the number of genes in the pathway targeted by the miRNA. It is obtained by intersecting information on miRNA targets with the list of genes and TFs for the pathway. Np- represents the number of non-targeted genes in the pathway, while Nm+ represents the total number of miRNA targets. Finally, Nm- represents the total number of genes not targeted by the miRNA. 10.1371/journal.pone.0161771.g003Fig 3 miRNA vs. pathway contingency table. Np+: number of targeted genes in the pathway, Np-: number of non-targeted genes in the pathway, Nm+: number of total targets of the miRNA, Nm-: number of total genes not targeted by the miRNA. To perform a fair analysis, Nm- was computed by merging the list of targets of all considered miRNAs and subtracting the targets of the miRNA under evaluation; and Np- was computed by excluding those genes in the pathway that had never been annotated as a miRNA target in the considered database. For each miRNA-pathway pair we applied a one-tailed Fisher's exact test to compute the enrichment of the miRNA targets in the selected pathway [36][37]. The one-tailed test for a 2x2 case computes p-values directly using the hypergeometric distribution. It provides exact probabilities and is appropriate for small number statistics [37]. By means of the Fisher's test we were able to test, for each miRNA-pathway pair, the hypothesis of a miRNA having its target genes enriched in a given pathway. Finally, significance levels computed for each miRNA when paired with each pathway were combined together in order to test the significance of the miRNA enrichment over the joint set of pathways. First, false discovery rate (FDR) was applied to the p-values for each miRNA-pathway pair, in order to correct for multiple-hypothesis testing, thus implementing the correction as previously described in [38]. Second, corrected p-values were combined by applying the modified Lancaster's method proposed in [39]. Lancaster's method is a meta-analysis algorithm, which can be used to combine the results of more than one test bearing upon the same hypothesis. Compared to the Fisher's combined probability test widely used in other enrichment studies to combine p-values [40], Lancaster's method enables the introduction of weighting functions for the combined p-values. Weighting functions allowed for the incorporation of prior biological information in the analysis leading to more meaningful results. In particular, in our study, p-values were weighted according to the relevance class of the related pathway, reported in the S1 Table. Class A pathways where weighted the highest (2), equal to the traditional Fisher's combined probability test [39]. Class B pathways were weighted 1 to account for reduced relevance in the analysis and, finally, class C pathways were weighted 0.5. Finally, the modified Lancaster's method previously proposed in [39] and applied in this study accounts for correlation among the single tests. This is important in our setup since pathways often share common sets of genes. Traditional Fisher's combined probability tests are based on the independence assumption, which does not hold in our setup. Based on the result of the combined p-value test, we were able to assign to each miRNA a single enrichment significance level. Only highly significant enriched miRNAs (p<0.01) were retained in the final reduced panel of miRNAs as candidates for laboratory analysis. As an example, Fig 4 reports a sample of the computations performed in the analysis of a single miRNA (miR-34a) using VMTs. The analysis starts from acquisition of the list of miRNA targets and ends with the computation of an enrichment score for the miRNA in the considered set of pathways. The source code for the scripts used to perform this analysis is available in the “4-Enrichment-Analisys” folder of the BOS-miRNA-Enrichment repository on GitHub. 10.1371/journal.pone.0161771.g004Fig 4 Example of step by step processing of a single miRNA using VMTs. (1) miR-34a, used as an example in the figure, is queried on TargetHUB to obtain its list of targets (VMTs); (2) The list of targets is intersected with the list of genes for each considered pathway to obtain a matrix showing how many targets belong to each pathway; (3) The matrix built during step 2, together with the total gene count for each pathway, is used to construct a set of contingency tables (one for each miRNA/pathway pair) as shown in Fig 3; (4) For each miRNA/pathway pair, the related contingency table is used to perform a one-tailed Fisher’s exact test to compute the enrichment of the miRNA targets in the pathway, obtaining a p-value that measures the significance of the test; and (5) p-values obtained for each miRNA/pathway pair are combined into a single significance score by the application of the modified Lancaster's method proposed in [39]. Final filtering Finally, the scored list of miRNAs was filtered according to two different criteria. First we selected the miRNAs identified as being expressed in relevant tissue samples. This step is important to focus on the miRNAs specifically active in the target tissue. To perform this operation, we consulted mimiRNA, a database of miRNA expression across different tissues and cell lines [41]. mimiRNA incorporates a sample classification algorithm that groups identical miRNA or mRNA experiments from separate sources. This enables mimiRNA to provide reliable expression profiles and to discover functional relations between miRNAs and mRNAs such as miRNA targets. The mimiRNA database for the selected tissues was downloaded and processed by means of a php script in order to extract a list of miRNAs significantly expressed in lung tissues and in the following cells involved in specific immunity and non-specific inflammation in BOS: Lung tissues; B lymphocytes; Natural killer lymphocytes; Monocytes; Dendritic cells; Granulocytes; and A549 (A549 cells derive from a peripheral well-differentiated lung adenocarcinomas, which retain several features of mature deep airway epithelium, and are therefore frequently used in experimental studies as a "model" of human bronchiolo-alveolar epithelium). The second step was to select intragenic miRNA co-expressed by genes or transcription factors in the selected pathways. To perform this operation, we consulted the miRIAD Intragenic Microrna database [42]. Each gene and transcription factor belonging to the considered pathways was automatically queried on miRIAD to identify if it potentially hosts an intragenic miRNA. Intragenic miRNAs selected in this step are likely to be co-expressed with their host genes [43] and therefore have a good chance of being expressed within biological processes involved in the pathogenesis of BOS. The full list of 349 miRNAs selected along with the tissue or cell in which they are expressed is reported in the S1 File, while the code for the script used to perform the filtering is available in the “3-Mirna-Processing” folder of the BOS-miRNA-Enrichment repository on GitHub. Wet lab experiments ISH for two candidate miRNAs (miR-34a and miR-21) was performed on formalin-fixed/paraffin-embedded samples of normal human and rat lungs, lung explants of BOS patients and rat models of acute and chronic lung rejection (outbred CD SPF/VAF), and on mesenchymal cells obtained from bronchoalveolar lavage (BAL) of lung recipients. The specimens were collected from the University of Padua and Foundation IRCSS San Matteo/University of Pavia Pathology Units. The study to develop an animal model of obliterans bronchiolitis was approved by the Research Ethics Committee of University of Padua (Protocol n. 0004959: Approval of informed consent from patients for tissue storage and research use. Date of approval: 27/01/2011. Project n. 23/2014: 244 Development of an animal model of obliterans bronchiolitis. Date of approval: 245 13/10/2014). All procedures used in this study were conformed to the rules and principles of the 2010/63/EU Directive. Lung explants of BOS patients were collected from IRCSS San Matteo/University of Pavia Pathology Units. The study was approved by Pavia Area Ethics Committee of Foundation IRCCS San Matteo (Protocol n. 20140003328: Isolation and characterization of mesenchymal cells obtained from bronchoalveolar lavage of lung recipients. Date of approval: 28/07/2014). Written informed consents of patients have been obtained before explant. Additional information regarding the Research Ethics Committee are provided in S2 File. miR-34a and miR-21 are double-DIG labeled miRCURY LNA®microRNA Detection Probes and have the sequences 5'-ACAACCAGCTAAGACACTGCCA-3', and 5'-TCAACATCAGTCTGATAAGCTA-3' respectively. U6 and a scramble probe were used respectively as positive and negative controls. All probes were purchased from Exiqon. ISH was performed according to the manufacturer's protocol provided by Exiqon (Exiqon, Vedbaek, Denmark) [44] with slight modifications described below. ISH experiments were performed on tissue sections and mesenchymal cells (MCs) isolated from BAL of BOS patients and cultured on Lab-Tek®II Chamber SlidesTM (Thermo Scientific) according to [45]. Normal skin fibroblasts (NHDF-cor; PromoCell GmbH, Heidelberg Germany) were used as control. Six micron thick unstained tissue sections were deparaffinized and digested with Proteinase-K (15 μg/mL) for 10' at 37°C using an Abbott Molecular StatSpin ThermoBrite Hybridizer System (Dako hybridizer). Hybridization with 40nM miR-21 probe, 1nM U6 probe and 20nM scramble probe was carried out at 50°C for 60 min. For miR-34a, a less-abundantly expressed miRNA, 100nM labeled probe was used and hybridization was carried out at 37°C for 20h. Furthermore, to prevent miR-34a release and extra-tissutal diffusion during hybridization, an additional miRNA fixation step, using 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide, was added on tissue sections to anchor miRNAs into the protein matrix [46]. Stringent washes were carried out in 5X SSC, 1X SSC and 0.2X SSC at 55°C. The digoxigenins were then recognized by a specific anti-DIG antibody, directly conjugated with the enzyme Alkaline Phosphatase and the cells were stained with NCIP/NBT chromogenic substrate. Images of miRNA signals were captured by an Olympus BX41 microscope equipped for conventional bright light and epifluorescence microscopy and connected to an image acquisition software (Cell F, Olympus). Consecutive unstained sections were stained with Movat pentachrome stain for connective tissue to highlight tissue morphology and connective tissue components, and immunoreacted with anti-actin HHF35 monoclonal antibody (Dako, Glostrup, DE) on the Dako Omnis autostainer platform to confirm the myofibroblast location of positive ISH signals. All reactions were analyzed under the light microscope and reviewed blindly by two skilled pathologists (PM and FC). Results Panel of miRNA involved in the pathogenesis of BOS Figs 5 and 6 report the main results obtained by the computational pipeline described in Fig 1. 10.1371/journal.pone.0161771.g005Fig 5 Ranked list of miRNAs most significantly enriched in BOS-related pathways (p<0.01) considering VMTs. Results are plotted using a -log(p-value). 10.1371/journal.pone.0161771.g006Fig 6 Ranked list of the top 30 miRNAs most significantly enriched in BOS-related pathways (p<0.01) considering CMTs. Results are plotted using a -log(p-value). The analysis performed with VMTs (Fig 5) identified a set of 29 miRNAs, out of the 349 miRNAs considered in the initial dataset, whose targets were significantly enriched (p-value<0.01) in the selected pathways. miRNAs were ranked according to the significance of their enrichment in the pathways. This result enabled us to narrow down considerably the set of candidate miRNAs to be investigated in the laboratory. The complete list of scores for the miRNAs considered is presented in the S1 File. Limiting the analysis to VMTs might have biased the results due to the limited amount of information available in the miRTarBase repository. To enlarge the exploratory scope of our study, we also submitted CMTs to analysis, and we were able to identify another ranked set of 174 miRNAs (Fig 6) whose targets (both computational and validated) were significantly enriched (p-value<0.01) in the set of selected pathways. The list produced in this case was, of course, larger than the one presented in Fig 5, since, unlike VMTs, CMTs are available for all considered miRNAs. This enabled us to enlarge the spectrum of possible interactions between miRNAs and pathways, and to account for miRNAs that might have escaped our attention. Nevertheless, the information obtained in this second analysis must be considered carefully. It is well known that computational methods for miRNA target analysis are not highly accurate and often generate a high number of false positives [47]. The complete list of scores for the considered miRNAs obtained with this second analysis is also available in the S1 File while Fig 6 shows the 30 top ranked miRNAs in this list. Finally, in order to complete the analysis, we performed a random control experiment by applying the computational pipeline to a random selection of 39 pathways from KEGG. Interestingly, the number of significantly enriched miRNA selected by the computational flow dropped to only 5 in the analysis performed with VMTs (Fig 7) and to only 2 miRNAs for the analysis performed with CMTs (Fig 8). This provides a good indication that the selected miRNAs depend on the considered set of BOS-related pathways. However, it is interesting to note that in both cases (VMTs and CMTs), the miRNAs identified in the random control were also reported in the list of miRNAs related to BOS. This result was not unexpected. First of all, the reader may note that the score obtained for this random control (-log(p-value)) is higher than the corresponding score assigned when considering BOS related pathways. This means reduced enrichment in the selected set of random pathways. Second, the presence of these miRNAs in the random control is probably an indication that they might be related to more general cell proliferation/differentiation mechanisms relevant in several disease processes. While they can still be found dysregulated in BOS patients (for this reason they have not been removed from the lists reported in Figs 5 and 6) they are less specific for the target disease. 10.1371/journal.pone.0161771.g007Fig 7 Ranked list of miRNAs highly significantly enriched in a set of 39 randomly selected pathways (p<0.01) considering VMTs. Results are plotted using a -log(p-value). 10.1371/journal.pone.0161771.g008Fig 8 Ranked list of miRNAs highly significantly enriched in a set of 39 randomly selected pathways (p<0.01) considering CMTs. Results are plotted using a -log(p-value). ISH results In order to validate the results of the computational analysis, we analyzed specific miRNA expression profiles in a set of lung tissue samples obtained from animals and patients with BOS and in mesenchymal cells from BAL of lung recipients, and compared these with results obtained from normal controls. Information regarding the Research Ethics Committee is provided in the S2 File. The ISH approach was preferred to qRT-PCR, since it not only provides evidence of local dysregulation but also allows morphological correlations and highlights the expression profiles in different cell types (inflammatory cells like macrophages or lymphocytes, normal or reactive epithelial cells, endothelial cells, and, finally, fibroblasts). Specifically, BO lesions involve a small proportion of lung tissue and therefore the quantitative approach could miss significant differences in expression levels occurring in limited, although extremely disease-specific, tissue components. To choose the first two miRNAs to be validated by wet-lab experiments we firstly focused on those factors present in both lists shown in Figs 5 and 6, but not identified in the random control set of miRNAs. The factors present in both lists were: let-7a, miR-34a, miR-21 and miR-9 family. The let-7a and miR-9 miRNAs were also among those identified in the random control set, and we excluded them as their dysregulation might be related to a more general cell proliferation/differentiation mechanism relevant in several disease processes, but not specifically related to BOS. Moreover, miR-21 and miR-34a had not yet been described in BOS in previous human studies. Given these premises, we selected miR-34a, the top-ranked miRNA in Fig 5 (VMTs) and miR-21, as candidate miRNAs for ISH analysis. Our preliminary results documented the dysregulated expression of miR-34a and miR-21 in human and rat transplanted lungs with BOS (Table 1). Specifically, ISH analysis showed that in normal human and rat lungs, miR-34a was diffusely expressed in bronchial and alveolar epithelial cells and in some inflammatory cells, mostly plasma cells (Fig 9A). In lung explants from BOS patients, miR-34a was strongly expressed in bronchiolar and reactive alveolar cells and moderate expression was also detectable in proliferating fibroblasts of BO lesions (Fig 9B). 10.1371/journal.pone.0161771.g009Fig 9 miR-34a and miR-21 expression in human and rat transplanted lungs. Upper panel: miR-34a expression in bronchial epithelial cells in normal human lung (A, blue staining), in myofibroblasts in lung explants from a BOS patient (B), in proliferating fibroblasts in rat lungs with chronic rejection (C) and in epithelial cells in remodeled areas in the animal model of acute cellular rejection (D). Lower panel: miR-21 ISH (E) and Movat pentachrome stains (F) highlighting miR-21 expression in BO myofibroblasts in lung explants from BOS patients. The airway lumen, which can be recognized by the peripheral elastic fibers, is totally occluded by collagen deposition and myofibroblasts (F), which were strongly positive for miR-21 (E, blue staining). In rat lungs with chronic rejection, miR-21 expression was similarly observed in BO myofibroblasts (G), while in acute cellular rejection, miR-21 expression was localized in epithelia and in interstitial fibroblasts associated with inflammatory infiltrates (H). 10.1371/journal.pone.0161771.t001Table 1 Profile and expression analysis of miR-34a and miR-21 in human and rat lung samples: normal lung, human chronic rejection (BOS), acute and chronic rat graft rejection. Human CTR(1) ACUTE REJECTION(1) BOS(5) miR-34a Bronchial and // BO fibroblasts (+++), alveolar epithelia (++) // Reactive pneumocytes (++), // Bronchial and alveolar epithelia (++) miR-21 Absent // BO fibroblasts (+++), // Reactive pneumocytes (++) Rat CTR(2) ACUTE REJECTION(1) CHRONIC REJECTION(1) miR-34a Bronchial and Bronchial and Reactive pneumocytes (++), alveolar epithelia (++) alveolar epithelia (++), Fibroblasts (++), Inflammatory infiltrates (++) Bronchial and alveolar epithelia (++), Endothelia (++) miR-21 Absent Reactive pneumocytes (+), Reactive pneumocytes (++), Fibroblasts (++), Fibroblasts (+++) Inflammatory infiltrates (+) BO: bronchiolitis obliterans; (+): scarce expression; (++): moderate expression; (+++): strong expression. Analogously, in the rat orthotopic lung transplantation with chronic bronchiolar rejection, miR-34a was strongly expressed in proliferating fibroblasts (Fig 9C), bronchial epithelial cells, endothelia and inflammatory cells, mostly plasma cells, while in the setting of acute cellular rejection, miR-34a expression was detectable only in epithelial cells and in inflammatory infiltrates (Fig 9D). As far as miR-21 was concerned, human and rat lungs with chronic bronchiolar rejection showed strong expression in the fibroblasts of bronchiolar lesions (Fig 9E and 9G) and in epithelial cells in remodeled areas, while miR-21 expression was not detectable in either human or animal normal lungs. In the acute rejection animal model, miR-21 was modestly expressed in epithelia and in interstitial fibroblasts associated with inflammatory infiltrates (Fig 9H). All observations of miR-34a and miR-21 expression were validated with alpha 1-actin and Movat (Fig 9F) connective tissue stains on consecutive sections. ISH results on mesenchymal cells from BAL confirmed the dysregulation observed in tissue samples. Specifically, miR-21 was not expressed in normal fibroblasts, while it was modestly expressed in BOS MCs. miR-34a expression was evident both in control cells and BOS MCs (data not shown). Discussion and Conclusions In this paper we applied for the first time a complex computational pipeline to perform enrichment analysis of miRNA targets in pathways involved in BOS pathogenesis. The analysis, which started with literature mining to identify a limited set of relevant pathways and integrated information from several public databases, ultimately produced a ranked panel of miRNAs potentially involved in BOS pathogenic process. The analysis considered the full set of miRNAs annotated in miRBase (version 21), and applied a sequence of filtering approaches and statistical analyses to reduce this set and rank the identified miRNAs. The full computational pipeline consulted only publicly available miRNA interaction data, and did not require any expression data to perform its prediction. It was therefore well suited to perform a preliminary screening before conducting actual laboratory experiments. Two different lists of miRNAs were produced: a first, smaller one, based on VMTs; and a second, larger one, based on CMTs. We proceeded to validate the analysis of the ranked lists of miRNAs by means of wet lab experiments (ISH) on two highly-ranked miRNAs that appeared in both lists: miR-34a and miR-21. As stated above, they were chosen among those not yet described in the most recent literature on BOS; and because of their presence in both lists (validated and computational factors). Fig 10 shows the p-values obtained from the enrichment analysis of these two miRNAs against all considered pathways. Subfigure A reports data computed with VMTs while subfigure B reports data computed with CMTs. These p-values have been used to compute the scores used to rank the miRNAs in Figs 5 and 6. 10.1371/journal.pone.0161771.g010Fig 10 P-values obtained from the miRNA targets enrichment analysis for miR-34a and miR-21 paired with all considered pathways. The color scale indicates the significance of the enrichment (green = significant enrichment). A: analysis with VMTs; B: analysis with CMTs. The first factor chosen for the validation phase was miR-34a, which ranked first in the VMTs list (Fig 5) and eleventh in the CMTs list (Fig 6). It is an intergenic factor that has been reported as a tumor suppressing gene, being consistently downregulated (mainly by promoter hypermethylation) in a variety of cancer cells (breast, prostate, bladder, pancreas, kidney and colon carcinoma, glioblastoma, myeloma, melanoma, and neuroblastoma) [48][49][50]. We were able to document diffuse expression of miR-34a in bronchial and alveolar epithelial cells, as well as in endothelial and inflammatory cells of normal adult lungs. In lung samples obtained from BOS patients, miR-34a was clearly expressed in normal bronchiolar, alveolar epithelia and reactive pneumocytes, but it was also expressed in proliferating fibroblasts, suggesting that its dysregulated expression might play a role in the fibrogenic process of BOS. Published evidence of a profibrogenic role of miR-34a and of its expression by proliferating fibroblasts is however very limited. It has been reported as being upregulated in macrophages in the bleomycin model of pulmonary fibrosis [51] and it has been shown to play a critical role in the progression of cardiac tissue fibrosis, mainly by targeting the TGF-beta/SMAD signal transduction pathway [52]. Recently, miR-34a and miR-34c have been shown able to down regulate peroxisome proliferator-activated receptor (PPAR) in hepatic stellate cells thus inducing their activation, a hallmark of liver fibrosis [53]. PPAR has, on the other hand, been recognized as a relevant anti-fibrotic mediator (mainly by inducing the inhibition of TGF-beta/SMAD signal transduction pathway), thus its down regulation might be important in driving fibroblast proliferation and extracellular matrix deposition, as recently suggested in an investigation of systemic sclerosis-associated skin and lung fibrosis [54]. Further data are required to clarify the molecular mechanisms underlying the pro-fibrotic role of miR-34a in BOS that we suggest in this study. The second factor, miR-21 ranked twenty-second in the VMTs list (Fig 5) and fifteenth in the CMTs list (Fig 6). Interestingly, our ISH results revealed a net overexpression of miR-21 in BO lesions: it was primarily expressed in fibroblasts and in activated epithelial cells in all human BOS cases and in rat grafts, while it was completely absent in normal human and rat lungs, thus clearly demonstrating upregulation in BO. Literature data in support of role of miR-21 in fibrogenesis are more consistent. It was found to be significantly upregulated in the lung tissue of animal models of bleomycin-induced pulmonary fibrosis [55]; and upregulated 2.7-fold by RT PCR in a mouse model of orthotopic tracheal transplantation [4]. In addition, miR-21 upregulation has been demonstrated in fibroblast foci as well as in the serum of patients with idiopathic pulmonary fibrosis, with a significant correlation between serum levels and the degree of lung function impairment [56]. Besides its involvement in fibro-proliferative processes in several organs [54][56][57][58][59][60][61], miR-21 has been shown to play an important role in several other physiological and pathological processes [62]. Its up-regulation has been observed in solid tumors (breast, colon, lung, pancreas, prostate and stomach, ovarian, cervical, head and neck carcinomas), in leukemias and in a variety of other human proliferative disorders, implying a function in regulating cell growth [62]. The mature miR-21 sequence is strongly conserved throughout evolution and is encoded by a single intergenic gene located on the plus strand of chromosome 17q23.2, where it overlaps with the protein-coding gene VMP1. Its transcription is under the control of AP-1, SRF, p53, Signal Transducer and Activator of Transcription 3 (STAT3) and many other TFs, as well as of epigenetic mechanisms [62]. It is also controlled at the post-transcriptional level by: TGF-beta and BMP via SMAD1/5 and SMAD2/3. Since the latter signaling pathway is important in epithelial-mesenchymal transition (EMT), it is reasonable to think that miR-21 plays a role in this process [62]. Moreover, hypoxia, which is a well-recognized stimulus for EMT [63], is a potent miR-21 inducer. Indeed, an HIF-1alpha binding site is present in the pri-miR-21 promoter [64]. miR-21 targets several factors responsible for cell growth and invasion suppression, cell cycle arrest, MMP and other protease inhibition, control of angiogenesis, cellular branching and migration, with a variety of regulatory feedback loops [62]. Their down-regulation has been reported in several cancers in which miR-21 is overexpressed [62]. Finally, miR-21 dysregulation has been demonstrated in bio-fluids [65][66] and anti-miR-21 has been used as targeted therapy in cancer, obtaining promising results both in vitro and in vivo [67,68]. On the basis of our novel finding and of previous evidence of the role of miR-21 in homeostasis and disease, its contribution to BOS pathogenesis can be inferred to be at the level of myofibroblast migration and proliferation in the airway lumen, while a possible additional regulatory activity in the mechanism of epithelial-mesenchymal transition cannot be excluded. Overall, this is the first study to systematically analyze and score a large number of miRNAs implicated in BOS. We were able to process all the miRNAs currently annotated in miRBase (version 21) and, by means of a sequence of computational filtering methods, we were able to filter them and then rank them according to their enrichment into the selected pathways. In addition to miR-34a and miR-21, other miRNAs have recently been described as being dysregulated in BOS. Two recent papers reported an integrated analysis of miRNA expression in the mouse models of orthotopic tracheal transplantation [4] [5]. The following miRNAs, also present in the VTMs list of Fig 5, were found by RT-PCR to be dysregulated in mouse lung tissue: miR-21, miR-146, miR-20, miR-302, miR-19, miR-98, let-7a, miR-15a. In comparison with lung recipients without BOS, clear dysregulation of miR-34a, miR-193b, miR-9 and miR-15a, likewise present in the VTM list in Fig 5, was also detected in peripheral mononuclear cells obtained from BOS patients in a RT-PCR evaluation of miRNA expression by Xu at al. [3]. Even if the above results were obtained on the one hand in animals by RT-PCR without a comprehensive analysis of cell type-restricted expression; and on the other in humans but on a specific peripheral inflammatory cell subset, without further confirmation on lung tissue, they might support the validation of our pipeline. Further analyses are, however, necessary to confirm the role of these factors in the pathogenesis of BOS. The main limitation of our study is that only two factors identified by our pipeline underwent validation. However, since our wet-lab experiments were intended only to confirm the results obtained with the computational approach, more extensive tissue analysis was beyond the scope of the study. In addition, we chose 2 miRNAs which had never been associated with BO in previous human studies. For these reasons, while we did demonstrate miR-34a and miR-21 dysregulation in fibroblasts obliterating the bronchiolar lumen, we cannot provide mechanistic insights into their role in BOS pathogenesis, and these should be addressed in specifically designed studies. Nevertheless, the data depicted in Fig 11 are a preliminary step in this direction. This shows a regulatory network depicting the interactions between miRNAs identified in the VMTs list in Fig 5 and all genes identified during the initial literature search as being involved in BOS (see S1 Table). The network was constructed using the CyTransfinder Cytoscape plugin [69]. Interestingly, 17 out of the 29 identified miRNAs were involved in the regulation of 14 out of the 27 identified genes. Even more interestingly, none of the identified relations involves a direct interaction between a miRNA and a target gene, but all are mediated through one transcription factor introduced while processing the considered pathways. All miRNAs and genes not included in this network involve longer chains of regulation that we could not easily reconstruct through the CyTransfinder tool and require further investigation. 10.1371/journal.pone.0161771.g011Fig 11 Gene regulatory network showing interaction between miRNAs identified in the VMTs list and relevant genes implicated in BOS as identified from the literature (see the S1 Table). The network is hierarchically organized into four levels. Brown squares: miRNA host genes for intragenic miRNAs; red and purple rhombi: intragenic and intergenic miRNAs respectively; cyan squares TFs; green hexagons: target genes. Supporting Information S1 File This file contains the final significance scores (VMTs and CMTs) computed for the analyzed miRNAs including the tissue or cell in which they are expressed. (XLSX) Click here for additional data file. S2 File This file contains research ethics committee information regarding the experiments described in the paper. (PDF) Click here for additional data file. S1 Table List of analyzed pathways. For each pathway the table reports: the unique KEGG Identifier, the pathway name, the assigned class based on its relevance, the relevant genes and the related references that identify the pathway as relevant for BOS. (PDF) Click here for additional data file. ==== Refs References 1 Estenne M , Hertz MI . Bronchiolitis obliterans after human lung transplantation . Am J Respir Crit Care Med . 2002 8 ;166 (4 ):440 –4 . 12186817 2 Todd JL , Palmer SM . Bronchiolitis obliterans syndrome: the final frontier for lung transplantation . Chest . 2011 8 ;140 (2 ):502 –8 . 10.1378/chest.10-2838 21813529 3 Xu Z , Nayak D , Yang W , Baskaran G , Ramachandran S , Sarma N , et al Dysregulated MicroRNA Expression and Chronic Lung Allograft Rejection in Recipients With Antibodies to Donor HLA . Am J Transplant . 2015 7 ;15 (7 ):1933 –47 . 10.1111/ajt.13185 25649290 4 Dong M , Wang X , Zhao HL , Chen XL , Yuan JH , Guo JY , et al Integrated analysis of transcription factor, microRNA and LncRNA in an animal model of obliterative bronchiolitis . Int J Clin Exp Pathol . 2015 ;8 (6 ):7050 –8 . 26261598 5 Wang J , Cao H , Hong X , Chen GH , Fan HM , Li QC , et al MicroRNA screening and functional study of obliterative bronchiolitis in a rat model simulating lung transplantation . Genet Mol Res . 2015 ;14 (4 ):19309 –16 . 10.4238/2015.December.29.40 26782583 6 Filipowicz W , Bhattacharyya SN , Sonenberg N . Mechanisms of post-transcriptional regulation by microRNAs: are the answers in sight? Nat Rev Genet . 2008 2 ;9 (2 ):102 –14 . 10.1038/nrg2290 18197166 7 Stefani G , Slack FJ . Small non-coding RNAs in animal development . Nat Rev Mol Cell Biol . 2008 3 ;9 (3 ):219 –30 . 10.1038/nrm2347 18270516 8 Ambros V . The functions of animal microRNAs . Nature . 2004 9 ;431 (7006 ):350 –5 . 15372042 9 Bartel DP . MicroRNAs: genomics, biogenesis, mechanism, and function . Cell . 2004 1 ;116 (2 ):281 –97 . 14744438 10 Di Carlo S , Politano G , Savino A , Benso A . A systematic analysis of a mi-RNA inter-pathway regulatory motif . J Clin Bioinforma . 2013 ;3 (1 ):20 10.1186/2043-9113-3-20 24152805 11 Dai R , Ahmed SA . MicroRNA, a new paradigm for understanding immunoregulation, inflammation, and autoimmune diseases . Transl Res . 2011 4 ;157 (4 ):163 –79 . 10.1016/j.trsl.2011.01.007 21420027 12 Farooqi AA , Qureshi MZ , Coskunpinar E , Naqvi SKUH, Yaylim I, Ismail M. MiR-421, miR-155 and miR-650: emerging trends of regulation of cancer and apoptosis . Asian Pac J Cancer Prev . 2014 ;15 (5 ):1909 –12 . 24716910 13 Heneghan HM , Miller N , Kerin MJ . MiRNAs as biomarkers and therapeutic targets in cancer . Curr Opin Pharmacol . 2010 10 ;10 (5 ):543 –50 . 10.1016/j.coph.2010.05.010 20541466 14 Zampetaki A , Mayr M . MicroRNAs in vascular and metabolic disease . Circ Res . 2012 2 ;110 (3 ):508 –22 . 10.1161/CIRCRESAHA.111.247445 22302757 15 Wilflingseder J , Reindl-Schwaighofer R , Sunzenauer J , Kainz A , Heinzel A , Mayer B , et al MicroRNAs in kidney transplantation . Nephrol Dial Transplant . 2015 6 ;30 (6 ):910 –7 . 10.1093/ndt/gfu280 25170095 16 Subramanian A , Tamayo P , Mootha VK , Mukherjee S , Ebert BL , Gillette MA , et al Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles . Proceedings of the National Academy of Sciences . 2005 ;102 (43 ):15545 –15550 . 17 Steinfeld I , Navon R , Ach R , Yakhini Z . miRNA target enrichment analysis reveals directly active miRNAs in health and disease . Nucleic Acids Research . 2012 ; Available from: http://nar.oxfordjournals.org/content/early/2012/12/02/nar.gks1142.abstract. 18 Kelder T , van Iersel MP , Hanspers K , Kutmon M , Conklin BR , Evelo CT , et al WikiPathways: building research communities on biological pathways . Nucleic Acids Res . 2012 1 ;40 (Database issue):D1301 –7 . 10.1093/nar/gkr1074 22096230 19 Kanehisa M , Goto S . KEGG: kyoto encyclopedia of genes and genomes . Nucleic Acids Res . 2000 1 ;28 (1 ):27 –30 . 10592173 20 Nagasaki M , Saito A , Doi A , Matsuno H , Miyano S . Foundations of systems biology: using Cell Illustrator and pathway databases vol. 13 Springer Science & Business Media ; 2009 . 21 Sayers E. "A general introduction to the E-utilities." (2010) Available from: http://www.ncbi.nlm.nih.gov/books/NBK25497/. 22 Hsieh WT , Tzeng KR , Ciou JS , Tsai JJ , Kurubanjerdjit N , Huang CH , et al Transcription factor and microRNA-regulated network motifs for cancer and signal transduction networks . BMC Syst Biol . 2015 ;9 Suppl 1 :S5 10.1186/1752-0509-9-S1-S5 25707690 23 Riba A , Bosia C , El Baroudi M , Ollino L , Caselle M . A combination of transcriptional and microRNA regulation improves the stability of the relative concentrations of target genes . PLoS Comput Biol . 2014 2 ;10 (2 ):e1003490 10.1371/journal.pcbi.1003490 24586138 24 Politano G , Benso A , Savino A , Di Carlo S . ReNE: a cytoscape plugin for regulatory network enhancement . PLoS One . 2014 ;9 (12 ):e115585 10.1371/journal.pone.0115585 25541727 25 Chen YA , Tripathi LP , Mizuguchi K . TargetMine, an integrated data warehouse for candidate gene prioritisation and target discovery . PLoS One . 2011 ;6 (3 ):e17844 10.1371/journal.pone.0017844 21408081 26 Linhart C , Halperin Y , Shamir R . Transcription factor and microRNA motif discovery: the Amadeus platform and a compendium of metazoan target sets . Genome research . 2008 ;18 (7 ):1180 –1189 . 10.1101/gr.076117.108 18411406 27 Orenstein Y , Linhart C , Shamir R . Assessment of Algorithms for Inferring Positional Weight Matrix Motifs of Transcription Factor Binding Sites Using Protein Binding Microarray Data . PLoS ONE . 2012 9 ;7 (9 ):e46145 Available from: doi: http://dx.doi.org/10.1371%2Fjournal.pone.0046145 23029415 28 Montgomery S , Griffith OL , Sleumer MC , Bergman CM , Bilenky M , Pleasance E , et al ORegAnno: an open access database and curation system for literature-derived promoters, transcription factor binding sites and regulatory variation . Bioinformatics . 2006 ;22 (5 ):637 –640 . 16397004 29 Kozomara A , Griffiths-Jones S . miRBase: integrating microRNA annotation and deep-sequencing data . Nucleic Acids Res . 2011 1 ;39 (Database issue):D152 –7 . 10.1093/nar/gkq1027 21037258 30 Manyam G , Ivan C , Calin GA , Coombes KR . targetHub: a programmable interface for miRNA-gene interactions . Bioinformatics . 2013 10 ;29 (20 ):2657 –8 . 10.1093/bioinformatics/btt439 24013925 31 Hsu SD , Lin FM , Wu WY , Liang C , Huang WC , Chan WL , et al miRTarBase: a database curates experimentally validated microRNA-target interactions . Nucleic Acids Res . 2011 1 ;39 (Database issue):D163 –9 . 10.1093/nar/gkq1107 21071411 32 Lewis BP , Burge CB , Bartel DP . Conserved seed pairing, often flanked by adenosines, indicates that thousands of human genes are microRNA targets . Cell . 2005 1 ;120 (1 ):15 –20 . 15652477 33 Krek A , Gr•un D , Poy MN , Wolf R , Rosenberg L , Epstein EJ , et al Combinatorial microRNA target predictions . Nat Genet . 2005 5 ;37 (5 ):495 –500 . 15806104 34 John B , Enright AJ , Aravin A , Tuschl T , Sander C , Marks DS . Human MicroRNA targets . PLoS Biol . 2004 11 ;2 (11 ):e363 15502875 35 Chou CH , Chang NW , Shrestha S , Hsu SD , Lin YL , Lee WH , et al miRTarBase 2016: updates to the experimentally validated miRNA-target interactions database . Nucleic Acids Research . 2015 ;Available from: http://nar.oxfordjournals.org/content/early/2015/11/19/nar.gkv1258.abstract. 36 Beissbarth T , Speed TP . GOstat: find statistically overrepresented Gene Ontologies within a group of genes . Bioinformatics . 2004 6 ;20 (9 ):1464 –5 . 14962934 37 Rivals I , Personnaz L , Taing L , Potier MC . Enrichment or depletion of a GO category within a class of genes: which test? Bioinformatics . 2007 ;23 (4 ):401 –407 . 17182697 38 Benjamini Y , Hochberg Y . Controlling the false discovery rate: a practical and powerful approach to multiple testing . Journal of the Royal Statistical Society Series B (Methodological) . 1995 ;p. 289 –300 . 39 Dai H , Leeder JS , Cui Y . A modified generalized Fisher method for combining probabilities from dependent tests . Front Genet . 2014 ;5 :32 10.3389/fgene.2014.00032 24600471 40 Littell RC , Folks JL . Asymptotic optimality of Fisher's method of combining independent tests . Journal of the American Statistical Association . 1971 ;66 (336 ):802 –806 . 41 Ritchie W , Flamant S , Rasko JEJ . mimiRNA: a microRNA expression profiler and classification resource designed to identify functional correlations between microRNAs and their targets . Bioinformatics . 2010 1 ;26 (2 ):223 –7 . 10.1093/bioinformatics/btp649 19933167 42 Hinske LC , Franca GS , Torres HAM , Ohara DT , Lopes-Ramos CM , Heyn J , et al miRIAD-integrating microRNA inter- and intragenic data . Database (Oxford) . 2014 ;2014 . 43 Rodriguez A , Griffiths-Jones S , Ashurst JL , Bradley A . Identification of mammalian microRNA host genes and transcription units . Genome Res . 2004 10 ;14 (10A ):1902 –10 . 15364901 44 J rgensen S , Baker A , M ller S , Nielsen BS . Robust one-day in situ hybridization protocol for detection of microRNAs in paraffin samples using LNA probes . Methods . 2010 12 ;52 (4 ):375 –81 . 10.1016/j.ymeth.2010.07.002 20621190 45 Cova E , Colombo M , Inghilleri S , Morosini M , Miserere S , Pe~naranda-Avila J , et al Antibody-engineered nanoparticles selectively inhibit mesenchymal cells isolated from patients with chronic lung allograft dysfunction . Nanomedicine . 2015 ;10 (1 ):9 –23 . 10.2217/nnm.13.208 24559038 46 Pena JTG , Sohn-Lee C , Rouhanifard SH , Ludwig J , Hafner M , Mihailovic A , et al miRNA in situ hybridization in formaldehyde and EDC-fixed tissues . Nat Methods . 2009 2 ;6 (2 ):139 –41 . 10.1038/nmeth.1294 19137005 47 Hammell M . Computational methods to identify miRNA targets . Semin Cell Dev Biol . 2010 9 ;21 (7 ):738 –44 . 10.1016/j.semcdb.2010.01.004 20079866 48 Li XJ , Ren ZJ , Tang JH . MicroRNA-34a: a potential therapeutic target in human cancer . Cell Death Dis . 2014 ;5 :e1327 10.1038/cddis.2014.270 25032850 49 Lodygin D , Tarasov V , Epanchintsev A , Berking C , Knyazeva T , Korner H , et al Inactivation of miR-34a by aberrant CpG methylation in multiple types of cancer . Cell Cycle . 2008 8 ;7 (16 ):2591 –600 . 18719384 50 Qiao P , Li G , Bi W , Yang L , Yao L , Wu D . microRNA-34a inhibits epithelial mesenchymal transition in human cholangiocarcinoma by targeting Smad4 through transforming growth factor-beta/Smad pathway . BMC Cancer . 2015 ;15 :469 10.1186/s12885-015-1359-x 26077733 51 Honeyman L , Bazett M , Tomko TG , Haston CK . MicroRNA profiling implicates the insulin-like growth factor pathway in bleomycin-induced pulmonary fibrosis in mice . Fibrogenesis Tissue Repair . 2013 ;6 (1 ):16 10.1186/1755-1536-6-16 23987664 52 Huang Y , Qi Y , Du JQ , Zhang Df . MicroRNA-34a regulates cardiac fibrosis after myocardial infarction by targeting Smad4 . Expert Opin Ther Targets . 2014 12 ;18 (12 ):1355 –65 . 10.1517/14728222.2014.961424 25322725 53 Li X , Chen Y , Wu S , He J , Lou L , Ye W , et al microRNA-34a and microRNA-34c promote the activation of human hepatic stellate cells by targeting peroxisome proliferator-activated receptor . Mol Med Rep . 2015 2 ;11 (2 ):1017 –24 . 10.3892/mmr.2014.2846 25370690 54 Dantas AT , Pereira MC , de Melo Rego MJB , da Rocha LF Jr, Pitta IdR , Marques CDL , et al The Role of PPAR Gamma in Systemic Sclerosis . PPAR Res . 2015 ;2015 :124624 10.1155/2015/124624 26064084 55 Lino Cardenas CL , Kaminski N , Kass DJ . Micromanaging microRNAs: using murine models to study microRNAs in lung fibrosis . Drug Discov Today Dis Models . 2013 ;10 (3 ):e145 –e151 . 25328532 56 Pandit KV , Milosevic J , Kaminski N . MicroRNAs in idiopathic pulmonary fibrosis . Transl Res . 2011 4 ;157 (4 ):191 –9 . 10.1016/j.trsl.2011.01.012 21420029 57 Cavarretta E , Condorelli G . miR-21 and cardiac fibrosis: another brick in the wall? Eur Heart J . 2015 5 ;. 58 O'Reilly S . MicroRNAs in fibrosis: opportunities and challenges . Arthritis Res Ther . 2016 ;18 (1 ):11 .26762516 59 Zanotti S , Gibertini S , Curcio M , Savadori P , Pasanisi B , Morandi L , et al Opposing roles of miR-21 and miR-29 in the progression of fibrosis in Duchenne muscular dystrophy . Biochim Biophys Acta . 2015 7 ;1852 (7 ):1451 –64 . 10.1016/j.bbadis.2015.04.013 25892183 60 Bd Tong , Xiao MY , Zeng JX , Xiong W. MiRNA-21 promotes fibrosis in orbital fibroblasts from thyroid-associated ophthalmopathy . Mol Vis . 2015 ;21 :324 –34 . 25873777 61 Chung ACK , Lan HY . MicroRNAs in renal fibrosis . Front Physiol . 2015 ;6 :50 10.3389/fphys.2015.00050 25750628 62 Krichevsky AM , Gabriely G . miR-21: a small multi-faceted RNA . J Cell Mol Med . 2009 1 ;13 (1 ):39 –53 . 10.1111/j.1582-4934.2008.00556.x 19175699 63 Zhao F , Malm SW , Hinchman AN , Li H , Beeks CG , Klimecki WT . Arsenite-induced pseudo-hypoxia results in loss of anchorage-dependent growth in BEAS-2B pulmonary epithelial cells . PLoS One . 2014 ;9 (12 ):e114549 10.1371/journal.pone.0114549 25513814 64 Xu X , Kriegel AJ , Liu Y , Usa K , Mladinov D , Liu H , et al Delayed ischemic preconditioning contributes to renal protection by upregulation of miR-21 . Kidney Int . 2012 12 ;82 (11 ):1167 –75 . 10.1038/ki.2012.241 22785173 65 Xie Y , Todd NW , Liu Z , Zhan M , Fang H , Peng H , et al Altered miRNA expression in sputum for diagnosis of non-small cell lung cancer . Lung Cancer . 2010 2 ;67 (2 ):170 –6 . 10.1016/j.lungcan.2009.04.004 19446359 66 Li Y , Li W , Ouyang Q , Hu S , Tang J . Detection of lung cancer with blood microRNA-21 expression levels in Chinese population . Oncol Lett . 2011 9 ;2 (5 ):991 –994 . 22866162 67 Devulapally R , Sekar NM , Sekar TV , Foygel K , Massoud TF , Willmann JK , et al Polymer nanoparticles mediated codelivery of antimiR-10b and antimiR-21 for achieving triple negative breast cancer therapy . ACS Nano . 2015 3 ;9 (3 ):2290 –302 . 10.1021/nn507465d 25652012 68 Seca H , Lima RT , Lopes-Rodrigues V , Guimaraes JE , Almeida GM , Vasconcelos MH . Targeting miR-21 induces autophagy and chemosensitivity of leukemia cells . Curr Drug Targets . 2013 9 ;14 (10 ):1135 –43 . 23834154 69 Politano G , Orso F , Raimo M , Benso A , Savino A , Taverna D , et al CyTRANSFINDER: a Cytoscape 3.3 plugin for three-component (TF, gene, miRNA) Signal Transduction Pathway construction . BMC Bioinformatics . 2016 ;.

PMC005xxxxxx/PMC5001702.txt

==== Front PLoS CurrPLoS CurrPLoS CurrplosPLoS Currents2157-3999Public Library of Science San Francisco, USA 2761716910.1371/currents.outbreaks.13b4e5e36a5c0831a1663fbdb5713fe9Research ArticleGenetic Studies of Vibrio cholerae in South West Cameroon—A Phylogenetic Analysis of Isolates from the 2010-2011 Epidemic Ngwa Moise C. Emerging Pathogens Institute, University of Florida, Gainesville, Florida, USA; Department of Environmental and Global Health, College of Public Health and Health Professions, University of Florida, Gainesville, Florida, USA*Masalla Thomas Laboratory for Emerging Infectious Diseases, University of Buea, Buea, South West Region, CameroonEsemu Seraphine Laboratory for Emerging Infectious Diseases, University of Buea, Buea, South West Region, CameroonFumoloh Foche Francis Laboratory for Emerging Infectious Diseases, University of Buea, Buea, South West Region, Cameroon*Kracalik Ian Cella Eleonora Emerging Pathogens Institute, University of Florida, Gainesville, Florida, USAAlam Meer Taifur Emerging Pathogens Institute, University of Florida, Gainesville, Florida, USAAkoachere Jane-Francis Laboratory for Emerging Infectious Diseases, University of Buea, Buea, South West Region, CameroonLiang Song Emerging Pathogens Institute, University of Florida, Gainesville, Florida, USA; Department of Environmental and Global Health, College of Public Health and Health Professions, University of Florida, Gainesville, Florida, USASalemi Marco Emerging Pathogens Institute, University of Florida, Gainesville, Florida, USAMorris J. Glenn Emerging Pathogens Institute, University of Florida, Gainesville, Florida, USA; Department of Medicine, College of Medicine, University of Florida, Gainesville, Florida, USAAli Afsar Emerging Pathogens Institute, University of Florida, Gainesville, Florida, USA; Department of Environmental and Global Health, College of Public Health and Health Professions, University of Florida, Gainesville, Florida, USANdip Lucy M. Laboratory for Emerging Infectious Diseases, University of Buea, Buea, South West Region, Cameroon; Department of Microbiology, University of Buea, Buea, South West Region, Cameroon12 8 2016 8 ecurrents.outbreaks.13b4e5e36a5c0831a1663fbdb5713fe9© 2016 Ngwa, Masalla, Esemu, Fumoloh, Kracalik, Cella, Alam, Akoachere, Liang, Salemi, Morris, Ali, Ndip, et al2016Ngwa, Masalla, Esemu, Fumoloh, Kracalik, Cella, Alam, Akoachere, Liang, Salemi, Morris, Ali, Ndip, et alThis is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited. Introduction: During the cholera outbreak from 2010 to 2011 in Cameroon, 33,192 cases with 1,440 deaths (case fatality ratio 4.34%) were reported to the World Health Organization. Of these, the South West Region reported 3,120 clinical cases. This region is in the Equatorial Monsoon climatic subzone of Cameroon, close to the coast, raising questions as to whether cases were linked with development of environmental reservoirs. Methods: In an investigation conducted by the Laboratory for Emerging Infectious Diseases, University of Buea, toxigenic V. cholerae O1 were isolated from diarrheal stool samples from 18 patients, with ages ranging from <3 to 70 years. Coordinates for clinical centers at which cases were identified were obtained using a handheld GPS, and were mapped using ArcGIS. Antibiotic susceptibility testing was performed using the Kirby ‘Bauer agar disc diffusion method. The full genomes of these strains were sequenced with the Illumina MiSeq platform. De novo assembly of cholera genomes and multiple sequence alignment were carried out using the bioinformatics pipeline developed in the Emerging Pathogens Institute laboratory at the University of Florida. Results/Discussion: Genetic comparisons showed that isolates were closely related, with pairwise p-distances ranging from 2.25 to 14.52 10-5 nt substitutions per site, and no statistically significant correlation between the pairwise genetic distances and the geographic distances among sampling locations. Indeed, the phylogeny of the Cameroonian strains displays the typical star-like topology and intermixing of strains from different locations that are characteristic of an exponential outbreak localized around a relatively restricted area with occasional spillover to other parts of the country, likely mediated by direct human contact and human movement. Findings highlight the utility of whole genome sequencing and phylogenetic analysis in understanding transmission patterns at the local level. This study was supported by supplement to NIAID grant No. R01AI097405 awarded to J Glenn Morris Jr. ==== Body Introduction Between 2010 and 2011, Cameroon recorded its worst cholera outbreak since1971, when the disease was first reported in the country. In the 2010-2011 epidemic, 33,192 cholera cases with 1,440 deaths (case fatality ratio 4.34%) were reported to the World Health Organization.1 , 2 In the southern plateau of Cameroon, the South West Region was one of the heavily affected regions during the 2010/11 outbreak. This region is in the Equatorial Monsoon3 climatic sub zone of Cameroon, close to the coast, raising questions as to whether transmission was linked with the development of environmental reservoirs or human movement. This is a key question in developing cholera intervention strategies; it is also a question which has generated considerable controversy, with at least one group hypothesizing that cholera transmission in Africa is solely a function of human movement,4 , 5 without the environmental reservoirs as reported in Asia and other parts of the world. Further, given the magnitude of the outbreak, another key question was whether or not the outbreak’s causative isolate was resistant to the prevailing antibiotics. We sought to address these questions for the South West Region-Cameroon, making use of isolates and whole genome sequence data from the 2010/11 cholera epidemic. The Study While clinical cases are incomplete for 2010, 2011 case data obtained from the ministry of public health of Cameroon showed evidence of three disease peaks, which did not correlate with seasonal patterns of rainfall (Figure 1); very few cases were recorded in 2012 while cases have not been identified in subsequent years. Toxigenic V. cholerae O1 altered biotype Ogawa was obtained from 18 patients (12% of rice-water stool samples collected) who were living in diverse geographic locations ranging from estuarine environmental sites along the coastline and inland sites, including Mundemba (Figure 2A). The ages of the patients ranged from <3 to 70 years. Samples were collected from October 2010 through to June 2011, and sampling sites included all health facilities (clinics, hospitals, and cholera treatment centers) reporting at least one case of cholera. Global positioning system (GPS) coordinates of each clinical sampling site were obtained using GARMIN eTrex 30 handheld GPS and maps were produced using ArcGIS 10.2. Total number of weekly cholera cases in the South West Region of Cameroon in 2010 and 2011. This region is in the Equatorial Monsoon climate sub zone of Cameroon. Cholera exhibits a trimodal transmission pattern with peaks in the dry (week 3) and rainy (weeks 19 and 27) seasons. Cholera clinical data was obtained from the ministry of public health of Cameroon. Total number of weekly cholera cases in the South West Region of the Equatorial Monsoon climate of Cameroon for 2010 and 2011. Cholera exhibit a trimodal transmission pattern with peaks in the dry (week 3) and rainy (weeks 19 and 27) seasons. Geographic distribution of sampled V. cholerae strains in South West Region-Cameroon. Panel A: red dots indicate the sample location of each strain in South West-Cameroon; the purple circle includes strains from the high-density populated area around the capital that likely harbored the origin of the outbreak. Panel B: pairwise genetic distances (y-axis) and geographic distances (in Km) among sampling locations (x-axis) were plotted to investigate specific migration trends of V. cholerae from the epicenter of the outbreak (purple circle) to the periphery. The two-tailed Mantel test failed to find any statistical correlation between the matrices of pair-wise genetic and geographic distances (p=0.74), indicating that strains genetic divergence does not increase with sampling locations progressively more distant and suggesting a homogeneous outbreak localized around a center, with occasional spillover in distant areas. Panel C: Maximum likelihood phylogenetic tree of full genome cholera strains from South West Region-Cameroon. The sampling location of each strain is given in parenthesis; ‘Point A’ represents areas enclosed within the purple circle in the map. Geographic distribution of sampled V. cholerae strains in South West Region-Cameroon. Panel A: red dots indicate the sample location of each strain in southwest Cameroon; the purple circle includes strains from the high-density populated area around the capital that likely harbored the origin of the outbreak. Panel B: pairwise genetic distances (y-axis) and geographic distances (in Km) among sampling locations (x-axis) were plotted to investigate specific migration trends of V. cholerae from the epicenter of the outbreak (purple circle) to the periphery. The two-tailed Mantel test failed to find any statistical correlation between the matrices of pair-wise genetic and geographic distances (p=0.74), indicating that strains genetic divergence does not increase with sampling locations progressively more distant and suggesting a homogeneous outbreak localized around a center, with occasional spillover in distant areas. Panel C: Maximum likelihood phylogenetic tree of full genome cholera strains from South West Region-Cameroon. The sampling location of each strain is given in parenthesis; ‘Point A’ represents areas enclosed within the purple circle in the map. Fecal samples from suspect cases were analyzed in the Laboratory for Emerging Infectious Diseases, University of Buea, Cameroon. After initial isolation on thiosulfate citrate bile salts sucrose (TCBS) agar and passage on brain heart infusion agar (BHIA), oxidase-positive isolates from BHIA were presumptively identified as V. cholerae O1 by slide agglutination assay using polyvalent antiserums specific to V. cholerae O1 and O139. Polyvalent antiserum positive V. cholerae O1 isolates were differentiated to sub-serotypes by testing against monovalent antiserum (Ogawa, Inaba, and Hikojima). Each V. cholerae O1 isolate was further characterized using a panel of PCR and mismatch amplification mutation assay (MAMA) PCR primers targeting V. cholerae species specific and virulence specific genes, including ctxA, ctxB and tcpA genes, as described previously.6 Furthermore, we PCR amplified the entire ctxB gene from each Cameroonian isolate and the PCR amplicon was sequenced in the Interdisciplinary Center for Biotechnology Research (ICBR) at University of Florida. Sequence comparison showed that all 18 isolates have identical ctxB gene which aligned (100%) to the ctxB7 7 sequence of V. cholerae classical strain, O395, but differed from ctxB3 sequence of N16961, an El Tor V. cholerae strain.8 Our results confirm that the South West Region-Cameroonian isolates are V. cholerae strains carrying ctxB7 gene mirroring classical ctxB in El Tor background as corroborated by the MAMA PCR. Antibiotic susceptibility testing was performed using the Kirby ‘Bauer agar disc diffusion method following CLSI (Clinical and Laboratory Standards Institute)9 guidelines9 , 10 , 11 and disk were collected from Oxoid, Hampshire, UK. Susceptibility was determined to Amikacin (30 μg), Ampicillin (10 μg), Azithromycin (15 μg), Cefixime (five μg), Ceftazidime (30 μg), Ceftriaxone (30 μg), Cefotaxime (30 μg), Ciprofloxacin (5 μg), Chloramphenicol (30 μg), Doxycycline (30 μg), Gentamicin (120 μg), Tetracycline (30 μg), Nalidixic acid (30 μg) and Trimethoprim /Sulfamethoxazole (25 μg). Our results indicate that all the strains showed the same pattern of antibiotic susceptibility. They are resistant against Nalidixic acid and Trimethoprim/Sulfamethoxazole, but intermediate against Chloramphenicol and Ampicillin, and susceptible to the rest of the antibiotics used in the study (Supplemental Table S1) The full genomes of all the 18 V. cholerae O1 isolates were sequenced with the Illumina MiSeq platform. De novo assembly of cholera genomes and multiple sequence alignment were carried out using our previously described bioinformatic pipeline.12 There were a total of 2673 single nucleotide polymorphisms (SNPs) within the outbreak genomes after excluding repetitive and putative recombinant (identified with the program Gubbins13) regions. Sequence data have been deposited at DDBJ/ENA/GenBank under the accession XXXX00000000. The version described in this paper is version XXXX01000000 (Supplemental Table S2). Genetic comparisons showed that isolates were closely related, with pairwise p-distances ranging from 2.25 to 14.52 10-5 nt substitutions per site. Potential correlation between the pairwise genetic distances and geographic distances among strains sampling locations was investigated with a two-tailed Mantel test, using GPS coordinates of the sampling sites. No statistically significant correlation was found (p=0.74, Figure 2B). In addition, we evaluated whether there is a correlation between the genetic distance from the common ancestor (root of tree) and the geographical distance from the epicenter of the epidemic (i.e., the city of Buea). There was no association between the genetic distance from the root and geographic distance from the epicenter (Figure 3), which may reflect the impact of human mobility. Using phylogenetic analysis, genomic relatedness among the strains was further investigated. A maximum likelihood phylogenetic tree was inferred from the multiple alignments of the core genomes of the sampled strains using the HKY+G nucleotide substitution model and 1000 bootstrap replicates to assess statistical support for the tree internal branches, as previously described.12 Correlation between the distance from root and the geographic distance of the sampling location of each strain from the epicenter of the epidemic. No correlation was found between genetic distances from root and geographic distance from epicenter, which may reflect the impact of human mobility. Correlation between the distance from root and the geographic distance of the sampling location of each strain from the epicenter of the epidemic. No correlation was found between genetic distances from root and geographic distance from epicenter, which is expected in an exponential outbreak. The maximum likelihood phylogeny of the strains (Figure 2C) displays the typical star-like topology and intermixing of strains from different locations consistent with an exponential outbreak around a central area, resulting in the rapid spread of closely related strains. Indeed, no internal branch in the tree had significant bootstrap support (values were all consistently <60%), except for the branch clearly separating strain CMR001, one of the earlier strains collected in the epicenter of the outbreak (purple circle, Figure 2A), from all other strains (100% bootstraps). A potential limitation of this study is that the low diversity of the sample may diminish the power of phylogenetic inference, although it may in itself be an indication of exponential growth and the absence of repeated transmission from a more diverse environmental reservoir. Conclusions This study provides data on whole genome sequence analysis of V. cholerae isolates from the South West Region-Cameroon. While the epidemic curve shows multiple peaks, there was no evidence of seasonality, nor has there been major recurrence of illness in subsequent years: i.e., we did not see the annual seasonal epidemics described in Asia and other regions of the world. Movement of strains through local environmental reservoirs might be expected to show correlation between genetic distance and geographic distance of sampling locations, with genetic heterogeneity generally increasing with progressively more distant locations from the initial epicenter of epidemic. Moreover, phylogenetic relationships among sampled strains would show highly supported monophyletic clades, corresponding to different transmission chains. On the contrary, the star-like topology of the tree and the lack of correlation between genetic and geographic distance seen with the South West Region outbreak are highly consistent with an exponential outbreak localized around a relatively restricted area (i.e., the city of Buea) with occasional spillover to other parts of the region likely mediated by occasional human-to-human contact. Intriguingly, one of the earliest sampled strains from the epicenter of the epidemic was significantly separated from all other strains, and appeared to be a natural outgroup for the tree; and thus, this phylogenetic analysis suggests this natural outgroup as the origin of the epidemic around the urban area (i.e., the city of Buea). In separate studies of spatial-temporal clustering of cholera cases in the Equatorial Monsoon3 climate subzone, we have found that cholera cases clustered in health districts with highways (RR = 3.99, CI95% 1.08-12.40) (Ngwa et al., unpub. data), again supporting the idea of transmission through human movements. Taken together, the evidence is in support of the model that an index case probably migrated and triggered the outbreaks in the region as opposed to the ingestion of V. cholerae from an environmental reservoir. Antimicrobial resistance patterns mirror those reported from other areas of Cameroon during the 2010/11 epidemics;7 , 8 , 14 of note, resistance was not seen to any of the antibiotics used as “front line” agents against cholera, including tetracycline, ciprofloxacin, and azithromycin. Over the past decades, sub-Sahara Africa has emerged as the primary contributor to the global cholera disease burden. From 2001-2009, 93% to 98% of all reported cases worldwide were from Africa, and in 2014, Africa was the leading global source of cholera burden, with 105,287 cases and 1,882 deaths reported from 19 countries.15 Some 20 years ago, Colwell proposed what has come to be known as the ‘cholera paradigm’,16 the concept that establishment of cholera endemicity requires the presence of aquatic environmental reservoirs, with seasonal/weather related increases in toxigenic V. cholerae O1 in these reservoirs serving as the primary trigger for human epidemics.17 , 18 In recent systematic reviews, Rebaudet et al. have challenged this conceptual framework, insisting that the ‘cholera paradigm’ does not apply to Africa.4 , 5 Instead, they hypothesize that cholera transmission in Africa is a function of human-to-human interactions, with human displacements (internal and external) serving as the major determinant of spread and trigger for recurrent epidemics. Our data, at least for the region under study in Cameroon, support this latter hypothesis. This, in turn, has implications for cholera management within Africa, underscoring the importance of monitoring (and limiting) movement of potentially infected persons into regions without active cholera epidemics. Competing Interests The authors have declared that no competing interest exists. Supplements Supplemental Table S1. Susceptibility results of V. cholerae strains from South West Region-Cameroon Susceptibility results of V. cholerae strains from South West Region-Cameroon Supplemental Table S2. Accession number of whole genome sequenced Isolates from South West Region, Cameroon cholera outbreak in 2010/2011 Accession number of whole genome sequenced Isolates from South West Region, Cameroon cholera outbreak in 2010/2011. M. Ngwa is a research assistant at the Emerging Pathogens Institute and the Department of Environmental and Global Health, College of Public Health and Health Professions, University of Florida. His interests are health economics, logistics and public health research. Foche was born in 1986 in Lobe. Education Foche attended three different primary schools (G.S. Ntoh, C.S. Bujong, and C.S Ntenefor) and attained his First School Leaving Certificate at C.S. Ntenefor in 1999. He enrolled in Government Bilingual High School (G.B.H.S.) Bamenda-nkwe where he attended secondary and high school education earning the G.C.E. Ordinary Level and Advanced Level, respectively, in 2004 and 2006. Thereafter, he enrolled as an undergraduate student at the University of Buea in 2006 where he obtained a B.Sc. in Microbiology with First Class Honours (3.65/4.0), emerging as the best Microbiology student in 2009 class, and overall third in the University of Buea 2009 graduation batch. During his undergraduate studies, he helped many of his class mates to go through the degree program within the required duration by organizing and facilitating a study group. After his undergraduate studies, he enrolled to study Masters in the same university in 2009 after which he obtained an M.Sc. in Microbiology in 2012 with Second Class Honours, Upper Division (3.5/4). His zeal to study abroad even before enrolling for Masters increased his reluctance to enroll into doctoral studies in Cameroon. He however, enrolled as a Ph.D student studying Microbiology at his former university in 2014. He is currently looking for funding opportunities and scholarships to complete his doctoral studies within or without Cameroon. Work 2010 – 2013: Secondary school teacher (CCAST Sandpit, Baird Memorial College) in Buea. 2013 – Present: Laboratory technologist at the Laboratory for Emerging Infectious Diseases, University of Buea. 2015 – Present: Part-time lecturer at the Saint Monica University, Muea, Cameroon. ==== Refs References 1 Cholera, 2010. Wkly Epidemiol Rec. 2011 Jul 29;86(31):325-39. PubMed PMID:21800468. 21800468 2 Cholera, 2011. Wkly Epidemiol Rec. 2012 Aug 3;87(31/32):289–304. PubMed PMID:22905370. 22905370 3 Molua EL, Lambi CM. Climate, hydrology and water resources in Cameroon. CEEPA DP33, South Africa, University of Pretoria; 2006. 4 Rebaudet S, Sudre B, Faucher B, Piarroux R. Environmental determinants of cholera outbreaks in inland Africa: a systematic review of main transmission foci and propagation routes. J Infect Dis. 2013 Nov 1;208 Suppl 1:S46-54. PubMed PMID:24101645. 24101645 5 Rebaudet S, Sudre B, Faucher B, Piarroux R. Cholera in coastal Africa: a systematic review of its heterogeneous environmental determinants. J Infect Dis. 2013 Nov 1;208 Suppl 1:S98-106. PubMed PMID:24101653. 24101653 6 Ali A, Chen Y, Johnson JA, Redden E, Mayette Y, Rashid MH, Stine OC, Morris JG Jr. Recent clonal origin of cholera in Haiti. Emerg Infect Dis. 2011 Apr;17(4):699-701. PubMed PMID:21470464. 21470464 7 Reimer AR, Van Domselaar G, Stroika S, Walker M, Kent H, Tarr C, Talkington D, Rowe L, Olsen-Rasmussen M, Frace M, Sammons S, Dahourou GA, Boncy J, Smith AM, Mabon P, Petkau A, Graham M, Gilmour MW, Gerner-Smidt P. Comparative genomics of Vibrio cholerae from Haiti, Asia, and Africa. Emerg Infect Dis. 2011 Nov;17(11):2113-21. PubMed PMID:22099115. 22099115 8 Kaas RS, Ngandjio A, Nzouankeu A, Siriphap A, Fonkoua MC, Aarestrup FM, Hendriksen RS. The Lake Chad Basin, an Isolated and Persistent Reservoir of Vibrio cholerae O1: A Genomic Insight into the Outbreak in Cameroon, 2010. PLoS One. 2016;11(5):e0155691. PubMed PMID:27191718. 27191718 9 Clinical and Laboratory Standards Institute. Methods for antimicrobial dilution and disk susceptibility testing of infrequently isolated or fastidious bacteria; approved guideline. 2nd ed. Wayne, PA: Clinical and Laboratory Standards Institute; 2010. p. 96. (M45A2). 10 Bauer AW, Kirby WM, Sherris JC, Turck M. Antibiotic susceptibility testing by a standardized single disk method. Am J Clin Pathol. 1966 Apr;45(4):493-6. PubMed PMID:5325707. 5325707 11 Centers for Disease Control and Prevention. Manual for the laboratory detection of antimicrobial resistance among acquired bacterial pathogens of public health concern in the developing world (draft)Atlanta, GA: Centers for Disease Control and Prevention; 2001. pp. 55–64. (CDC/WHO/USAID). 12 Azarian T, Ali A, Johnson JA, Mohr D, Prosperi M, Veras NM, Jubair M, Strickland SL, Rashid MH, Alam MT, Weppelmann TA, Katz LS, Tarr CL, Colwell RR, Morris JG Jr, Salemi M. Phylodynamic analysis of clinical and environmental Vibrio cholerae isolates from Haiti reveals diversification driven by positive selection. MBio. 2014 Dec 23;5(6). PubMed PMID:25538191. 25538191 13 Croucher NJ, Page AJ, Connor TR, Delaney AJ, Keane JA, Bentley SD, Parkhill J, Harris SR. Rapid phylogenetic analysis of large samples of recombinant bacterial whole genome sequences using Gubbins. Nucleic Acids Res. 2015 Feb 18;43(3):e15. PubMed PMID:25414349. 25414349 14 Quilici ML, Massenet D, Gake B, Bwalki B, Olson DM. Vibrio cholerae O1 variant with reduced susceptibility to ciprofloxacin, Western Africa. Emerg Infect Dis. 2010 Nov;16(11):1804-5. PubMed PMID:21029554. 21029554 15 Cholera, 2014. Wkly Epidemiol Rec. 2015 Oct 2;90(40):517-28. PubMed PMID:26433979. 26433979 16 Colwell RR. Global climate and infectious disease: the cholera paradigm. Science. 1996 Dec 20;274(5295):2025-31. PubMed PMID:8953025. 8953025 17 Morris JG Jr. Cholera--modern pandemic disease of ancient lineage. Emerg Infect Dis. 2011 Nov;17(11):2099-104. PubMed PMID:22099113. 22099113 18 Jutla A, Akanda AS, Huq A, Faruque AS, Colwell R, Islam S. A water marker monitored by satellites to predict seasonal endemic cholera. Remote Sens Lett. 2013;4(8):822-831. PubMed PMID:23878762. 23878762

PMC005xxxxxx/PMC5001704.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756441510.1371/journal.pone.0161898PONE-D-16-20281Research ArticleBiology and Life SciencesAnatomyOcular SystemOcular AnatomyRetinaMedicine and Health SciencesAnatomyOcular SystemOcular AnatomyRetinaBiology and Life SciencesPhysiologyImmune PhysiologyComplement SystemMedicine and Health SciencesPhysiologyImmune PhysiologyComplement SystemBiology and Life SciencesImmunologyImmune SystemComplement SystemMedicine and Health SciencesImmunologyImmune SystemComplement SystemBiology and Life SciencesImmunologyImmune System ProteinsComplement SystemMedicine and Health SciencesImmunologyImmune System ProteinsComplement SystemBiology and Life SciencesBiochemistryProteinsImmune System ProteinsComplement SystemResearch and Analysis MethodsModel OrganismsAnimal ModelsMouse ModelsBiology and Life SciencesPhysiologyImmune PhysiologyComplement SystemComplement ActivationMedicine and Health SciencesPhysiologyImmune PhysiologyComplement SystemComplement ActivationBiology and Life SciencesImmunologyImmune SystemComplement SystemComplement ActivationMedicine and Health SciencesImmunologyImmune SystemComplement SystemComplement ActivationBiology and Life SciencesImmunologyImmune System ProteinsComplement SystemComplement ActivationMedicine and Health SciencesImmunologyImmune System ProteinsComplement SystemComplement ActivationBiology and Life SciencesBiochemistryProteinsImmune System ProteinsComplement SystemComplement ActivationBiology and Life SciencesAnatomyCardiovascular AnatomyBlood VesselsRetinal VesselsMedicine and Health SciencesAnatomyCardiovascular AnatomyBlood VesselsRetinal VesselsBiology and Life SciencesAnatomyOcular SystemOcular AnatomyRetinal VesselsMedicine and Health SciencesAnatomyOcular SystemOcular AnatomyRetinal VesselsResearch and Analysis MethodsImmunologic TechniquesImmunoassaysEnzyme-Linked ImmunoassaysMedicine and Health SciencesGeriatricsGeriatric OphthalmologyMacular DegenerationMedicine and Health SciencesOphthalmologyGeriatric OphthalmologyMacular DegenerationMedicine and Health SciencesOphthalmologyRetinal DisordersMacular DisordersMacular DegenerationMedicine and Health SciencesOphthalmologyRetinal DisordersRetinal DegenerationMacular DegenerationResearch and Analysis MethodsSpecimen Preparation and TreatmentStainingImmunostainingRegulation of C3 Activation by the Alternative Complement Pathway in the Mouse Retina C3 Activation in the RetinaWilliams Jennifer A. E. 1Stampoulis Dimitris 1Gunter Chloe E. 1Greenwood John 1Adamson Peter 2¤http://orcid.org/0000-0002-7304-8879Moss Stephen E. 1*1 Department of Cell Biology, UCL Institute of Ophthalmology, 11–43 Bath Street, London EC1V 9EL, United Kingdom2 Ophthiris Discovery Performance Unit and Department of Laboratory Animal Science, GlaxoSmithKline, Medicines Research Centre, Gunnelswood Road, Stevenage, Herts SG1 2NY, United KingdomLangmann Thomas EditorUniversity of Cologne, GERMANYCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: JG PA SM. Data curation: JW SM. Formal analysis: JW SM. Funding acquisition: SM JG. Investigation: JW CG DS. Methodology: PA SM JW JG. Project administration: SM. Resources: PA SM JG. Supervision: SM. Validation: JW DS CG. Visualization: JW SM JG. Writing – original draft: SM JW. Writing – review & editing: PA JG. ¤ Current address: ProQR Therapeutics, Darwinweg 24, 2333 CR Leiden, The Netherlands * E-mail: s.moss@ucl.ac.uk26 8 2016 2016 11 8 e016189823 5 2016 12 8 2016 © 2016 Williams et al2016Williams et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.The purpose of this study was to examine the retinas of mice carrying hemizygous and null double deletions of Cfb-/- and Cfh-/-, and to compare these with the single knockouts of Cfb, Cfh and Cfd. Retinas were isolated from wild type (WT), Cfb-/-/Cfh-/-, Cfb-/-/Cfh+/-, Cfh-/-/Cfb+/-, Cfb-/-, Cfh-/- Cfd-/-, and Cfd+/- mice. Complement proteins were evaluated by western blotting, ELISA and immunocytochemistry, and retinal morphology was assessed using toluidine blue stained semi-thin sections. WT mice showed staining for C3 and its breakdown products in the retinal vasculature and the basal surface of the retinal pigment epithelium (RPE). Cfb-/- mice exhibited a similar C3 staining pattern to WT in the retinal vessels but a decrease in C3 and its breakdown products at the basal surface of the RPE. Deletion of both Cfb and Cfh restored C3 to levels similar to those observed in WT mice, however this reversal of phenotype was not observed in Cfh-/-/Cfb+/- or Cfb-/-/Cfh+/- mice. Loss of CFD caused an increase in C3 and a decrease in C3 breakdown products along the basal surface of the RPE. Overall the retinal morphology and retinal vasculature did not appear different across the various genotypes. We observed that C3 accumulates at the basal RPE in Cfb-/-, Cfb-/-/Cfh-/-, Cfb-/-/Cfh+/-, Cfd-/- and WT mice, but is absent in Cfh-/- and Cfh-/-/Cfb+/- mice, consistent with its consumption in the serum of mice lacking CFH when CFB is present. C3 breakdown products along the surface of the RPE were either decreased or absent when CFB, CFH or CFD was deleted or partially deleted. http://dx.doi.org/10.13039/100004440Wellcome TrustStephen Mosshttp://dx.doi.org/10.13039/100004330GlaxoSmithKlineStephen MossThe work was funded by an Educational Grant from GSK, and a project grant from the Wellcome Trust (090669). Data AvailabilityAll relevant data are within the paper.Data Availability All relevant data are within the paper. ==== Body Introduction Dysregulation of the complement system is a recognised characteristic of patients with age-related macular degeneration (AMD). However, AMD is a multifactorial disease and isolating the specific contributions of individual complement proteins to disease pathology is not straightforward. One approach to studying the role of the complement system in the maintenance of retinal health is to evaluate and compare the retinal phenotype of mice carrying mutations or deletions of specific complement genes. In AMD there are several complement gene single nucleotide polymorphisms (SNP) known to associate with increased susceptibility to or protection against AMD. The complement factor H (CFH) gene contains the most common complement-associated high risk SNP for AMD that changes a tyrosine at position 402 to a histidine [1–4]. CFH is a fluid-phase regulator of the alternative pathway of complement activation, exerting its activity in three ways, by i) preventing C3 convertase formation, ii) promoting its dissociation and iii) participating in the breakdown of the active split product C3b. Formation of the C3 convertase is driven by complement factor B (CFB), and consequent downstream activation of C5 and generation of the membrane attack complex. In previous studies we showed that Cfh-/- mice exhibit visual dysfunction at 2 years, with C3 and autofluorescent debris accumulating at Bruch’s membrane, disorganisation of retinal pigment epithelial (RPE) cell organelles and thinning of Bruch’s membrane [5]. Early compensatory signs of complement dysregulation are evident at 1 year in Cfh-/- mice where CD59 mRNA expression in the neuroretina is reduced [6] and CD55 (decay-accelerating factor) expression in Müller cells is up-regulated [7]. Recently it has been shown that the Y402H SNP in CFH affects its binding to the lipid peroxidation product malondialdehyde, which accumulates with age [8]. SNPs that provide protection against AMD have also been identified such as the R32Q variant of complement factor B (CFB) [9]. CFB is expressed by the RPE [10] and upon down-regulation by siRNA or complete gene knock-out in mice, has been shown to provide protection against choroidal neovascularisation after laser burn of Bruch’s membrane [11,12]. Complement factor D (CFD), which is a serine protease required for the generation of the C3 convertase, is currently a focus of interest due to on-going clinical trials of the function-blocking antibody Lampalizumab in dry AMD [13,14], and Cfd gene knockout has been shown to be protective in a mouse model of light-induced phototoxicity [15]. However phenotypic characterisation of the Cfb-/- and Cfd-/- retinas has not been reported. A caveat of studies using Cfh-/- mice is that their serum is almost completely depleted of C3 since without CFH there is no fluid phase regulator to prevent its breakdown. Furthermore, these mice also lack CFB as this is consumed in the breakdown of C3. Therefore these mice are close to being a CFH, C3, CFB triple knock-out. Unlike the Cfh-/- mice, Cfb-/-/Cfh-/- double knockout mice would be expected to have normal to higher levels of C3 since without CFB, C3 cannot be broken down via the alternative pathway. We would also expect this to be the case in Cfd-/- mice. Without CFB or CFD we would not expect to see C3 breakdown products since the C3 convertase cannot form without both activators. In this study, we address these questions by characterising the retinas of Cfb-/-, Cfh-/- and Cfd-/- mice at 12 months to identify signs of retinal abnormalities, and for the first time report the phenotype of the Cfb-/-/Cfh-/- double knock-out mouse retina, with Cfh-/-/Cfb+/- and Cfb-/-/Cfh+/-mice examined as a gene dosage control for CFB and CFH. Materials and Methods Animals Wild-type, Cfb-/-, Cfh-/-, Cfb-/-/Cfh-/-, Cfb-/-/Cfh+/- and Cfh-/-/Cfb+/- mice were housed for 12 months at Charles River Laboratories France, Domaine des Oncins, BP 10969592, L’ARBRESLE CEDEX. Cfd+/+, Cfd+/- and Cfd-/- mice were housed at GSK. All mice were on the C57/Bl6J background. This study was ethically reviewed and approved by the Institute of Ophthalmology Animal Welfare and Ethical Review Board, and carried out in accordance with Animals (Scientific Procedures) Act 1986 and the GSK Policy on the Care, Welfare and Treatment of Animals. All mice were sent to UCL where they were culled by a rising concentration of CO2 with subsequent cervical dislocation and the eyes removed immediately. Semithin sections Eyes were fixed in Karnovsky’s fixative (3% glutaraldehyde (EM grade-TAAB, G002), 1% paraformaldehyde in 0.07 M sodium cacodylate (Agar Scientific, R1104), pH 7.4) for 2 h at RT. Sectioning and staining were performed as previously described [16]. Immunofluorescence After cervical dislocation, eyes were enucleated and immediately placed in 2% paraformaldehyde (1x PBS) for 7 min on ice. Eyes were then transferred to 2 x PBS for up to 30 min on ice prior to dissection. Eyes were then fixed again in 4% paraformaldehyde for 30 min at RT and subsequently cryopreserved in OCT embedding matrix and cut into 12 μm sections. Sections were blocked and permeabilised in PBS containing 1% bovine serum albumin, 0.5% Triton X-100 and 0.12% sodium azide. Sections were either stained with a C3 antibody conjugated to FITC fluorophore (Cappel, 1:100 dilution) or a C3b/iC3b/C3c antibody (Rat monoclonal antibody, Hycult, 1:50 dilution) overnight at 4°C. Sections stained with C3b/iC3b/C3c antibody were then washed and stained with anti-rat IgG conjugated to AlexaFluor 594 (1:200 dilution) for 1 h at RT. Sections were again washed and incubated with 1μg/ml DAPI for 2 min at RT before mounting in Mowiol mounting medium. Retinal wholemounts Mouse eyes were fixed in 4% paraformaldehyde for 30 min and remained in PBS overnight before dissecting. The neuroretina was peeled away from the RPE and incisions made to flatten before fixing in ice-cold methanol. Flatmounts were blocked and permeabilised for 1h at RT in 2X PBS containing 1% bovine serum albumin, 3% Triton X-100, 0.5% Tween-20 and 0.12% sodium azide. Flatmounts were stained using Collagen IV (Rabbit polyclonal, AbD Serotec, 1:500) antibody overnight at RT, followed by anti-rabbit IgG conjugated to AlexaFluor 488 and anti-rat IgG conjugated to AlexaFluor 594 (1:200 dilution). Sections were again washed and incubated with 1μg/ml DAPI for 2 min at RT before mounting in Mowiol mounting medium. The entire ribbon containing the optic nerve was analysed for each sample. We did not detect noticeable differences in C3 or C3 breakdown products along the ribbon, and therefore took images halfway along on either side of the optic nerve. Western blot analysis of mouse C3, CFB, CFH, CFD and C5 Mouse blood was collected by cardiac puncture in the presence of ethylenediamine tetraacetic acid (EDTA) from 6 month old mice of the genotypes listed earlier. Samples were chilled on ice and then plasma-separated by centrifugation at 4°C for 20 min. Proteins were separated using sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE): 4–12% gel under reducing conditions for C3, CFB, CFH and CFD, and 4–12% gel under non-reducing conditions for C5. PVDF membranes were blocked in 7% w/v non-fat dry milk in PBS. The buffer used for diluting the primary and secondary antibodies was 3.5% w/v non-fat dry milk in 0.025% Tween20/PBS. All the intermediate washing steps were performed in 0.05% Tween20/PBS. Antibodies used were peroxidase-conjugated goat anti-mouse complement C3 (product no. 55557; MP Biomedicals, 1:500), goat antisera to human CFH (product no. A312; Quidel, 1:500), goat antisera to human CFB (product no. A311; Quidel, 1:500), polyclonal sheep immunoglobulin (IgG) to mouse CFD (product no. AF5430; R&D Systems, 1:500) and goat antisera to human C5 (product no. A306; Quidel, 1:500). Secondary antibodies were polyclonal rabbit anti-goat HRP (product no. P0449; Dako, 1:1000), polyclonal rabbit anti-sheep HRP (product no. P0163; Dako, 1:1000), and peroxidase AffiniPure sheep anti-mouse IgG (product no. 515-035-071, Jackson Immunoresearch, 1:1000). Blots were visualized using Amersham enhanced chemiluminescence (ECL) Western Blotting Detection Reagent (product no. RPN2106, GE Healthcare Life Sciences). ELISA for plasma C3 Microtiter plates were coated overnight at 4°C with polyclonal goat IgG to mouse complement C3 (product no. 55463, MP Biomedicals, 1:8000) in PBS. The plates were then washed in 0.2% Tween20/PBS and blocked with 2% BSA in 0.2% Tween20/PBS at RT for 60 min. After washing twice, mouse plasma samples were loaded at 1:6000 and 1:12000 dilutions and left for 60 min at RT. Mouse C3 was detected with peroxidase-conjugated goat IgG fraction to mouse complement C3 (product no. 55557, MP Biomedicals, 1:25000) in 0.2% Tween20/PBS. Plates were developed using TMB substrate Reagent Pack (product no. DY999, R&D Systems). The concentration of plasma C3 was estimated by reference to a calibration curve constructed with 11 serial dilutions (1:2) of pooled mouse plasma of known C3 concentration (220 μg/ml). Plasma samples from 5–6 mice were tested for each genotype. Unpaired, two-tailed Student’s t-tests were applied to data. Results The three pathways of complement activation converge upon C3, which is the central component of the complement cascade. In order to determine the impact of null mutations in the Cfb, Cfh, and Cfd genes on the levels of circulating complement proteins, plasma samples from the corresponding mutant mice were examined by western blotting and ELISA (Fig 1). As expected, gene knockout of Cfb, Cfh and Cfd resulted in the absence of the cognate gene product from the plasma. However, whilst CFB was not detected in the plasma of Cfb-/- mice, it was also absent in samples from the Cfh-/-, Cfb-/-/Cfh-/-, Cfh-/-/Cfb+/- and Cfb-/-/Cfh+/- strains. In the Cfd+/- and Cfd-/- mice we observed that CFB levels were elevated relative to the WT Cfd+/+ mice. CFH was not present in detectable levels in the plasma of the Cfh-/- mice and neither was CFH detected in the plasma of any of the other strains carrying a homozygous Cfh deletion. Restoring CFH expression to heterozygous levels in the Cfb-/-/Cfh+/- mice resulted in detectable but slightly reduced levels of CFH compared to the WT. 10.1371/journal.pone.0161898.g001Fig 1 Expression of complement proteins in multiple complement gene knock-out mice. Plasma samples were taken from mice at age 6 months corresponding to the genotypes indicated. For western blotting, 1 μl whole plasma was analysed using specific antisera as detailed in the Materials and Methods. Representative blots are shown for each protein. For C3 ELISA, plasma samples from 5–6 mice were analysed per genotype. Data are means ± S.D, unpaired, two-tailed Student’s t-tests were applied to data. *P = < 0.05, **P = < 0.01, ***P = < 0.001. In the Cfd-/- and Cfd+/- mice, CFH was present with no apparent difference in expression compared to the WT Cfd+/+. Indeed, CFD was present at similar levels in all tested plasma samples apart from the Cfd-/- mice. The presence of plasma C5 was also analysed and the expression pattern across the nine genotypes mirrored that of C3. Thus, C5 was not detected in the Cfh-/- mice, but deletion of Cfb in order to generate the Cfb-/-/Cfh-/- mice restored the expression of C5 to levels comparable to those observed in WT mice. In contrast, in Cfh-/-/Cfb+/- mice, decreasing CFB expression to heterozygous levels did not lead to recovery of C5. In the Cfd-/- and Cfd+/- mice C5 was present but slightly reduced when compared to the WT Cfd+/+ mice. Complement C3 was evaluated by ELISA to obtain actual plasma concentrations, and also by western blotting to visualize the α and ß chains. In the Cfh-/- mice C3 was reduced to almost undetectable levels whereas in the Cfb-/- mice, levels of C3 were slightly though not significantly elevated compared to WT, consistent with C3 degradation being dependent on CFB. Deletion of Cfb in the Cfh-/- mice, to generate the Cfb-/-/Cfh-/- mice, reversed the level of C3 in the plasma to that observed in WT mice. However, in Cfh-/-/Cfb+/- mice, decreasing CFB expression to heterozygous levels was insufficient to rescue C3, which remained at almost undetectable levels. In contrast, restoring CFH expression to hemizygous levels in the Cfb-/-/Cfh+/- mice had no effect on plasma C3. Loss or partial loss of CFD in the Cfd-/- and Cfd+/- mice respectively, caused no significant changes in plasma C3 levels. The effects of these multiple complement gene knockouts on components of the alternative pathway are broadly in line with previous reports [17]. In order to identify any gross anatomical differences in retinal morphology caused by the loss of the complement genes under investigation here, semi-thin sections were cut and stained with toluidine blue (Fig 2). As pilot studies across the range of genotypes revealed no obvious changes at 6 months, animals were examined at 12 months to increase the chances of detecting slow onset phenotypic changes. However, all genotypes presented a similar morphology, with no marked differences between them with regard to retinal thickness or numbers of photoreceptor nuclei. We next examined retinal sections for C3 deposition, as accumulation of C3 and its breakdown products is a well-established characteristic of retinas from patients with AMD [18,19]. In WT mice, we observed C3 immunostaining within the retinal blood vessels, corresponding to the circulating pool and, consistent with previous reports, also along the basal surface of the RPE (Fig 3) [5,20]. In Cfb-/- mice, C3 staining was evident in the retinal blood vessels and along the basal surface of the RPE but to a markedly lesser extent than that observed in WT mice. As one would expect, given the results in Fig 1, the Cfh-/- mouse lacked any C3 staining in the retina, whereas in the Cfb-/-/Cfh-/- mouse this phenotype was reversed such that C3 returned to levels comparable to those observed in WT mice. Heterozygosity for CFH in the Cfb-/-/Cfh+/- mouse led to reduced staining of C3 in the RPE/Bruch’s. In Cfh-/-/Cfb+/- mice, decreasing CFB expression to heterozygous levels was insufficient to rescue C3 stability, which remained absent from the retinal vessels and the basal surface of the RPE. In the Cfd-/- mouse, but not the Cfd+/-, the staining of C3 in the RPE/Bruch’s was notably more intense than in control animals, perhaps because the absence of CFD would result in failure to generate the C3 convertase, leading to accumulation of C3. 10.1371/journal.pone.0161898.g002Fig 2 Loss of Cfb, Cfh or Cfd does not affect retinal morphology in one year mice. The images show toluidine blue stained 2 μm semithin sections of mouse retinas of indicated genotypes at 12 months. Scale bar = 50 μm. Images are representative of n = 3. 10.1371/journal.pone.0161898.g003Fig 3 Immunostaining of C3 is restored to wild type levels in the retinal vessels and RPE/Bruch’s space in Cfh-/- mice when in the genetic background of Cfb-/- but not Cfb+/-. 12 μm PFA fixed sections were taken of mouse retinas of the indicated genotypes at age 12 months. Sections were stained for C3-FITC (green) and nuclei (blue). Yellow arrows highlight staining in RPE/Bruch’s. Z-stacks were imaged using confocal microscopy and merged to form a maximum intensity projection. Scale bar = 50 μm. Images are representative of n = 6. Activation of the complement system leads to the breakdown of C3 into C3b, which in turn can be further degraded by CFI and CFH into iC3b and C3c. To analyse the presence of these breakdown products in the retina, sections were stained with an antibody specific to C3b/iC3b/C3c (Fig 4). In WT mice, these breakdown products were evident at the basal surface of the RPE, but in contrast to whole C3 none were visible within the retinal blood vessels in WT or any of the mutant strains. Where CFH was deleted in Cfh-/- and Cfh-/-/Cfb+/- mice, we observed no staining for C3b/iC3b/C3c consistent with the lack of circulating C3. Where CFB was deleted in Cfb-/- and Cfb-/-/Cfh+/- mice we observed only faint punctate staining for C3b/iC3b/C3c probably due to the inability of little C3 convertase to form in the absence of CFB. Surprisingly where both CFH and CFB were deleted in the Cfb-/-/Cfh-/- double knockout mouse, we observed C3b/iC3b/C3c in RPE/Bruch’s in this strain at similar levels to WT. In the Cfd-/- mouse, the C3b/iC3b/C3c staining pattern followed that of C3, with no C3b/iC3b/C3c detectable, and weak staining in the Cfd+/- mouse. To aid comparison of the different genotypes, qualitative analysis of the staining of C3 and its breakdown products is summarised in Table 1, together with the actual values of circulating C3 in the serum. 10.1371/journal.pone.0161898.g004Fig 4 C3 breakdown products in RPE/Bruch’s are absent in Cfh-/- mice but restored in the double knock-out Cfb-/-/Cfh-/- mouse. 12 μm PFA fixed sections were taken of mouse retinas of indicated genotypes at 12 months. Sections were stained for C3b/iC3b/C3c (red) and nuclei (blue). Z-stacks were imaged using confocal microscopy and merged to form a maximum intensity projection (MIP). MIPs were merged with digital interference contrast and arrows highlight the presence of C3 breakdown products in RPE/Bruch’s. Scale bars = 50 μm. Images are representative of n = 6. 10.1371/journal.pone.0161898.t001Table 1 Values of circulating C3 (mg/L ± SEM) are presented for each genotype. N = 6 for all genotypes except Cfd-/- where n = 5. Staining intensity for C3 and iC3b in the RPE/Bruch’s membrane was scored as follows: +++ = strong, ++ = moderate, + = weak,— = not detected. Genotype C3 in serum (mg/L) C3 in RPE/Bruch’s iC3b in RPE/Bruch’s WT 495 ± 117 +++ +++ Cfb-/- 749 ± 146 + + Cfh-/- 16.5 ± 2.6 - - Cfb-/-/Cfh-/- 350 ± 30.3 ++ +++ Cfb+/-/Cfh-/- 31.9 ± 4.4 - - Cfh+/-/Cfb-/- 401 ± 111 + + Cfd+/+ 415 ± 89 ++ ++ Cfd+/- 390 ± 51 ++ + Cfd-/- 354 ± 91 +++ - It has been previously reported, by fluorescein angiography, that retinal blood vessels in the deep plexus of 1 year Cfh-/- mice were withered due to the accumulation of C3 breakdown products on blood vessel walls [21]. Here we failed to detect C3 breakdown products anywhere in the retina (data not shown), but we nevertheless examined retinal vessel morphology and density by immunostaining neuroretinal flatmounts for collagen IV (Fig 5). In all genotypes tested, each retinal vessel plexus was morphologically comparable to that of WT mice with no sign of withering vessels or other vascular abnormalities. The absence of a distinct vascular phenotype in this study is consistent with the majority of reports of complement mutant mice, perhaps indicating that secondary environmental factors such as microbiome or diet may modulate the phenotype in some instances. 10.1371/journal.pone.0161898.g005Fig 5 Organisation of the retinal blood vessels is unaffected by the loss of Cfb or Cfh. Neuroretinas from mice of indicated genotypes aged 12 months were flatmounted and fixed in methanol. Vessels were stained for Collagen IV (green) and z-stacks imaged using confocal microscopy. Images represent a maximum intensity projection of each plexus ranging from 8–12 μm for the inner, intermediate and deep plexuses. Scale bars = 100 μm. Images are representative of n = 3. Discussion CFB and CFH are key mediators in the alternative pathway of complement activation. CFB is required for the formation of the C3 convertase whereas CFH is a key regulator of C3 convertase formation, and also acts as a co-factor for CFI in the breakdown of C3b. Investigating the roles of these proteins within the retina is of particular interest since mutations in both genes have been identified as either protective in the case of CFB (R32Q) [9] or associated with increased risk of developing AMD for CFH (Y402H) [1–4]. It has been shown in the mouse that both proteins are involved in choroidal neovascularisation, in that reduction in CFB has a protective effect upon laser-induced injury whereas loss of CFH leads to enhanced pathological choroidal neovascularisation [11,12,22]. These observations are consistent with the alternative pathway of complement activation playing a role in the development of vascular pathology. Both proteins are present in serum but whether they access the retina from the circulation is unknown. Within the retina, a potentially important local source of CFB and CFH is the RPE, and the expression of both proteins by RPE cells has been reported to increase in response to inflammatory stimuli [10,23]. Several studies have been performed on the retinas of Cfh-/- mice [5–7,20,21]. We previously showed functional changes in scotopic electroretinography (ERG) that were mild at 1 year and more pronounced by 2 years. Loss of CFH also led to structural changes in the RPE, photoreceptors and Bruch’s membrane. Overall these studies suggest that the retina requires CFH for complement homeostasis and normal visual function. However, in interpreting these results in the context of CFH function, it must be kept in mind that a confounding feature of these mice is the almost total absence of C3 and CFB from the serum due to its non-regulated breakdown by the absence of CFH in the fluid phase [17]. Indeed, the complete absence of C3 in 1 year old Cfh-/-/C3-/- double knock-out mice causes more pronounced visual dysfunction than in C3-/- or Cfh-/- single knock-out mice at this age, suggesting that CFH is required for more than just the regulation of C3 convertase formation and breakdown [20]. In patients with AMD, C3 in the serum is not depleted [24,25] and therefore studying the effect of the loss of CFH when C3 is present is of interest. Here we have partially addressed this problem by crossing Cfb-/- mice with Cfh-/- mice to create the double knockout, and show that in these animals the level of C3 in the retina, as judged by immunostaining, and in the plasma measured by ELISA, is approximately equivalent to that observed in control mice. Our results show that the loss of CFB, CFH or both proteins together does not lead to any significant anatomical or morphological changes in the retina, at least in the first year of life. This suggests that at 12 months, and in the absence of pathological challenge, infection or environmental stress, the retina is stable and healthy in the absence of a functional alternative pathway. In a recent microarray study of the retinas of young, 7–8 week Cfh-/- mice, we observed no significant changes at the transcriptional level in any complement regulatory genes [6]. Examination of the retinal vasculature and microglial infiltration (not shown) similarly revealed no significant differences between genotypes. As expected, the main consequences of deleting CFB, CFH or both proteins, were in relation to the levels of C3 and its breakdown products, C3b, iC3b and C3c, in the retina. Our data confirm those of Pickering et al., who first reported that C3 is no longer present in the serum in Cfh-/- mice [17]. The loss of C3 from the basal surface of the RPE in Cfh-/- and Cfh-/-/Cfb+/- mice, when C3 is low in the serum, suggests that the C3 that deposits at this location in control, Cfb-/- and Cfb-/-/Cfh+/- mice originates from the serum. Without CFB, C3 convertases are no longer able to form via the alternative pathway and therefore C3 cannot be broken down unless the classical or lectin pathways become activated. Indeed we did see immunostaining of C3 in the retinal vasculature and Bruch’s membrane of the Cfb-/- mice. Low levels of C3 were also present along the basal surface of the RPE reinforcing the notion that these C3 deposits are derived from the serum. However, in contrast to WT animals, C3 breakdown products were almost undetectable in the Cfb-/- and Cfb-/-/Cfh+/- mice, consistent with C3 convertases being prevented from forming. In the double knock-out mice the loss of both CFH and CFB restored C3 expression to levels similar to those observed in WT or Cfb-/- mice. But without the ability to form C3 convertases via the alternative pathway, the regulatory role of CFH becomes redundant. Despite this, in the Cfb-/-/Cfh-/- mice there was significant immunostaining of C3 breakdown products at the basal surface of the RPE suggesting that in these mice, C3b is generated by either the lectin or classical pathway. We can speculate that the loss of CFH in addition to CFB causes an imbalance that is not present in Cfb-/- mice. When CFB expression was partially restored in the Cfh-/-/Cfb+/- mice C3 expression in the retina was again absent as in the Cfh-/- mice. This is because with the restoration of CFB, C3 convertases are again able to form, but without CFH to regulate them, all the C3 would be broken down. As in the Cfh-/- mice, without C3 on the basal surface of the RPE, no breakdown products were detected at this site. We included CFD mutant mice in our analysis as there are few ocular studies that have used this strain, despite CFD being the key serine protease that cleaves C3bB to form the C3 convertase C3bBb, and with the CFD blocking antibody Lampalizumab now in trials for dry AMD [13,14]. Interestingly, we observed particularly intense staining of C3 on the basal RPE/Bruch’s in the Cfd-/- mouse retina, even though these mice have normal serum levels of C3, suggesting that production of CFD by the RPE might provide local control of C3 build-up under normal circumstances. It was not surprising that we were unable to detect C3 breakdown products on RPE/Bruch’s since without CFD, the C3 convertase was not able to form via the alternative pathway. In future studies it would be interesting to generate a Cfd-/-/Cfh-/- double knock-out mouse to see whether loss of CFH as in the Cfb-/-/Cfh-/- mouse causes activation of classical or lectin pathways and restores C3 breakdown at the RPE/Bruch’s surface. In summary we have shown, using a range of single and double null mutant mice, that the alternative pathway of complement activation is critical in regulating C3 activation on RPE/Bruch’s. We also show that regardless of genotype the effects of disrupting Cfb, Cfh and Cfd on retinal anatomy are mild at one year of age. Some of our observations cannot be readily explained on the basis of systemic complement activation alone, and hint instead at RPE-derived pools having specific roles that may be important in the development of AMD. We are grateful to Marieta Ruseva and Matthew Pickering from Imperial College London for help with the C3 ELISAs, Shannon Conder at UCL for help with cardiac puncture, and Matt Hayes and Peter Munro at the UCL IoO imaging facility, and we acknowledge GSK for an educational grant and the Wellcome Trust for financial support. ==== Refs References 1 Klein RJ , Zeiss C , Chew EY , Tsai JY , Sackler RS , Haynes C , et al (2005 ) Science . Complement factor H polymorphism in age-related macular degeneration . 308 :385 –389 . 15761122 2 Edwards AO , Ritter R 3rd, Abel KJ , Manning A , Panhuysen C , Farrer LA . (2005 ) Complement factor H polymorphism and age-related macular degeneration . Science . 308 :421 –424 . 15761121 3 Haines JL , Hauser MA , Schmidt S , Scott WK , Olson LM , Gallins P , et al (2005 ) Complement factor H variant increases the risk of age-related macular degeneration . Science . 308 :419 –21 . 15761120 4 Hageman GS , Anderson DH , Johnson LV , Hancox LS , Taiber AJ , Hardisty LI , et al (2005 ) A common haplotype in the complement regulatory gene factor H (HF1/CFH) predisposes individuals to age-related macular degeneration . Proc Natl Acad Sci USA . 102 :7227 –7232 15870199 5 Coffey PJ , Gias C , McDermott CJ , Lundh P , Pickering MC , Sethi C , et al (2007 ) Complement factor H deficiency in aged mice causes retinal abnormalities and visual dysfunction . Proc Natl Acad Sci U S A . 104 :16651 –16656 . 17921253 6 Faber C , Williams J , Juel HB , Greenwood J , Nissen MH , Moss SE . (2012 ) Complement factor H deficiency results in decreased neuroretinal expression of Cd59a in aged mice . Invest Ophthalmol Vis Sci . 53 :6324 –6330 . 22918646 7 Williams JA , Greenwood J , Moss SE . (2013 ) Retinal changes precede visual dysfunction in the complement factor H knockout mouse . PLoS One . 8 :e68616 10.1371/journal.pone 23844226 8 Weismann D , Hartvigsen K , Lauer N , Bennett KL , Scholl HP , Charbel Issa P , et al (2011 ) Complement factor H binds malondialdehyde epitopes and protects from oxidative stress . Nature 478 :76 –81 . 10.1038/nature10449 21979047 9 Spencer KL , Hauser MA , Olson LM , Schmidt S , Scott WK , Gallins P , et al (2007 ) Protective effect of complement factor B and complement component 2 variants in age-related macular degeneration . Hum Mol Genet . 16 :1986 –1992 . 17576744 10 Chen M , Muckersie E , Robertson M , Forrester JV , Xu H . (2008 ) Up-regulation of complement factor B in retinal pigment epithelial cells is accompanied by complement activation in the aged retina . Exp Eye Res . 87 :543 –550 . 10.1016/j.exer.2008.09.005 18926817 11 Rohrer B , Coughlin B , Kunchithapautham K , Long Q , Tomlinson S , Takahashi K , et al (2011 ) The alternative pathway is required, but not alone sufficient, for retinal pathology in mouse laser-induced choroidal neovascularization . Mol Immunol . 48 (6–7 ):e1 –8 . 10.1016/j.molimm.2010.12.016 21257205 12 Schnabolk G , Coughlin B , Joseph K , Kunchithapautham K , Bandyopadhyay M , O'Quinn EC , et al (2015 ) Local production of the alternative pathway component factor B is sufficient to promote laser-induced choroidal neovascularization . Invest Ophthalmol Vis Sci . 56 :1850 –1863 . 10.1167/iovs.14-15910 25593023 13 Jack LS , Sadiq MA , Do DV , Nguyen QD . (2016 ) Emixustat and Lampalizumab: Potential Therapeutic Options for Geographic Atrophy . Dev Ophthalmol . 55 :302 –309 . 10.1159/000438954 26501510 14 Rhoades W , Dickson D , Do DV . (2015 ) Potential role of lampalizumab for treatment of geographic atrophy . Clin Ophthalmol . 9 :1049 –1056 . 10.2147/OPTH.S59725 26089637 15 Rohrer B , Guo Y , Kunchithapautham K , Gilkeson GS . (2007 ) Eliminating complement factor D reduces photoreceptor susceptibility to light-induced damage . Invest Ophthalmol Vis Sci . 48 :5282 –5289 . 17962484 16 Law AL , Ling Q , Hajjar KA , Futter CE , Greenwood J , Adamson P , et al (2009 ) Annexin A2 regulates phagocytosis of photoreceptor outer segments in the mouse retina . Mol Biol Cell . 20 :3896 –3904 . 10.1091/mbc.E08-12-1204 19587120 17 Pickering MC , Cook HT , Warren J , Bygrave AE , Moss J , Walport MJ , et al (2002 ) Uncontrolled C3 activation causes membranoproliferative glomerulonephritis in mice deficient in complement factor H . Nat Genet . 31 :424 –428 . 12091909 18 Mullins RF , Russell SR , Anderson DH , Hageman GS . (2000 ) Drusen associated with aging and age-related macular degeneration contain proteins common to extracellular deposits associated with atherosclerosis, elastosis, amyloidosis, and dense deposit disease . FASEB J . 14 :835 –846 . 10783137 19 Anderson DH , Mullins RF , Hageman GS , Johnson LV . (2002 ) A role for local inflammation in the formation of drusen in the aging eye . Am J Ophthalmol . 134 :411 –431 . 12208254 20 Hoh Kam J , Lenassi E , Malik TH , Pickering MC , Jeffery G . (2013 ) Complement component C3 plays a critical role in protecting the aging retina in a murine model of age-related macular degeneration . Am J Pathol . 183 :480 –492 . 10.1016/j.ajpath.2013.04.008 23747511 21 Lundh von Leithner P , Kam JH , Bainbridge J , Catchpole I , Gough G , Coffey P , et al (2009 ) Complement factor h is critical in the maintenance of retinal perfusion . Am J Pathol . 175 :412 –421 . 10.2353/ajpath.2009.080927 19541934 22 Lyzogubov VV , Tytarenko RG , Jha P , Liu J , Bora NS , Bora PS . (2010 ) Role of ocular complement factor H in a murine model of choroidal neovascularization . Am J Pathol . 177 :1870 –1880 . 10.2353/ajpath.2010.091168 20813971 23 Luo C , Zhao J , Madden A , Chen M , Xu H . (2013 ) Complement expression in retinal pigment epithelial cells is modulated by activated macrophages . Exp Eye Res . 112 :93 –101 . 10.1016/j.exer.2013.04.016 23644095 24 Ristau T , Paun C , Ersoy L , Hahn M , Lechanteur Y , Hoyng C , et al (2014 ) Impact of the common genetic associations of age-related macular degeneration upon systemic complement component C3d levels . PLoS One . 9 :e93459 10.1371/journal.pone.0093459 . eCollection 2014. 24675670 25 Warwick A , Khandhadia S , Ennis S , Lotery A . (2014 ) Age-Related Macular Degeneration: A Disease of Systemic or Local Complement Dysregulation? J Clin Med . 3 :1234 –1257 . 10.3390/jcm3041234 26237601

PMC005xxxxxx/PMC5001705.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756467910.1371/journal.pone.0161871PONE-D-16-13964Research ArticleBiology and Life SciencesOrganismsAnimalsVertebratesAmniotesBirdsBiology and life sciencesOrganismsVirusesRNA virusesFlavivirusesWest Nile virusBiology and life sciencesMicrobiologyMedical microbiologyMicrobial pathogensViral pathogensFlavivirusesWest Nile virusMedicine and health sciencesPathology and laboratory medicinePathogensMicrobial pathogensViral pathogensFlavivirusesWest Nile virusBiology and life sciencesOrganismsVirusesViral pathogensFlavivirusesWest Nile virusPeople and placesGeographical locationsSouth AmericaArgentinaBiology and Life SciencesPhysiologyImmune PhysiologyAntibodiesMedicine and Health SciencesPhysiologyImmune PhysiologyAntibodiesBiology and Life SciencesImmunologyImmune System ProteinsAntibodiesMedicine and Health SciencesImmunologyImmune System ProteinsAntibodiesBiology and Life SciencesBiochemistryProteinsImmune System ProteinsAntibodiesMedicine and Health SciencesEpidemiologyDisease VectorsInsect VectorsMosquitoesBiology and Life SciencesOrganismsAnimalsInvertebratesArthropodaInsectsMosquitoesMedicine and Health SciencesInfectious DiseasesInfectious Diseases of the Nervous SystemEncephalitisMedicine and Health SciencesNeurologyInfectious Diseases of the Nervous SystemEncephalitisBiology and Life SciencesEcologyDisease EcologyEcology and Environmental SciencesEcologyDisease EcologyMedicine and Health SciencesEpidemiologyDisease EcologyBiology and Life SciencesOrganismsAnimalsVertebratesActivity Patterns of St. Louis Encephalitis and West Nile Viruses in Free Ranging Birds during a Human Encephalitis Outbreak in Argentina SLEV and WNV Activity Patterns in Urban Birds, Argentinahttp://orcid.org/0000-0001-5953-2907Diaz Luis Adrián 12*Quaglia Agustín Ignacio 1Konigheim Brenda Salomé 1Boris Analia Silvana 1Aguilar Juan Javier 1Komar Nicholas 3Contigiani Marta Silvia 11 Laboratorio de Arbovirus—Instituto de Virología “Dr. J. M. Vanella”–Facultad de Ciencias Médicas–Universidad Nacional de Córdoba, Córdoba, Argentina2 Instituto de Investigaciones Biológicas y Tecnológicas–CONICET–Universidad Nacional de Córdoba, Córdoba, Argentina3 Centers for Diseases Control and Prevention, Fort Collins, Colorado, United States of AmericaCoffey Lark L. EditorUniversity of California Davis, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: LAD MSC. Formal analysis: LAD AIQ NK. Funding acquisition: MSC. Investigation: LAD BSK JJA ASB. Resources: LAD NK MSC. Supervision: LAD MSC. Validation: LAD BSK JJA. Visualization: LAD AIQ. Writing – original draft: LAD AIQ BSK NK MSC. Writing – review & editing: LAD AIQ BSK NK MSC. * E-mail: adrian.diaz@conicet.gov.ar26 8 2016 2016 11 8 e01618716 4 2016 12 8 2016 This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.St. Louis encephalitis virus (SLEV) (Flavivirus) is a reemerging arbovirus in the southern cone of South America. In 2005, an outbreak of SLEV in central Argentina resulted in 47 human cases with 9 deaths. In Argentina, the ecology of SLEV is poorly understood. Because certain birds are the primary amplifiers in North America, we hypothesized that birds amplify SLEV in Argentina as well. We compared avian SLEV seroprevalence in a variety of ecosystems in and around Córdoba city from 2004 (before the epidemic) and 2005 (during the epidemic). We also explored spatial patterns to better understand the local ecology of SLEV transmission. Because West Nile virus (WNV) was also detected in Argentina in 2005, all analyses were also conducted for WNV. A total of 980 birds were sampled for detection of SLEV and WNV neutralizing antibodies. SLEV seroprevalence in birds increased 11-fold from 2004 to 2005. Our study demonstrated that a high proportion (99.3%) of local birds were susceptible to SLEV infection immediately prior to the 2005 outbreak, indicating that the vertebrate host population was primed to amplify SLEV. SLEV was found distributed in a variety of environments throughout the city of Córdoba. However, the force of viral transmission varied among sites. Fine scale differences in populations of vectors and vertebrate hosts would explain this variation. In summary, we showed that in 2005, both SLEV and to a lesser extent WNV circulated in the avian population. Eared Dove, Picui Ground-Dove and Great Kiskadee are strong candidates to amplify SLEV because of their exposure to the pathogen at the population level, and their widespread abundance. For the same reasons, Rufous Hornero may be an important maintenance host for WNV in central Argentina. Competence studies and vector feeding studies are needed to confirm these relationships. Fundacion Bunge y BornDiaz Luis A http://dx.doi.org/10.13039/501100006668Fondo para la Investigación Científica y TecnológicaDiaz Luis A http://dx.doi.org/10.13039/501100006668Fondo para la Investigación Científica y TecnológicaContigiani Marta Silvia Secretaría de Ciencia y Técnica, Universidad Nacional de Córdoba (AR)Diaz Luis A This work was funded by grants from Ministerio de Ciencia y Tecnología de la Nación Argentina (MINCYT, PICT 2013/1779, MSC), Consejo Nacional de Ciencia y Tecnología (CONICET PIP 11220120100544), Secretaría de Ciencia y Tecnología Universidad Nacional de Córdoba (SECYT 203/14, LAD) and Fundación Bunge y Born. AIQ is a recipient of a doctorate scholarship from Council for Scientific and Technical Research (CONICET). LAD´s internship at the CDC Ft. Collins was supported by the International Union of Microbiological Societies (http://www.iums.org/) and International Society of Infectious Diseases (http://www.isid.org/). Data AvailabilityAll relevant data are within the paper.Data Availability All relevant data are within the paper. ==== Body Introduction St. Louis encephalitis virus (SLEV) (Flavivirus, Flaviviridae) is a reemerging arbovirus in the southern cone of South America (Argentina and Brazil) [1,2]. In 2005, an outbreak of St. Louis encephalitis in central Argentina resulted in 47 human cases with 9 deaths [1]. SLEV has been considered a serious public health threat since 1933, when it was first discovered during a large human encephalitis outbreak that initiated in Missouri, USA [3]. In North America, SLEV is maintained through transmission between Culex mosquito vectors and certain passeriform and columbiform birds. In North America, House Sparrow (Passeridae; Passer domesticus) and House Finch (Fringillidae; Haemorhousmexicanus) are thought to be principal amplifiers of SLEV in urban locations, with Mourning Dove (Columbidae; Zenaidamacroura) also contributing in rural regions [3]. In Argentina, the ecology of SLEV is poorly understood. The virus has been isolated from humans (Buenos Aires Province), Culex spp. mosquitoes (Córdoba and Santa Fe provinces) and wild rodents (Córdoba) [4,5]. SLEV has never been isolated from birds in Argentina. However, serological evidence indicates frequent infection of SLEV in Argentine birds [6]. In temperate and subtropical regions of Argentina, neutralizing antibodies were detected in a wide range of birds belonging to several families (Furnariidae, Columbidae, Tyrannidae, Fringillidae, Icteridae, Ardeidae and Cotingidae) [6]. In 2005, the first human encephalitis outbreak attributed to SLEV outside the USA took place in Córdoba Province in central Argentina [1], providing an opportunity to characterize the role of avian hosts in the ecology of SLEV in this locale. Because certain birds are the primary amplifiers in North America, we hypothesized that birds amplify SLEV in Argentina as well. Thus, we should detect a significant increase in viral activity in birds during an epidemic period compared to an enzootic period. To test this hypothesis, we compared avian SLEV seroprevalence in a variety of ecosystems in and around Córdoba city from 2004 (before the epidemic) and 2005 (during the epidemic). We also explored spatial patterns to better understand the local ecology of SLEV transmission. Because West Nile virus (Flaviviridae; WNV) was also detected in Argentina in 2005 [7, 8], and is closely related to SLEV, all analyses were also conducted for WNV. Material and Methods Study area and collection sites Bird captures were carried out during summer and fall of 2004 and 2005 in 4 sites located in Córdoba city (31°24'30”S, 64°11'02” W) (Córdoba Province, Argentina) (Fig 1). This city of 1.3 million inhabitants is situated at 450 m above sea level, and encompasses an area of 576 km2 of which 37.2% is urbanized. The area belongs to the phytogeographic region called Espinal, Chaqueño Domain, or “Chaco Thorn-Forest” [9]. This region is characterized by semi-arid thorn scrub habitat but has been modified intensively by human activities (soy and fruit agriculture, cattle ranching, industrial activity). Isolated patches of natural habitat surrounding the city are comprised of shrub forest. The climate is temperate and semi-arid due to high evapotranspiration in spite of a relatively high precipitation level (750–800 mm) [10]. Four sampling sites were selected based upon accessibility, owners’ authorization and feasibility for mist net use; most of them are located in the periphery of the city of Córdoba (Fig 1): Bajo Grande (BG) sewage treatment plant (31°23′38″ S; 64°04′36″ W). The site is surrounded by aquatic vegetation, reservoirs, low income human settlements and crop lands (vegetables and fruits). Vegetation is dominated by non-native chinaberry (Melia azedarach) and white mulberry (Morus alba) deciduous trees alternating with grasslands. Botanical Garden (BT) (31°23′13″ S; 64°14′58″ W). The vegetation is characterized by patches of aquatic environments, grasslands, croplands and native thorn forest. It is surrounded by low and middle income human settlements. Camino San Carlos (CS) farmland (31°28′46″ S; 64°09′26″ W). Two habitats present include native thorn forest dominated by Acacia sp., Prosopis sp., as well as cinacina (Parkinsonia aculeata), and grassland characterized by Poa sp., Stipa sp., Spanish needle (Biden spilosa) and introduced Chinese privet (Ligustrum lucidum) close to the family house. Villa Gran Parque (VP) (31°19′55″ S; 64°10′28″ W). This vacant lot is surrounded by human settlements and croplands. The dominant habitat is grassland with Spanish needle, Sorghum sp., artichoke thistle (Cynara cardunculus) and alfalfa (Medicago sativa) with scattered white mulberry trees. 10.1371/journal.pone.0161871.g001Fig 1 Geographic location of sampling sites in Córdoba city and neutralizing antibodies seroprevalence (%) in wild birds for St. Louis encephalitis virus and West Nile virus per site and year. BG: Bajo Grande, CS: Camino San Carlos, BT: Botanical Garden, VP: Villa Gran Parque. Bar graphs present seroprevalence values expressed as percentages (number of positive sera/number of analyzed sera). Permission for sampling work in Bajo Grande and Botanical Garden were obtained from the Municipality of Córdoba. Private landowner permission was obtained for study sites at Camino San Carlos and Villa Gran Parque. Bird captures and sample collection Birds were captured with mist nets (AFO Mist Nets, Manomet, Inc., Manomet, MA, USA). Four mist nets per site were installed and operated at dawn and late afternoon. The bird capture was authorized by the Córdoba Province Environmental Agency. Captured birds were identified, weighed, sexed and aged when possible. Wild birds were marked with uniquely numbered aluminum leg bands. Before released, blood-sampled birds were rehydrated with sugar water. Whole blood was collected by jugular (most species) or brachial venipuncture (columbids). Birds that weighed less than 10 grams were not blood-sampled. Blood was placed into a centrifuge tube containing 0.45 mL or 0.9 mL (according to blood sample volume: 0.1 ml or 0.2 ml, respectively) of Minimum Essential Medium (MEM) for an approximate 1:10 serum dilution, held at ambient temperature for 20–30 min for coagulation and placed into coolers. At the laboratory, samples were centrifuged at 5,000g for 15 min for separation of serum. Sera were stored at -20°C. Prior to analysis, sera were heat-inactivated for 30 minutes at 56°C to inactivate non-specific inhibitors of virus neutralization. Ethics Statement The bird capture was authorized by the Córdoba Province Environmental Agency (586869-053-107). Birds were handled following guidelines for the use of wild birds in research elaborated by the Ornithological Council (http://naturalhistory.si.edu/BIRDNET/documents/guidlines/Guidelines_August2010.pdf). Field studies did not involve endangered or protected species. All blood sampling procedures were specifically approved as part of the obtaining field permit. Viral stocks preparation Low-passage SLEV CbaAr-4005 and WNV NY99-4132 strains were used for serologic assays. CbaAr-4005strain of SLEV was isolated from Culex quinquefasciatus mosquitoes collected in Córdoba during the human encephalitis outbreak of 2005 [5]. NY99-4132 strain of WNV was obtained from the brain of an American Crow (Corvus brachyrhynchos) collected in New York during 1999. Viral stocks were obtained from infected Vero cell monolayers harvested on day 7 and 5 post-inoculation for SLEV and WNV, respectively. Serological assays and data interpretation Avian serum samples were screened for the presence of WNV- and SLEV-reactive antibodies by the plaque-reduction neutralization test (PRNT). A suspension of approximately 100 plaque-forming units (PFU) of virus in 75 μL of MEM was added to an equal volume of diluted avian serum, bringing the final serum dilution to 1:20. The mixture was incubated for 1 hour at 37°C. Vero cell monolayers grown in 24-well culture plates (Costar, Cambridge, MA, USA) were inoculated with 0.1 mL of the serum-virus mixture and incubated for 1 hour at 37°C, 5% CO2. Cells were overlaid with an initial 1 mL of 0.5% agarose in M-199 medium supplemented with 350 mg/L sodium bicarbonate, 29.2 mg/L L-glutamine, and antibiotics (penicillin, streptomycin, gentamycin, and amphotericin B). After 3 and 6 days of incubation with WNV and SLEV, respectively, cultures were overlaid again with the same nutrient–agarose mix but also containing 0.004% neutral red for staining and visualization of viral plaques. Plaques were counted the following day. Serum samples that neutralized > 80% of the challenge virus (relative to a serum-free control) were selected for further titration against both WNV and SLEV. Flavivirus titers of serum samples that tested positive in screen tests were determined as follows. Seven serial two-fold dilutions of serum in MEM were prepared beginning with a dilution of 1:10. Virus mixtures were added as described above, resulting in final serum dilutions of 1:20, 1:40, 1:80, 1:160, 1:320, 1:640 and 1:1280. Endpoint titers were assigned as the reciprocal of the greatest dilution in which >80% neutralization of the challenge virus was achieved. Experiments that evaluated cross reaction among SLEV and WNV in heterologous inoculation scenarios in Common Quail (Coturnix coturnix) indicated no cross reaction among SLEV and WNV (Contigiani MS personal communication).Based on this evidence, plus studies by Patiris et al. [11] and Ledermann et al. [12], we considered all serum samples with antibody titers higher than 20 positive. Therefore samples with titers higher than 20 for both viruses were considered as multiple heterologous infections. Statistical analysis Infection prevalence with 95% confidence intervals were calculated for both viruses. Infection proportions within the year and month of sampling were compared using the Fisher exact test or Pearson chi-square test. P-values were considered significant at a threshold of α = 0.05. The homology between spatial activity for SLEV and the bird community was explored by means of a Procrustes analysis [13]. First, we selected a subset of 6 species that accounted for the majority of captures (51%) and 70% of SLEV seroprevalence, which included Eared Dove (Zenaida auriculata), Picui Ground-Dove (Columbina picui), Rufous Hornero (Furnarius rufus), House Sparrow (Passer domesticus), Great Kiskadee (Pitangus sulphuratus) and Bay-winged Cowbird (Agelaioides badius). These 6 species were represented in all sampled sites. Abundance and absolute prevalence matrix were built and data exploration showed the need to remove VP from the dataset (due to double zeros abundance). These frequency matrices were used to produce unconstrained ordinations analysis with chi-square distance (community assemblage matrix) and Non-metric multidimensional scaling (NMDS) (SLEV seroprevalence matrix). Finally, the ordination configuration homology was estimated by a Procrustes analysis and its statistical significance was calculated after 9999 permutations [13]. All analyses were run using vegan and core packages within the R platform [14,15,16]. In order to explore the effect of new species detected as infected on the overall avian seroprevalence among the study periods, we compared the proportion of infected birds of the selected species (same subset as used for the spatial analysis) vs. the proportion of infected birds of the non-selected species, for a given period. Results A total of 980 birds were sampled for detection of SLEV- and WNV-neutralizing antibodies. Analyzed sera belonged to 65 species and 27 families of free-ranging birds. SLEV seroprevalence in birds increased 11-fold from 2004 to 2005 (Table 1). Although the difference in seroprevalence observed between years was significant (Fisher exact test, p <0.001), no significant difference was detected between months within each year (2004, χ2 = 1.51, df = 3, p = 0.68; 2005, χ2 = 0.44, df = 2, p = 0.80). During January-April 2004 only 3 birds tested positive for SLEV-neutralizing antibodies (0.7%). SLEV activity was found in 3 resident (i.e. non-migratory) bird species: Rufous Hornero (2.0%, 1/49), Bay-winged Cowbird (8.3%, 1/12) and Great Kiskadee (1.2%, 1/80). These three seropositive individuals were distributed among three of the four study sites (Fig 1). One of these (Rufous Hornero) seroconverted from January to March (Table 2). Of 14 other recaptured birds in 2004, none seroconverted. In the second period (January-March 2005), an overall SLEV seroprevalence of 7.7% (42/543) was registered, with seropositive birds detected at all four study sites. Multiple infections for SLEV and WNV were detected in 6 serum samples (1.1%).Neutralizing antibody titer ranged from 40 to 1280, with titers of 40 and 160 the most frequently detected. SLEV seroprevalence increased between years in all three paired-year sites, showing significance differences between 2004 and 2005 (BG: χ2 = 136; p<0.001; CS: χ2 = 58; p = 0.01 and VP: χ2 = 56.35; p = 0.01) (Fig 1). Differences were also observed among sites within 2005 (χ2 = 24.53; df = 3; p = 0.006). 10.1371/journal.pone.0161871.t001Table 1 St. Louis encephalitis virus-neutralizing antibody prevalence among all bird species combined, by months and years, Córdoba, Argentina. Month 2004 2005 Pos/Total %[CI] Pos/Total %[CI] January 1/110 0.90[0.05–5.69] 12/145 8.28[4.54–14.32] February 0/137 0.00 [0.00–3.40] 15/178 8.43[4.96–13.77] March 1/76 1.31[0.07–8.11] 15/220 6.81[4.00–11.21] April 1/114 0.88[0.05–5.50] ND ND Total 3/437 0.69[0.18–2.17]* 42/543 7.73[5.69–10.39]* *p-value<0.001. ND: no data available. CI: 95% confidence interval. 10.1371/journal.pone.0161871.t002Table 2 Seroprevalence among free-ranging birds collected in Córdoba city during 2005 and tested by PRNT for St. Louis encephalitis- and West Nile virus-reactive antibodies. Family Species Migratory Status SLEV WNV Pos/Tested %[CI] Pos/Tested %[CI] Falconidae American Kestrel (Falco sparverius) Resident 0/2 0 [0–65.76] 1/2 50 [9.45–90.55] ChimangoCaracara(Milvago chimango) Resident 1/1 100 [20.65–100] 0/1 0 [0–79.35] Columbidae Eared Dove (Zenaidaauriculata) Resident 4/33 12.12 [4.82–27.33] 0/33 0 [0–10.43] Picui Ground-Dove (Columbinapicui) Resident 10/65 15.38 [8.58–26.06] 0/65 0 [0–5.58] Picidae Green-barred Woodpecker (Colaptesmelanochloros) Resident 2/6 33.33 [9.68–70] 0/6 0 [0–39.03] Furnariidae Brown Cacholote (Pseudoseisuralophotes) Resident 2/7 28.57 [8.22–64.11] 0/7 0 [0–35.43] Rufous Hornero (Furnariusrufus) Resident 2/54 3.7 [1.02–12.54] 5/54 9.26 [4.02–19.91] Tyrannidae 0 [0–19.36] Elaenia spp. (E. albiceps/parvirostris) Migratory 1/16 6.25 [1.11–28.33] 0/16 Great Kiskadee(Pitangussulphuratus) Resident 12/91 13.19 [7.71–21.65] 1/91 1.10 [0.19–5.96] Vermilion Flycatcher (Pyrocephalusrubinus) Migratory 1/8 12.50 [2.24–47.09] 0/8 0 [0–32.44] Turdidae Creamy-bellied Thrush (Turdusamaurochalinus) Resident 1/20 5 [0.89–23.61] 0/20 0 [0–16.11] Thraupidae Black-and-rufous Warbling-Finch (Poospizanigrorufa) Resident 1/7 14.29 [2.57–51.31] 0/7 0 [0–35.43] Black-crested Finch (Lophospinguspusillus) Resident 1/1 100 [20.65–100] 0/1 0 [0–79.35] Emberizidae Rufous-collared Sparrow (Zonotrichiacapensis) Resident 1/9 11.11 [1.99–43.5] 0/9 0 [0–29.91] Icteridae Bay-winged Cowbird (Agelaioidesbadius) Resident 1/34 2.94 [0.52–14.92] 0/34 0 [0–10.15] Passeridae HouseSparrow(Passerdomesticus) Resident 2/51 3.92 [1.08–13.22] 1/51 1.96 [0.35–10.30] TOTAL 42/543 7.73 [5.77–10.29] 8/543 1.47 [0.75–2.88] Seronegative species belonging to seropositive families: Columbidae: Patagioenasmaculosa(1), Leptotilaverreauxi(2); Furnariidae: Asthenesbaeri(1), Tyrannidae: Knipolegusaterrimus(1), K. striaticeps (1), Machetornisrixosus (6), Myiarchustyrannulus(1), Pseudocolopteryxacutipennis(2), Serpophagasubcristata(2), Tyrannus savanna (1), Pachyramphuspolichopterus(1); Non identified (9); Turdidae: Turdusrufiventris(1); Thraupidae: Saltatoraurantirostris (2), Pipraeideabonariensis(1), Coryphospinguscucullatus(1), Embernagraplatensis (1), Poospiza ornate (1), Sicalisflaveola(12), S. luteola(1), Sporophilacaerulescens (10); Emberizidae: Rhynchospizastrigiceps(1), Ammodramushumeralis(1); Icteridae: Chrysomusruficapillus (2), Molothrusrufoaxillaris(9), M. bonariensis(10), Agelasticusthilius (3), Chrysomusruficapillus(2), Non identified (1). Seronegative families and species: Charadriidae: Vanelluschilensis(1); Psittacidae: Myiopsittamonachus (2); Cuculidae: Coccyzusmelacoryphus (4); Bucconidae: Nystalusmaculatus(1); Thamnophilidae: Taraba major (3); Cotingidae: Phytotomarutila (9); Vireonidae: Cyclarhisgujanensis(1); Troglodytidae: Troglodytes aedon (6); Mimidae: Mimustriurus (21); Parulidae: Geothlypisaequinoctialis (2); Fringillidae: Spinusmagellanica (4) In 2005, the number of SLEV-seropositive bird species increased from 3 in the previous year to 15 (Table 3). The avian families with the highest seroprevalence for SLEV-neutralizing antibodies were: Columbidae (14.0%; 14/101); Tyrannidae (10.1%; 14/138); Furnariidae (6.3%; 4/63); Thraupidae (5.5%; 2/36); Turdidae (4.5%; 1/22); Passeridae (3.9%; 2/51); and Icteridae (1.7%; 1/59). The seroprevalence by bird species ranged between 0% and 28.6%. Highest values were observed in Brown Cacholote (Pseudoseisura lophotes, 28.6%; 2/7) and Picui Ground-Dove (15.4%; 10/65). Two out of 25 recaptured birds seroconverted for SLEV during the January-March 2005 period (Table 2). 10.1371/journal.pone.0161871.t003Table 3 St. Louis encephalitis and West Nile virus seroconversion events in recaptured resident birds during 2004 and 2005. Species Site Months sampled(*) PRNT80titer Virus SLEV WNV 1st sample 2nd sample 1st sample 2nd sample Rufous Hornero CS Jan-Mar '04 (56) <20 40 <20 <20 SLEV Totalrecapturedindividuals 2004 15 Bay-wingedCowbird BG Jan-Feb '05 (31) <20 80 <20 <20 SLEV Great Kiskadee BG Jan-Mar '05 (57) <20 640 <20 <20 SLEV Great Kiskadee BT Feb-March'05 (54) <20 <20 <20 40 WNV Rufous Hornero CS Jan-Feb '05 (34) <20 <20 <20 640 WNV Rufous Hornero CS Jan-Mar '05 (71) <20 <20 <20 640 WNV Rufous Hornero BG Jan-Mar '05 (58) <20 <20 <20 80 WNV Total recaptured individuals 2005 25 *Days between date capture-recaptured Bird community assemblage ordination showed pairwise overlapping (BG-CS, BG-BT) total inertia 0.62. Two first axes explained 78% of the variation (Fig 2A). Great Kiskadee is a common element in the BG-CS paired cluster. Picui Ground-Dove, Eared Dove and Bay-winged Cowbird/House Sparrow abundances are key species in the ordination of CS, BG and BT respectively. Rufous Hornero was common among all sites analyzed. Two convergent solutions were found after six iterations running NMDS analyses (stress: 0.1267; non-metric fit R2 = 0.9840; RMSE = 2.77−5). SLEV prevalence ordination shows a complete overlapping among analyzed sites suggesting a similar composition of infected bird species (Fig 2B). In site BG, infected individuals of all six selected species were detected. Infected Shiny Cowbird (Molothrus bonariensis) was only detected here. Sites BT and CS share the absence of infection in Shiny Cowbird, Eared Dove and Rufous Hornero (Fig 2A). Procrustes test was not significant (m2 = 0.7775; t0 = 0.4717; p-value = 0.3456) and failed to detect an association pattern between SLEV prevalence and bird abundance by sites. 10.1371/journal.pone.0161871.g002Fig 2 Ordination analyses based on a subset of 6 bird species selected collected and tested in three sites of Córdoba city from January to March 2005. A) Correspondence analysis ordination bi-plot for bird species abundance derived from frequency of capture in mist-nets. For each axis, the amount of variation explained as a part of the total variation in the model is shown. B) Non-metric multidimensional scaling (NMDS) analysis ordination bi-plot for SLEV seroprevalence. Colored areas represent amount of variation for analyzed variables. During 2005, whereas the SLEV seroprevalence among birds appeared to be relatively stable, with no significant change among sampling months, the number of infected bird species increased throughout the sampling period. Moreover, the relative importance of the 6 selected species used in the spatial analysis diminished from January to March as more species were detected as seropositive each month (Fig 3A). The monthly infected bird species richness increased by 67% from January to March (Fig 3B). Only Picui Ground-Dove, Great Kiskadee and Eared Dove were detected as seropositive in all three months surveyed in 2005. Picui Ground-Dove and Great Kiskadee together contributed the largest number of infected individuals by month (approximately 50% of all seropositive birds each month). 10.1371/journal.pone.0161871.g003Fig 3 St. Louis encephalitis virus host infection dynamics during January-March 2005. A) SLEV infection prevalence and the ratio of selected vs non-selected infected bird species. B) Species composition of seropositive birds for SLEV. No WNV-neutralizing antibody was detected during January–April, 2004, among 437 bird sera analyzed. However, during January–March, 2005, the WNV seroprevalence observed was 1.5% (8/543, binomial 95% CI: 0.6–2.9%). Highest WNVseroprevalence was observed in March (2.3%; 95%CI = 1.7% - 9.0%;5/125) compared to January (0.9%;95% CI = 0.2% - 4.8%; 1/114) and February (0.5%; 1/209; 95%CI = 0.1% - 2.7%).WNV seropositive birds were observed in BG (1.3%; 2/160; 95%CI = 0.3% - 4.4%), BT (1.1%; 2/176; 95%CI = 0.3% - 4.1%) and CS (2.0%; 3/151; 95%CI = 0.7% - 5.7%). Only four out of 52 bird species were positive for WNV-neutralizing antibodies: American Kestrel (Falco sparverius); Rufous Hornero; House Sparrow; and Great Kiskadee (Table 2). Four WNV seroconversion events were registered out of 25 opportunities in 2005, most of them in Rufous Hornero (Table 3). Discussion Arbovirus activity is influenced in a complex manner by several biological and environmental factors [17]. As in North America, it appears thatcertain birds play an important role in the amplification of SLEV during outbreaks of human disease. Our study indicated that a high proportion (99.3%) of local birds were susceptible to SLEV infection immediately prior to the 2005 outbreak, indicating that the vertebrate host population was primed to amplify SLEV. Another biologic factor that may have promoted the SLEV outbreak could be the sustained increase in abundance of the main vector Culex quinquefasciatus which has been documented elsewhere [18,19]. In temperate areas, arbovirus transmission is highly influenced by mosquito vector abundance [20,21]. In urban sites of central Argentina, SLEV is mainly transmitted by Culex quinquefasciatus [22] and potentially also by Culex interfor [5]. These mosquitoes’ populations have two annual peaks: a small one in late spring-early summer (November-December) and a large one in late summer-early fall (February-March) [18,19]. Based on this vector pattern one would expect the presence of two peaks of SLEV activity, which indeed was observed during this human encephalitis outbreak [1]. The high avian SLEV seroprevalence detected in January, followed by no discernable increase in antibody prevalence during February and March suggests that much of the transmission to birds had already occurred by January 2005 or earlier, which corresponds with the earlier peak of vector abundance. Some transmission continued in birds after January as evidenced by two seroconversion events (Table 3) and an expanding number of seropositive bird species (Fig 3B). The apparent reduction in avian seroprevalence for SLEV in March 2005 can be explained in several ways: 1. By March, conditions for SLEV amplification among birds were no longer favorable, and the detectable seroprevalence dropped as a consequence of normal bird movements (emigration of seropositive individuals and immigration of seronegative individuals from regions where SLEV was not active); 2. Diminishing detectability of SLEV antibodies as birds aged and their antibodies, derived from infections several months earlier, began to wane to undetectable levels. SLEV antibodies in some birds infected during December to January could become undetectable in March due to diminishing neutralizing antibody titers below the detection threshold (20) [23,24]. SLEV was found distributed in a variety of environments (urbanized neighborhoods as well as patches of native habitat) throughout the city of Córdoba, indicating that ecological requirements for its maintenance are present. However, variation in the force of viral transmission among sites was observed. The highest activity was registered in the eastern peri-urban site (BG). This area contains large polluted water reservoirs and trash dumps, which support an abundant population of Culex quinquefasciatus. These landscape elements agree with those associated with risk of human infection identified by Vergara Cid et al. [25]. Differences in transmission among sites would be due to fine scale differences in populations of vectors and vertebrate hosts. Infected bird species found during the SLEV epidemic period belong to the same families (Columbidae, Furnariidae, Icteridae, Tyrannidae) as those detected during enzootic periods further north in Argentina [6]. Presumably, a strain of SLEV would utilize similar vector and vertebrate host species for its maintenance/amplification throughout its geographic range. Hence, spatiotemporal variation in transmission would result from differing ecological features such as vector and vertebrate host abundance and distribution, habitat use and population dynamics. Within the epidemic period, we observed an increase in the number of SLEV-infected bird species between months (Fig 3). Whether this was due to a change in transmission ecology (i.e. a shift of transmission to novel amplifier hosts) or spillover to tangential avian hosts remains unknown. However, infected individuals of Picui Ground-Dove and Great Kiskadee were reliably present at all sampling sites. Multivariate analysis used in the geospatial analyses compared the bird species composition and SLEV seroprevalence among the study sites, but failed to detect a clear separation pattern. In our study, avian community assemblage varied among sites (Fig 2A) yet SLEV seroprevalence ordination overlapped with all sites. The overlapping pattern observed in SLEV seroprevalence could be explained by the presence of identical/similar mosquito vector(s) with similar host preferences. Such was the case for WNV in the eastern USA where viral activity is driven by mosquito host preference [26]. Although WNV and SLEV share ecological requirements, this hypothesis must be corroborated for SLEV. Eared Dove and Picui Ground-Dove (Columbidae), Great Kiskadee (Tyrannidae) and Brown Cacholote (Furnariidae) had high seroprevalence (>12%), are widely dispersed and are abundant in several ecosystems (urban and rural habitats). Moreover, seropositive individuals of Picui Ground-Dove and Great Kiskadee were frequently detected during the epidemic period (Fig 3). These features make these species candidates for amplifying hosts of SLEV. Experimental infections of adult Eared Dove demonstrated this species’ competence to infect a local strain of Culex quinquefasciatus with the enzootic 78V-6507 strain of SLEV [27]. Another interesting species is Rufous Hornero. Although its infection prevalence was low (3.7%), it is abundant in urban and rural areas. Since it is highly territorial and easy to collect with mist nets, and it is frequently infected with both SLEV and WNV, it could represent an important amplifier as well as a good sentinel for flavivirus transmission activity. The House Sparrow is the main SLEV host in largely urban areas of the USA[3]. However, previous reports from Argentina highlighted the absence of SLEV-infected House Sparrow (more than 200 sera analyzed) and questioned its role as amplifying host for SLEV in Argentina [6]. For the first time in South America, our study detected SLEV-infected individuals of House Sparrow (seroprevalence<5%). Host competence assays carried out with North American House Sparrow and SLEV strains isolated in South America suggested a poor amplification potential for House Sparrow [28,29]. However, without data on host competence derived from infection studies matching Argentinean strains of SLEV with a sympatric population of House Sparrow, we cannot speculate regarding its role as an amplifier of SLEV in Argentina. The first evidence of autochthonous circulation for West Nile virus (WNV) in the American continent was detected in 1999 in New York City [30]. Thereafter, it rapidly spread throughout the New World [31,32]. In Argentina this virus caused epizootic events in equines [7]. Molecular characterization indicated that the introduced strain in Argentina belongs to the I2 lineage (NY99 strain cluster) [33]. A serosurvey of free-ranging birds indicated widespread transmission, with no evidence for disease detected in Argentinean birds [8]. The earliest detection of WNV-neutralizing antibodies in birds from Argentina was January 2005, indicating that WNV circulation in this country had begun by the end of 2004. In the present study, we document four birds that seroconverted to WNV sometime between January and March 2005. One of them was negative in January and had seroconverted to positive 34 days later in February. Host competence assays carried out in urban birds have indicated that the Picui Ground-Dove is competent to amplify WNV [34]. However none of 65 representatives of this species was found to have been infected by WNV in Córdoba. In summary, we showed that in 2005, both SLEV and to a lesser extent WNV circulated in a variety of avian species. Seroprevalence combined with avian abundance may be used to identify candidate amplifier hosts. Competence studies are needed to verify the importance of these candidates as amplifiers, as well as vector feeding studies to confirm vector-amplifier host contact. Eared Dove, Picui Ground-Dove and Great Kiskadee are strong candidates to amplify SLEV. Rufous Hornero and Brown Cacholote are also frequently infected and may be useful sentinels because of their local residence, territoriality and ease of capture. Rufous Hornero may be an important maintenance host for WNV in central Argentina. Authors want thanks to Eric Edwards for his invaluable support in serological determinations at CDC (Ft. Collins). AIQ is a recipient of a doctorate scholarship from Council for Scientific and Technical Research (CONICET). LAD´s internship at the CDC Ft. Collins was supported by the International Union of Microbiological Societies (http://www.iums.org/) and International Society of Infectious Diseases (http://www.isid.org/). ==== Refs References 1 Spinsanti LI , Díaz LA , Glatstein N , Arselán S , Morales MA , Farías AA , et al Human outbreak of St. Louis encephalitis detected in Argentina, 2005 . J ClinVirol . 2008 ; 42 :27 –33 . 10.1016/j.jcv.2007.11.022 . 2 Rocco IM , Santos CL , Bisordi I , Petrella SM , Pereira LE , Souza RP , et al St. Louis encephalitis virus: first isolation from a human in São Paulo State, Brazil . Rev Inst Med Trop Sao Paulo . 2005 ;47 :281 –285 . .16302112 3 Reisen WK . Epidemiology of St. Louis encephalitis virus . Adv Virus Res . 2003 ;61 :139 –183 . .14714432 4 Sabattini MS , Aviles G , Monath TO . Historical, epidemiological and ecological aspects of arboviruses in Argentina: Flaviviridae, Bunyaviridae and Rhabdoviridae In: Travassos da Rosa APA , Vasconcelos PFC , Travassos da Rosa JFS , editors. An Overview of Arbovirology in Brazil and Neighbouring Countries . Belem , Instituto Evandro Chagas , 1998 pp.113 –134 . 5 Diaz LA , Ré V , Almirón WR , Farías A , Vázquez A , Sánchez-Seco MP , et al Genotype III Saint Louis encephalitis virus outbreak, Argentina, 2005 . Emerg Infect Dis . 2006 ;12 :1752 –1754 ..17283629 6 Monath TP , Sabattini MS , Pauli R , Daffner JF , Mitchell CJ , Bowen GS , et al Arbovirus investigations in Argentina, 1977–1980. IV. Serologic surveys and sentinel equine program . Am J Trop Med Hyg . 1985 ;34 :966 –975 . .2863991 7 Morales MA , Barrandeguy M , Fabbri C , Garcia JB , Vissani A , Trono K , et al West Nile virus isolation from equines in Argentina, 2006 . Emerg Infect Dis . 2006 ;12 :1559 –1561 . .17176571 8 Diaz LA , Komar N , Visintin A , Dantur Juri MJ , Stein M , Lobo Allende R , et al West Nile virus in birds, Argentina . Emerg Infect Dis . 2008 ;14 :689 –691 . 10.3201/eid1404.071257 .18394305 9 Cabrera AL (1971 ) Fitogeografía de la República Argentina . Bol Soc Argent Bot . 14 :1 –42 . 10 Jarsún B, JA Gorjas, E Zamora, E Bosnero. Caracterización general de la provincia. In: Gorjas JA and Tassile JL, editors. Recursos Naturales de la Provincia de Córdoba, Córdoba, BR Copias.2003. pp. 23–60. 11 Patiris PJ , Oceguera LF , Peck GW , Chiles RE , Reisen WK , Hanson CV Serologic diagnosis of West Nile and St. Louis encephalitis virus infections in domestic chickens . Am J Trop Med Hyg . 2008 ;78 :434 –441 . .18337340 12 Ledermann J , Loroño-Pino M , Ellis C , Shaw K , Blitvich B , Beaty B , et al The evaluation of widely used diagnostic tests to detect West Nile virus infections in horses previously infected with St. Louis encephalitis or dengue virus . Clin Vaccine Inmunol . 2011 ;18 : 580 –587 . 13 Stevens MHH . A primer of ecology with R. Chapter 10: Community composition and diversity New York Springer ; 2009 . 14 Goslee SC , Urban DL . The ecodist package for dissimilarity-based analysis of ecological data . J Stat Software . 2007 ;22 :1 –19 . 15 R Development Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. 2008. 16 Oksanen J, Blanchet FG, Kindt R, Legendre P, Minchin PR, O'Hara RB, et al. 2013. Vegan: Community Ecology Package. R package version 2.0–10. http://CRAN.R-project.org/package=vegan. 17 Reisen WK . Landscape epidemiology of vector-borne diseases . Ann Rev Entomol . 2010 ;55 :461 –483 . 10.1146/annurev-ento-112408-085419 .19737082 18 Díaz LA , Almirón WR , Ludueña Almeida F , Spinsanti LI , Contigiani MS . Vigilancia del virus encefalitis de San Luis y mosquitos (Diptera: Culicidae) en la provincia de Córdoba, Argentina . EntomolVect . 2003 ;10 :551 –566 . 19 Batallán GP , Estallo EL , Flores FS , Sartor P , Contigiani MS , Almirón WR . St. Louis encephalitis virus mosquito vectors dynamics in three different environments in relation to remotely sensed environmental conditions .Acta Trop . 2015 ;146 :53 –59 . 10.1016/j.actatropica.2015.03.009 25792419 20 Reisen WK , Fang Y , Martinez VM . Avian host and mosquito (Diptera: Culicidae) vector competence determine the efficiency of West Nile and St. Louis encephalitis virus transmission . J Med Entomol . 2005 ;42 :367 –375 . .15962789 21 Barker CM , Johnson WO , Eldridge BF , Park BK , Melton F , Reisen WK . Temporal connections between Culextarsalis abundance and transmission of western equine encephalomyelitis virus in California . Am J Trop Med Hyg . 2010 ;82 :1185 –1193 . 10.4269/ajtmh.2010.09-0324 .20519621 22 Diaz LA , Flores FS , Beranek M , Rivarola ME , Almirón WR , Contigiani MS . Transmission of endemic St Louis encephalitis virus strains by local Culexquinquefasciatus populations in Cordoba, Argentina . Trans R Soc Trop Med Hyg . 2013 ;107 :332 –334 . 10.1093/trstmh/trt023 .23474474 23 Reisen WK , Kramer LD , Chiles RE , Green EN , Martinez VM . Encephalitis virus persistence in California birds: Preliminary studies with house finches (Carpodacusmexicanus) . J Med Entomol . 2001 ;38 :393 –399 . .11372964 24 Reisen WK , Chiles RE , Martinez VM , Fang Y , Green EN . Encephalitis virus persistence in California birds: experimental infections in mourning doves (Zenaiduramacroura) . J Med Entomol . 2004 ;41 :462 –466 . .15185951 25 Vergara Cid C , Estallo EL , Almirón WR , Contigiani MS , Spinsanti LI . Landscape determinants of Saint Louis encephalitis human infections in Córdoba city, Argentina during 2010 . Acta Trop . 2013 ;125 :303 –308 . 10.1016/j.actatropica.2012.12.005 .23257661 26 Kilpatrick AM , Kramer LD , Jones MJ , Marra PP , Daszak P . West Nile virus epidemics in North America are driven by shifts in mosquito feeding behavior . PLoS Biol . 2006 ;4 (4 ):e82 .16494532 27 Diaz LA , Occelli M , Almeida FL , Almirón WR , Contigiani MS . Eared dove (Zenaidaauriculata, Columbidae) as host for St. Louis encephalitis virus (Flaviviridae, Flavivirus) .Vector Borne Zoonotic Dis . 2008 ;8 :277 –282 . 10.1089/vbz.2007.0168 .18429694 28 Bowen GS , Monath TP , Kemp GE , Kerschner JH , Kirk LJ . Geographic variation among St. Louis encephalitis virus strains in the viremic responses of avian hosts . Am J Trop Med Hyg . 1980 ;29 :1411 –1419 . .7446828 29 Diaz LA , Nemeth NM , Bowen RA , Almiron WR , Contigiani MS . Comparison of argentinean Saint Louis encephalitis virus non-epidemic and epidemic strain infections in an avian model . PLoSNegl Trop Dis . 2011 ;5 (5 ):e1177 10.1371/journal.pntd.0001177 .21629729 30 Hayes EB , Komar N , Nasci RS , Montgomery SP , O'Leary DR , Campbell GL . Epidemiology and transmission dynamics of West Nile virus disease . Emerg Infect Dis . 2005 ;11 :1167 –1173 . .16102302 31 Komar N , Clark GG . West Nile virus activity in Latin America and the Caribbean . RevPanam Salud Publica . 2006 ;19 :112 –117 . .16551385 32 DiazL A ,Quaglia A , Flores FS , Contigiani MS .Virus West Nile en Argentina: un agente infeccioso emergente que plantea nuevos desafíos . Hornero . 2011 ;26 :5 –28 . 33 Fabbri CM , Morales MA , García JB , Luppo VC , Levis S . Análisis filogenético de una cepa de virus West Nile caracterizada de un equino con sintomatología neurológica de la provincia de Córdoba, Argentina . RevArgMicrobiol .2011 ;S1,43 :90 . 34 Diaz LA , Flores FS , Contigiani MS . Viremia profiles and host competence index for West Nile virus (Flavivirus, Flaviviridae) in three autochthonous birds species from Argentina . J Ornithol .2011 ;152 :21 –25 . 10.1007/s10336-010-0538-4

PMC005xxxxxx/PMC5001706.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756424010.1371/journal.pone.0161892PONE-D-16-17368Research ArticleBiology and Life SciencesOrganismsBacteriaActinobacteriaMycobacterium TuberculosisMedicine and Health SciencesInfectious DiseasesBacterial DiseasesTuberculosisMedicine and Health SciencesTropical DiseasesTuberculosisPeople and PlacesGeographical LocationsAfricaGhanaBiology and Life SciencesBiogeographyPhylogeographyEcology and Environmental SciencesBiogeographyPhylogeographyEarth SciencesGeographyBiogeographyPhylogeographyBiology and Life SciencesEvolutionary BiologyPopulation GeneticsPhylogeographyBiology and Life SciencesGeneticsPopulation GeneticsPhylogeographyBiology and Life SciencesPopulation BiologyPopulation GeneticsPhylogeographyMedicine and Health SciencesInfectious DiseasesZoonosesBovine TuberculosisMedicine and Health SciencesInfectious DiseasesInfectious Disease ControlBiology and Life SciencesMicrobiologyMedical MicrobiologyMicrobial PathogensViral PathogensImmunodeficiency VirusesHIVMedicine and Health SciencesPathology and Laboratory MedicinePathogensMicrobial PathogensViral PathogensImmunodeficiency VirusesHIVBiology and Life SciencesOrganismsVirusesViral PathogensImmunodeficiency VirusesHIVBiology and Life SciencesOrganismsVirusesImmunodeficiency VirusesHIVBiology and life sciencesOrganismsVirusesRNA virusesRetrovirusesLentivirusHIVBiology and Life SciencesMicrobiologyMedical MicrobiologyMicrobial PathogensViral PathogensRetrovirusesLentivirusHIVMedicine and Health SciencesPathology and Laboratory MedicinePathogensMicrobial PathogensViral PathogensRetrovirusesLentivirusHIVBiology and Life SciencesOrganismsVirusesViral PathogensRetrovirusesLentivirusHIVPeople and PlacesGeographical LocationsAfricaCameroonSpatio-Temporal Distribution of Mycobacterium tuberculosis Complex Strains in Ghana Distribution of Mycobacteria Strainshttp://orcid.org/0000-0002-1663-6287Yeboah-Manu Dorothy 1*Asare P. 1Asante-Poku A. 1Otchere I. D. 1Osei-Wusu S. 1Danso E. 1Forson A. 2Koram K. A. 1Gagneux Sebastien 341 Noguchi Memorial Institute for Medical Research, University of Ghana, Legon, Accra, Ghana2 Department of Chest Diseases, Korle-Bu Teaching Hospital, Korle-bu, Accra, Ghana3 Department of Medical Parasitology and Infection Biology, Swiss Tropical and Public Health Institute, Basel, Switzerland4 University of Basel, Basel, SwitzerlandSreevatsan Srinand EditorUniversity of Minnesota, UNITED STATESCompeting Interests: The authors have declared that no competing interest exist. Conceptualization: DYM SG. Formal analysis: IDO PA AAP. Funding acquisition: DYM. Investigation: ED IDO DYM AAP SO. Methodology: DYM AAP KK. Project administration: DYM. Resources: AF DYM KK. Supervision: DYM SG KK. Writing – original draft: DYM AAP IDO PA SG. Writing – review & editing: DYM AAP IDO PA SG. * E-mail: dyeboah-manu@noguchi.ug.edu.gh26 8 2016 2016 11 8 e016189229 4 2016 12 8 2016 © 2016 Yeboah-Manu et al2016Yeboah-Manu et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Background There is a perception that genomic differences in the species/lineages of the nine species making the Mycobacterium tuberculosis complex (MTBC) may affect the efficacy of distinct control tools in certain geographical areas. We therefore analyzed the prevalence and spatial distribution of MTBC species and lineages among isolates from pulmonary TB cases over an 8-year period, 2007–2014. Methodology Mycobacterial species isolated by culture from consecutively recruited pulmonary tuberculosis patients presenting at selected district/sub-district health facilities were confirmed as MTBC by IS6110 and rpoß PCR and further assigned lineages and sub lineages by spoligotyping and large sequence polymorphism PCR (RDs 4, 9, 12, 702, 711) assays. Patient characteristics, residency, and risks were obtained with a structured questionnaire. We used SaTScan and ArcMap analyses to identify significantly clustered MTBC lineages within selected districts and spatial display, respectively. Results Among 2,551 isolates, 2,019 (79.1%), 516 (20.2%) and 16 (0.6%) were identified as M. tuberculosis sensu stricto (MTBss), M. africanum (Maf), 15 M. bovis and 1 M. caprae, respectively. The proportions of MTBss and Maf were fairly constant within the study period. Maf spoligotypes were dominated by Spoligotype International Type (SIT) 331 (25.42%), SIT 326 (15.25%) and SIT 181 (14.12%). We found M. bovis to be significantly higher in Northern Ghana (1.9% of 212) than Southern Ghana (0.5% of 2339) (p = 0.020). Using the purely spatial and space-time analysis, seven significant MTBC lineage clusters (p< 0.05) were identified. Notable among the clusters were Ghana and Cameroon sub-lineages found to be associated with north and south, respectively. Conclusion This study demonstrated that overall, 79.1% of TB in Ghana is caused by MTBss and 20% by M. africanum. Unlike some West African Countries, we did not observe a decline of Maf prevalence in Ghana. http://dx.doi.org/10.13039/100004440Wellcome Trust097134/Z/11/Zhttp://orcid.org/0000-0002-1663-6287Yeboah-Manu Dorothy This work was supported by the Wellcome Trust fellowship grant 097134/Z/11/Z to DYM. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are contained within the paper and its Supporting Information files.Data Availability All relevant data are contained within the paper and its Supporting Information files. ==== Body Introduction One of the major threats to tuberculosis (TB) control is the emergence of strains that are resistant to most of the anti-TB drugs, which could make a treatable disease untreatable [1]. Other factors that limit current TB control efforts are lack of effective vaccines, lack of cheap but effective rapid diagnostics, emergence of HIV/AIDS pandemic and limited understanding of the diversity of circulating strains [1]. The increase in TB cases globally requires a concerted effort to control this global public health problem. This calls for improved understanding of the disease pathogenesis, epidemiology, and genetic variability within the causative agent. TB is caused by a group of closely related acid-fast gram-positive bacteria, together referred to as the Mycobacterium tuberculosis complex (MTBC) [2, 3]. The MTBC comprises M. tuberculosis sensu stricto (MTBss), M. Africanum (Maf), M. microti, M. bovis, M. caprae, M. mungi, M. suricattae, M. orygis and M. pinnipedii. They have varying host ranges: Mycobacterium microti affects voles, [4, 5] M. caprae a pathogen of goats and sheep [6]. M. mungi: Mangoose pathogen, M. orygis a pathogen of antelope [7], M. pinnipedii a pathogen of seals and sea lions [8]. Mycobacterium bovis displays the broadest spectrum of host affecting humans and animals [9]. Mycobacterium tuberculosis sensu stricto and M. africanum are the main causative agents of TB in humans; referred to as human adapted MTBC and the remaining seven species as animal adapted [3]. The human adapted MTBC comprises seven main phylogenetic lineages, which have been confirmed by single nucleotide polymorphisms (SNPs) and whole genome sequencing [3, 10–12]. These lineages were further found to exhibit a phylogeographical structure, which means that specific lineages are closely associated with specific geographic regions, and preferentially infect persons originating from these regions. Importantly, findings from recent genomic analysis indicate that some of these human MTBC lineages are as genetically distinct from each other as from the animal-adapted forms of MTBC [10] and have genomic differences that may influence host-pathogen interaction as well as applicability of control tools such as diagnostics and vaccine. Thus the lineages distribution needs to be taken into account in the development and testing of new control tools such as vaccines to account for any possible differential phenotypes. West Africa shows a unique mycobacterial population structure, as it is the only region worldwide where lineages of Maf are endemic [13]. Work done mainly in the Gambia, suggested that Maf is attenuated compared to MTB [13, 14]. While both transmit equally, the rate of progression to disease was slower in Maf infected contacts. Furthermore, MTBC lineage 6 (also known as Maf West Africa 2) was found to be associated with HIV co-infection and reduced ESAT6 secretion [15, 16]. Thus MTBss seems to have a competitive advantage that could lead to a replacement of Maf with the more virulent MTB. This might be particularly likely due to the large population increases in West African cities [17]. One could also argue that with the HIV pandemic and other immune suppression diseases, Maf will still be an important pathogen in West Africa. Recent publications from various countries, however, observed an interesting trend: the slow replacement of Maf with MTBss, especially the Cameroun sub-lineage of lineage 4. This phenomenon was first described in Guinea-Bissau, where lineage 6 decreased from 51% to 39% in about 2 decades [18]. Declines in prevalence of the other M. africanum lineage have also been observed in Côte d’Ivoire, and Cameroon [19–22]. At the same time, understanding of the genetic population structure of circulating MTBC strains is increasingly becoming important for TB control. Current genomic studies have revealed that substantial strain genetic diversity exists among the different members and genotypes of MTBC, which may have implications for the development and deployment of new TB vaccines and diagnostics [23]. In this study, we analysed the distribution of MTBC lineages and sub-lineages in Ghana, a country harbouring six of the seven identified MTBC [24, 25] over an 8-year period. Our findings indicated a fairly constant distribution of the two main MTBC species and lineages over time. In addition, we observed clustering of some MTBC lineages at specific geographical locations. Materials and Methods Ethics Statement The Scientific and Technical Committee and then the Institutional Review Board (IRB) of the Noguchi Memorial Institute for Medical Research with a federal wide assurance number FWA00001824 reviewed the protocols and procedures for this study and approved them. Written informed consent was obtained from participants using a designed form which was approved by the IRB. Methods for sputum sampling conformed to WHO guidelines (two sputa per patient) and patients’ identity was protected. Study Locale and Participants Data The study was conducted from July 2007 to December 2014 in Ghana, involving sputum smear positive TB cases. From July 2007 to December 2011, patients were recruited from five health facilities; Korle-Bu Teaching Hospital (KBTH) in the Greater Accra region, Agona Swedru Government Municipal Hospital (ASH), Winneba Government Hospital (WGH), St Gregory Catholic Clinic from the Central Region, all in the southern section of Ghana (Fig 1). Between 2012 and 2014, based on an on-going prospective study; sputum was collected from suspected TB cases reporting to the selected health facilities in the Accra Metropolitan Authority (AMA) and 2 districts (Mamprusi East (MamE) and Tamale Metropolis (TamM)) in the northern region of Ghana (Fig 1) after informed consent. The study sampling sites span 13 administrative districts with a combined population of 4,024,810 [26–29] in three regions according to the current administrative district division status created in 2013 (S1 Table). The AMA administrative district is made up of 10 sub-districts; Ablekuma South, Ablekuma North, Ablekuma Central, Ashiedu Keteke, Okai Koi South, Okai Koi North, Osu Klotey, Ayawaso East, Ayawaso Central and Ayawaso West Wogon with a combined population of 1,665,086 according to the 2010 population and housing census conducted in Ghana [26, 30]. The ten sub-districts within the AMA for the purposes of this study were merged into 5 sub-districts (Fig 1), based on the geographical demarcation existing as at 2007. These 5 sub-districts were: Ablekuma (Able), Ashiedu Keteke (AshK), Ayawaso (Ayaw), Okaikoi (Okai) and Osu Klottey (OsuK). The AMA covers a total land area of 136.674 square kilometres. Kpeshi, a former administrative sub-district of AMA is located on the eastern boundary of AMA and has been currently broken down into two districts; La Dade Kotopon Municipal and Ledzokuku/Krowor Municipal. We included the former Kpeshie demarcation in all targeted analysis involving AMA as TB patients still access facilities within AMA. Sites in the northern region, TamM and MamE, covers a land area of 790.5 and 1,823.6 square kilometres respectively with a combined population of 492,360 (TamM: 371,351 and MamE: 121,009). Together, TamM and MamE constitute 19.9% of the total population in the northern region of Ghana [29]. 10.1371/journal.pone.0161892.g001Fig 1 Map of sampling sites and study area. Sputum samples were obtained from fifteen sampling sites (health facilities) all located within three regions in Ghana; Greater Accra, Central and Northern regions. During the period September 2012 –December 2014, samples were obtained mainly from all the 13 diagnostic centres within the Accra metropolis, (serving more than 46% of the Greater Accra region populace) and the two health facilities located in the northern region (Tamale Teaching Hospital and BMC Nalerigu). The ArcMap program in ArcGIS v. 10.0 was used to create the map. Information on age, sex, nationality, ethnicity, employment status, previous history of TB, crowding, substance abuse and duration of symptoms were obtained from the patients with a structured questionnaire. Mycobacterial Isolation, Species and Lineage Classification Mycobacterial species were isolated by decontaminating sputum samples with equal amount of 5% oxalic acid solution, then inoculation on Lowenstein Jensen media and incubated as previously described [31]. Members of the MTBC were confirmed by PCR detection of the insertion sequence IS6110 and rpoβ as previously described [32]. Classification into the main phylogenetic lineages was achieved by large sequence polymorphism typing assay identifying regions of difference (RD) 4, 9, 12, 702, 711 [2, 3, 14] and also by spoligotyping following manufacturer’s directions (Isogen Bioscience BV Maarssen, The Netherlands). Data Management and Analysis Data obtained using the structured questionnaire was double entered using Microsoft Access and validated to correct entry errors. The questionnaire data primarily provided us with the year of diagnosis and residential address (location) of each TB case for the spatio-temporal analysis. In addition to these data, other demographic and clinical characteristics of each participant as indicated above were generated. The association of specific lineages and/or sub-lineages of the MTBC with time and/or geographical locations were explored with Fishers exact test using the Stata statistical package (Stata Corp., College Station, TX, USA). All analyses were run with significance level pegged at p < 0.05. To determine the TB case notification rates for the period 2012–2014, we obtained the projected population of the individual districts using the exponential growth rate formula; Pt = Poert (based on the assumption of constant population growth similar to compounded interest) [33]. Where; Pt = projected population, Po = initial population, e = base of the natural logarithm, r = intercensal growth rate and t = time elapsed after last census. The intercensal growth rates (S1 Table) used for the various regions were obtained from the Ghana statistical service 2010 population and housing census data [26]. The GIS co-ordinates of the participants’ self-reported district of residency was used to construct a pictorial plot of the distribution of the MTBC lineages analysed using the ArcMap (Economic and Social Research Institute, version 10.0) [34]. The district allocation data generated was linked to a molecular data of all TB isolates and was used for TB lineage clustering analysis. Spatial and Space-time Analysis Kulldorff’s scan statistics (SaTScan™ 9.4.2) tool [35], a commonly used tool for spatial and space-time cluster analysis for diseases in a wide variety of settings [36–40] was used for analysis of spatio-temporal clustering of TB cases using data obtained only within the time period; September 2012 to December 2014. TamM was excluded from analysis where 2012 data was used since we recorded no TB case in 2012. The Kulldorff’s scan statistics tool was used to detect significant MTBC clusters using the Monte Carlo simulations [41]. Three input files (cases, population and coordinates) were built using excel and saved in the required format for upload into the SaTScan software. The discrete Poisson model was used for the analysis with the assumption that the number of cases at each district had Poisson distribution with a known population at risk [41]. All other parameters were set at default for both spatial and space-time analysis (S4 Table). The results of the analyses were tabulated to add statistical significance to the inferences made using ArcMap. Normalization of TB cases for within district comparison To analyse the spatial and space-time distribution of MTBC cases at the district/sub-district level, we normalized the relative case frequencies against their respective reference population obtained from the Ghana Statistical Service [26, 27, 29, 30] Also, records of specific genotypes (or sub-lineages) were normalized using all recorded cases of the specified genotype (or lineage) within the specified time. For example, all Ghana and Cameroon sub-lineages per district were normalized using all Lineage 4 cases as the denominator. Results Characteristics of Patients Presenting with Tuberculosis Sputum smear positive patients from whom MTBC strains were isolated were 2551/3110 (82.0%) cases, comprising 70% (1789/2551) males and 30% (762/2551) females. Participants’ age ranged between 2 to 91 and a median age of 39 years. Ninety-one point seven percent (2339/2551) of the patients were from Southern Ghana and the remaining 8.3% (212/2551) from Northern Ghana (Fig 2). The HIV status of 1613 patients was indicated, of which 15.5% (250/1613) were HIV positive. The additional demographic and clinical characteristics of the cases are indicated in Table 1. 10.1371/journal.pone.0161892.g002Fig 2 Spatial distribution and prevalence of identified mycobacterium tuberculosis lineages. Diagram shows the spatial distribution of (A) 2551 Mycobacterium tuberculosis complex (MTBC) strains; (B) 2535 human adapted MTBC; (C) regional prevalence of 1883 Lineage 4 sub lineages; (D) regional prevalence of 516 Mycobacterium africanum (Maf) isolates from the geographical regions served by the health facility where sampling was carried out in Ghana. Animal strains were found to be associated with the North (p = 0.0389). Similarly, Lineage 2 was found to be associated with the North (p = 0.0006). The most dominant Lineage 4 sub-lineage in the North is Ghana (p = 0.0000) whereas in the South is Cameroon, even though the association is not statistically significant. The unknown sub-type of Lineage 4 is associated with South (p = 0.0001). 10.1371/journal.pone.0161892.t001Table 1 Demographic and clinical data of 2551 TB cases. Variable (Total number analysed) Number (Percentage) Sex (2551) Male 1789 (70.1) Female 762 (29.9) Age category (2551) 08–25 459 (18.0) 26–40 1082 (42.4) 41–77 980 (38.4) >77 30 (1.2) Residency (2551) North 212 (8.3) South 2339 (91.7) Occupation (2530) Skilled 490 (19.2) Unskilled 1890 (74.1) Unemployed 150 (5.9) Settlement (2550) Urban 2283 (89.5) Rural 267 (10.5) HIV status1 (1613) Yes 250 (15.5) No 1363 (84.5) Presence of BCG scar 2(1817) Yes 904 (49.8) No 913 (50.2) Income* (2551) None 771 (30.2) Low 1500 (58.8) High 280 (11.0) Drinking status3 (2071) Yes 556 (26.8) No 1515 (73.2) TB in the past4 (2029) Yes 218 (10.7) No 1811 (89.3) Education Level (2551) None 300 (11.7) Primary 605 (23.7) Secondary 1551 (60.7) Tertiary 95 (3.7) Smear Grade (2551) Scanty 235 (9.2) 1 985 (38.6) 2 531 (20.8) 3 800 (31.4) In Household number (2551) <5 589 (23.1) >5 1962 (76.9) Ethnicity (2551) Akan 800 (31.4) Ewe 339 (13.3) Ga 595 (23.3) Mole -Dagbon 36 (1.4) Gruma 5 (0.2) Guan 9 (0.4) Others 767 (30.0) Marital status (2551) Single 859 (33.7) Married 1037 (40.7) Divorced 221 (8.7) Widowed 104 (4.1) Co habiting 330 (12.9) Cough (2551) < 2 weeks 2117 (82.9) > 2 weeks 122 (4.8) Symptoms other than cough 312 (12.2) Night Sweat 5 (2219) Yes 1372 (61.8) No 847 (33.2) Hemoptysis 6 (2227) Yes 480 (21.6) No 1747 (78.4) House type (2551) Self-contained 486 (19.1) Compound House 1707 (66.9) Others 358 (14.0) Swollen glands 7 (2211) Yes 210 (9.5) No 2001 (91.5) Chest Pain 8 (2234) Yes 1717 (76.9) No 517 (23.1) Nationality 9 (2251) Ghana 2187 (97.2) Nigeria 21 (0.9) Togo 12 (0.5) Niger 11 (0.4) Ivory Coast 7 (0.3) Others West African Nationals 13 (0.7) Contact with TB patient (2551) Yes 316 (12.4) No 2234 (87.6) Smoking 10(2419) Yes 500 (20.7) No 1919 (75.3) 1 = 938 missed data for HIV status, 2 = 734 missed data for Presence of BCG scar, 3 = 480 missed data for Drinking status, 4 = 522 missed data for TB in the past, 5 = 332 missed data for Night Sweat, 6 = 324 missed data for Night Sweat, 7 = 340 missed data for Swollen glands, 8 = 317 missed data for Chest Pain, 9 = 300 missed data for Nationality, 10 = 132 missed data for Smoking. * Income below 1000GH₵ was defined as low whilst those above 1000GH₵ as high. The Population Structure of MTBC causing pulmonary TB in Ghana Two thousand six hundred and three mycobacterial isolates were obtained from 3110 samples giving a cumulative isolation rate of 83.7%. We identified 2551 of the isolates as members of the MTBC and 52 as non-tuberculous mycobacteria (NTM) (as well as those with negative mycobacteria) which were excluded from further analysis. Among those confirmed as MTBC, 2019 (79.1%) were MTBss, 516 (20.2%) were Maf, and 16 (0.6%) animal strains (15 M. bovis (SIT 1037, 482) and 1 M. caprae) (Fig 2A). Six of the seven lineages of the human adapted MTBC (Maf and MTBss) were identified in the following proportions: L1 (36; 1.4%), L2 (77; 3.0%), L3 (23; 0.9%), L4 (1883; 74.3%), L5 (338; 13.3%) and L6 (178; 7.0%), respectively (Fig 2B). The sub-lineages identified within the L4 were the Cameroon (1151:61.1%) followed by the Ghana (330; 17.5%), then Haarlem (119; 6.3%), LAM (39; 2.1%), Uganda I (38; 2.0%), Uganda II (5; 0.3%), New-1 (1; 0.1%), S (3; 0.2%) and H37Rv-like (2; 0.1%) (Fig 2C). Spatial Distribution of MTBC Genotypes The combined number of isolates analysed from the different geographical areas, identified species, lineages and sub-lineages are indicated in Fig 2A–2D, respectively. As shown in Fig 2D, there was no statistical difference in the Maf proportion between the north (17.0%; 36/212) and the south (21.9%; 378/1726) (p = 0.1099). However, we found the proportion of animal-adapted species (MTBC other than MTBss and Maf) in the north (1.9%; 4/212) to be more than twice the proportion in the south (0.7%; 12/1726) (p<0.0884; OR = 2.74). There was unequal spatial distribution of L4 sub-lineages and spoligotypes. The proportion of the Ghana sub-lineage was statistically higher in Northern Ghana (32.3%) compared to 20.1% in the south (p = 0.0016, OR = 1.9, 95%CI = 1.3–2.9). Whereas the Spoligotype international type (SIT) 61 was more likely to be found in the south (p = 0.0330; OR = 0.7; 95% CI = 0.4–0.9), the SIT 53 was more likely to be found in the north (p = 0.0015; OR = 2.0; 95% CI = 1.3–3.1). In addition, L2 was proportionally higher in the north (5.7%; 12/212) compared to the south (3.1%; 53/1726). Even though the sample size changed over time due to increase in case study sites, the proportion of distinct species did not change over time (Fig 3). The species/lineages/sub-lineages distributions of MTBC within the 13 administrative districts where participants resided are displayed in Fig 4. 10.1371/journal.pone.0161892.g003Fig 3 Temporal distribution and prevalence of human adapted mycobacterium tuberculosis complex (MTBC). Figure displays a stacked graph showing the temporal distribution of human adapted MTBC (left y-axis) and a linear graph showing the prevalence of Mycobacterium tuberculosis sensu stricto (MTBss) and Mycobacterium africanum (Maf) (right y-axis) over the entire 8-year study period. 10.1371/journal.pone.0161892.g004Fig 4 Spatial distribution of human adapted MTBC lineages and major sub-lineages within the eight-year study period. The figure shows the distribution of MTBC species/lineages/sub-lineages within the 13 districts where participants resided. The blue coloured panel shows the distribution of all the tuberculosis cases recruited with well-defined residential status. The red, brown and green coloured panels show the distribution of lineage 4, lineage 5 and lineage 6 respectively. All other sub-lineages/species have been indicated in the respective legends. This figure was created using the ArcMap program in ArcGIS v. 10.0. Abbreviations: MTBC, Mycobacterium tuberculosis complex; MTBss, Mycobacterium tuberculosis sensu stricto; Maf, Mycobacterium africanum; L4, Lineage 4; L5, Lineage 5; L6, Lineage 6; Ghana, Ghana genotypes (Ghana sub-lineage); Cam, Cameroon sub-lineage; MamE, Mamprusi East district; TamM, Tamale Metropolis; AgWM, Agona West Municipal; GomE, Gomoa East; AwuS, Awutu Senya; GaSM, Ga South Municipal; GaWM, Ga West Municipal; GaCM, Ga Central Municipal; GaEM, Ga East Municipal; AMA, Accra Metropolis; LaNM, La-Nkwantanang Madina Municipal; AdeM, Adenta Municipal; Kpes, Kpeshie Municipal. Spatial and space-time clustering analysis of MTBC cases at the district/sub-district level (2012–2014) Spatial and space-time analyses were carried out for districts where sampling was performed within the time period; September 2012 to December 2014. These districts were AMA (sub-divided into sub-districts due to the population density) in the south and in the north, MamE. The TB case notification rate ranged from 3 to 52 cases/100,000 individuals at risk within the districts/sub-district analyzed (Fig 5A) with the highest case notification rate occurring in 2013 (52 cases/100,000). In a purely spatial analysis, we found two significant clusters within the study period based on cases notified. The most likely cluster consisted of two sub-districts, AshK and OsuK (p = 0.0000, RR = 3.99) with a secondary cluster occurring at MamE, (p = 0.0000, RR = 2.16) (Table 2). Similar observations were made using a space-time analysis with the likely clusters occurring in 2013. To analyze the spatial and space-time distribution of the two human adapted MTBC (MTBss and Maf), we normalized the relative district case frequencies to that for all cases obtained per district/year (Fig 5B and 5C). We found that the normalized distribution of both MTBss and Maf fluctuated over the three-year period, and no particular district/sub-district showed constant high values (Fig 5B and 5C). In a purely spatial analysis, significant clusters (p = 0.0000, Table 2) were observed in 6 of the 8 districts (MamE, TamM, Ayaw, Okai, Kpes and OsuK). 10.1371/journal.pone.0161892.g005Fig 5 Spatial distribution of isolated MTBC within selected districts (2012–2014). This figure shows the; (A) Sum and case notification rate of all TB cases from September 2012 to December 2014, (B) Sum and normalized distribution of MTBss cases, (C) Sum and normalized distribution of Maf. The total number of cases per year was used as the denominator for normalization. Sampling from TamM did not meet our criteria for being included in analyses for case notification rate and so was excluded in all columns of panel A. Likewise we also recorded no cases in 2012 as such TamM was excluded from 2012 analysis (panel B and C). This figure was created using the ArcMap program in ArcGIS v. 10.0. Abbreviations: MTBss, Mycobacterium tuberculosis sensu stricto; Maf, Mycobacterium africanum; MamE, Mamprusi East district; TamM, Tamale Metropolis; AshK, Ashiedu Keteke; Ayaw, Ayawaso; Able, Ablekuma; OsuK, Osu Klottey; Okai, Okaikoi; Kpes, Kpeshie. 10.1371/journal.pone.0161892.t002Table 2 Most likely spatial clusters detected in the study area using SaTScan analysis. TB cases Reference population Cluster type Year (s) of observed cluster Clustered districts Observed cases Expected cases Log Likelihood ratio Relative risk P-value Type of analysis All TB cases District population Most likely 2012–2014 AshK, OsuK 213 61.2 123.9 3.99 0.000 Purely spatial All TB cases District population Secondary 2012–2014 TamM, MamE 138 67.9 29.8 2.16 0.000 Purely spatial All TB cases District population Most likely 2013 AshK, OsuK 123 29.7 85.1 4.48 0.000 Space-time All TB cases District population Secondary 2013 TamM, MamE 50 20.5 15.4 2.5 0.000 Space-time MTBss All cases per district Most likely 2012–2014 MamE, TamM, Ayaw, Okai, Kpes, OsuK 588 398.1 75.0 2.21 0.000 Purely spatial Maf All cases per district Most likely 2012–2014 MamE, TamM, Ayaw, Okai, Kpes, OsuK 158 107.8 18.9 2.12 0.000 Purely spatial L4 All L4 cases in 2012 Most likely 2012–2014 Able 285 279.1 0.1 1.03 1.000 Purely spatial L4 All L4 cases in 2012 Most likely 2012 Able 53 37.1 3.2 1.46 0.204 Space-time Gh All L4 cases in 2014 Most likely 2014 MamE, TamM 23 10.8 5.6 2.27 0.013 Space-time Cam All L4 cases in 2014 Most likely 2014 Kpes, Okai, Ayaw, TamM, MamE 122 80.0 11.5 1.68 0.000 Space-time Cam All L4 cases in 2013 Secondary 2013 AshK, OsuK 52 30.4 6.8 1.79 0.003 Space-time This table shows the most likely spatial clusters detected from the SaTScan analysis. The TB case in the first column shows the category of TB lineage/sub-lineage to which the spatial or space-time analysis was performed. The districts to which the clusters were observed are shown in column 5 with the respective year of cluster observation shown in column 4. The last column shows the type of cluster analysis performed. Abbreviations: TB, tuberculosis; MTBss, Mycobacterium tuberculosis sensu stricto; Maf, Mycobacterium africanum; L4, Lineage 4; Gh, Ghana genotypes (Ghana sub-lineage); Cam, Cameroon sub-lineage; MamE, Mamprusi East; TamM, Tamale Metropolis; AshK, Ashiedu Keteke; Ayaw, Ayawaso; Able, Ablekuma; OsuK, Osu Klottey; Okai, Okaikoi; Kpes, Kpeshie. The Ghana sub-lineage was found to significantly cluster in the north (p = 0.013, RR = 2.27). A space-time analysis revealed two significant clusters for the Cameroon sub-lineage with the most likely cluster occurring in 2014 (p = 0.000, RR = 1.68) consisting of five districts (Kpes, Okai, Ayaw, TamM and MamE). The second Cameroon sub-lineage cluster involved AshK and OsuK, which occurred in 2013 (p = 0.003, RR = 1.79, Fig 6C). Comparing the North and South for association with some risk factors showed association of the North with rural settings (P = 0.0000), farming (0.0000), contact with cattle (0.0003), compound housing (0.0000) whereas the South was associated with driving as occupation (0.0163) as shown in supplementary data (S6 Table). 10.1371/journal.pone.0161892.g006Fig 6 Spatial distribution of lineage 4 and major lineage 4 sub-lineages within selected districts (2012–2014). The figure shows the (A) Sum and normalized distribution of lineage 4 cases, (B) Sum and normalized distribution of Ghana sub-lineages cases, (C) Sum and normalized distribution of Cameroon sub-lineage cases. The total number of cases per year was used as the denominator for normalization. This study recorded no TB cases for TamM in 2012, consequently TamM was excluded from all analysis carried out using 2012 data. This figure was created using the ArcMap program in ArcGIS v. 10.0. Abbreviations: L4, Lineage 4; Ghana, Ghana genotypes (Ghana sub-lineage); Cam, Cameroon sub-lineage; MamE, Mamprusi East district; TamM, Tamale Metropolis; AshK, Ashiedu Keteke; Ayaw, Ayawaso; Able, Ablekuma; OsuK, Osu Klottey; Okai, Okaikoi; Kpes, Kpeshie. Discussion Our objective was to analyse in time and space the prevalence of MTBC species and genotypes among isolates obtained from sputum-positive TB cases over an 8-year period in Ghana. A secondary objective was to determine the space clustering of specific genotypes. Our longitudinal analysis indicated that: 1) the Beijing and Ghana genotypes of Lineages 2 and 4, respectively, as well as the animal adapted MTBCs are isolated more often from patients from Northern than Southern Ghana and 2) the proportion of Maf among the isolates over the study period remained fairly constant. Our first evidence of an association between the north and the Ghana genotype of Lineage 4 as well as the Beijing genotype of Lineage 2 reiterates the phylogeographical nature of the human-adapted MTBCs such that even within a single country there can be variations in the distribution of distinct genotypes within a lineage and specific geographical regions. For example in Senegal, it has been observed that the proportion of M. africanum causing TB varies by region [42]. The Ghana genotype clustered in the two districts (MamE and TamM) of the north (Fig 6B) whereas the Cameroon sub-lineage clustered in the South as two clusters (Fig 6C; Table 2). Thus clustering alone cannot be used as a proxy for active transmission of these genotypes in the specific geographical areas as the resolution of the molecular tool (spoligotyping) used for characterisation within this study is not enough to infer on-going transmission because of its low discriminatory power. However, they may be indicative of the areas of origin or introduction of these genotype/spoligotypes into the country. The Cameroon genotype as observed from various molecular studies is the most prevalent genotype causing TB in West Africa and our findings confirm this and also show that this genotype may have been introduced into the country through the south. Our findings call for studies to investigate the transmission dynamics of these sub-lineages within the respective geographical areas; this is of public health concern because evidence that MTBC genotypes might influence disease phenotype. For example, an association between Beijing strains (Lineage 2) and drug resistance has often been reported, and we recently showed that the Ghana genotype is also associated with drug resistance in Ghana (manuscript submitted). This means that effective control of TB in the Northern region of Ghana would be challenged with these two genotypes in higher proportions. We also found the animal-adapted MTBC to be statistically associated with Northern Ghana (p = 0.0381, OR = 3.73). There were 5 patients among our study population who had direct contact with cattle including 4 butchers and 1 farmer who owns cattle and all 5 were infected with animal strains of the MTBC. This finding supports previous observations that people who are in direct constant contact with cattle and/or their products may be at risk of infection with M. bovis [43]. However, there were some patients from whom animal strains were isolated but did not have constant direct contact with cattle or any other farm animal. This finding compares with a similar work done in Mexico where most of the patients from whom animal strains of the MTBC were isolated had no direct contact with livestock but rather had consumed unpasteurized milk products in the past [44]. Our work design did not include analysis of patients’ dietary lifestyle and therefore we have no data on whether those patients are in a similar situation. Nevertheless, this association could be due to the dominance of livestock in the Northern region as compared to the south [26] meaning more possible interaction of humans with animals in the north as compared to the south. The proportion of MTBss analysed among our data set remained stable and higher than that of Maf over the eight-year period (Figs 2A and 3). This result is in contrast with reports from other West African countries, indicating a decline of Maf [20, 21]. Various reasons have been suggested to explain the observed decline including non-specificity of biochemical assays used previously. Nevertheless, this study shows that Maf has remained fairly constant over the study period at an average of approximately 20% with significant fluctuations observed within four different periods {(2009/2010, P = 0.0422), (2010/2011, P = 0.0033), (2011/2012, P = 0.0002), (2013/2014, P = 0.0014)}. This finding in conjunction with others [45–47] indicates that the two TB causing pathogens have adapted well within the Ghanaian population. Moreover, two independent studies conducted by our group found a strong association between L5 and an ethnic group in Ghana [[25], Asante-Poku et al, submitted]. Study limitations We did not collect GPS coordinate of the residence of the study participants, so each participant residential district was generated using his or her residential addresses from the questionnaire. This might not be accurate and so future studies should be done using GPS coordinates taken from each individual participant’s actual residence. Supporting Information S1 Table Districts and population statistics within designated time points within the study period. The last population census conducted in Ghana was in 2010. As a results to obtain the population statistics for the period 2012 to 2014 (columns 5 to 7) we used the exponential growth rate formulae as described in methods. The intercensal growth rates used per region were: Greater Accra (3.1%), Central region (3.1%), Northern region (2.9%). *sub-districts within AMA; †Projected population from 2010 population census data. (PDF) Click here for additional data file. S2 Table Annual distribution of TB cases used for spatial or space-time analysis (2007–2014). The table lists the number of tuberculosis cases sampled within each district/sub-district for only participants with residential status. Periods within the 8 year study period where no sampling was done are marked N/A (not available). The final column and row contains total counts for each district/sub-district and year respectively. *sub-districts within AMA. (PDF) Click here for additional data file. S3 Table Lineage distribution of isolated MTBC recruited and used for spatial or space-time analysis (2007–2014). The table lists the distribution of tuberculosis species/major lineages/sub-lineages sampled within each district/sub-district for only participants with well-defined residential status. The final column and row contains total counts for each district/sub-district and TB species/lineage/sub-lineage respectively. Abbreviations: TB, tuberculosis; MTBss, Mycobacterium tuberculosis sensu stricto; Maf, Mycobacterium africanum; L4, Lineage 4; L5, Lineage 5; L6, Lineage 6; Gh, Ghana genotypes (Ghana sub-lineage); Cam, Cameroon sub-lineage. *sub-districts within Accra metropolis. (PDF) Click here for additional data file. S4 Table Default parameters used in SaTScan for clustering analysis. The table contains a list of the default settings used in performing the clustering analysis using the SaTScan software. (PDF) Click here for additional data file. S5 Table Primer sequences for large sequence polymorphism (LSP) assay. The table contains a list of primer sequences used for the LSP assay. (PDF) Click here for additional data file. S6 Table Comparison of some risk factors among the two regions. (PDF) Click here for additional data file. S7 Table Genotyping profile of 2551 MTBC isolates from Ghana. (PDF) Click here for additional data file. S8 Table Distribution of Species and Lineages of MTBC. (PDF) Click here for additional data file. We express our gratitude to all laboratory staff and study participants of the various health facilities for their time and cooperation during the study period. ==== Refs References 1 World Health Organization . Global tuberculosis report 2014 . Geneva : World Health Organization ; 2014 . 2 Brosch R , Gordon SV , Marmiesse M , Brodin P , Buchrieser C , Eiglmeier K , et al A new evolutionary scenario for the Mycobacterium tuberculosis complex . Proc Natl Acad Sci U S A . 2002 ; 99 (6 ): 3684 –9 . 10.1073/pnas.052548299 11891304 3 Gagneux S , Small PM . Global phylogeography of Mycobacterium tuberculosis and implications for tuberculosis product development . The Lancet Infectious Diseases . 2007 ; 7 (5 ): 328 –37 . 10.1016/s1473-3099(07)70108-1 17448936 4 Wells AQ , Oxon DM . Tuberculosis in wild voles . Lancet i . 1937 :1221 . 5 Frota CC , Hunt DM , Buxton RS , Rickman L , Hinds J , Kremer K , et al Genome structure in the vole bacillus, Mycobacterium microti, a member of the Mycobacterium tuberculosis complex with a low virulence for humans . Microbiology . 2004 ;150 (Pt 5 ):1519 –27 . Epub 2004/05/11. 10.1099/mic.0.26660-0 ; PubMed Central PMCID: PMCPmc2964484.15133113 6 Aranaz A , Liebana E , Gomez-Mampaso E , Galan JC , Cousins D , Ortega A , et al Mycobacterium tuberculosis subsp. caprae subsp. nov.: a taxonomic study of a new member of the Mycobacterium tuberculosis complex isolated from goats in Spain . Int J Syst Bacteriol . 1999 ;49 Pt 3 :1263 –73 . Epub 1999/07/30. 10.1099/00207713-49-3-1263 .10425790 7 Jakko van I , Zeaur R , Arnout M , Martin JB , Roxane S , Roland B , et al Characterization of Mycobacterium orygis as M. tuberculosis Complex Subspecies . Emerging Infectious Disease journal . 2012 ;18 (4 ):653 10.3201/eid1804.110888 8 Cousins DV , Bastida R , Cataldi A , Quse V , Redrobe S , Dow S , et al Tuberculosis in seals caused by a novel member of the Mycobacterium tuberculosis complex: Mycobacterium pinnipedii sp. nov . Int J Syst Evol Microbiol . 2003 ;53 (Pt 5 ):1305 –14 . Epub 2003/09/18. 10.1099/ijs.0.02401-0 .13130011 9 Garnier T , Eiglmeier K , Camus JC , Medina N , Mansoor H , Pryor M , et al The complete genome sequence of Mycobacterium bovis . Proc Natl Acad Sci U S A . 2003 ;100 (13 ):7877 –82 . Epub 2003/06/06. 10.1073/pnas.1130426100 ; PubMed Central PMCID: PMCPmc164681.12788972 10 Gagneux S , DeRiemer K , Van T , Kato-Maeda M , de Jong BC , Narayanan S , et al Variable host-pathogen compatibility in Mycobacterium tuberculosis . Proc Natl Acad Sci U S A . 2006 ; 103 (8 ): 2869 –73 . 10.1073/pnas.0511240103 16477032 11 Comas I , Homolka S , Niemann S , Gagneux S . Genotyping of genetically monomorphic bacteria: DNA sequencing in Mycobacterium tuberculosis highlights the limitations of current methodologies . PLOS ONE . 2009 ;4 (11 ): e7815 10.1371/journal.pone.0007815 19915672 12 Rebuma F , Stefan B , Elena H , Esther S , Balako G , Girume E , et al Mycobacterial Lineages Causing Pulmonary and Extrapulmonary Tuberculosis, Ethiopia . Emerging Infectious Disease journal . 2013 ; 19 (3 ): 460 10.3201/eid1903.120256 13 de Jong BC , Adetifa I , Walther B , Hill PC , Antonio M , Ota M , et al Differences between TB cases infected with M. africanum, West-African type 2, relative to Euro-American M. tuberculosis- an update . FEMS Immunol Med Microbiol . 2010 ; 58 (1 ):102 –5 . 10.1111/j.1574-695X.2009.00628.x . PMC2922882. 20002176 14 de Jong BC , Antonio M , Gagneux S . Mycobacterium africanum—review of an important cause of human tuberculosis in West Africa . PLoS Negl Trop Dis . 2010 ; 4 (9 ): e744 10.1371/journal.pntd.0000744 20927191 15 de Jong BC , Hill PC , Aiken A , Awine T , Antonio M , Adetifa IM , et al Progression to active tuberculosis, but not transmission, varies by Mycobacterium tuberculosis lineage in The Gambia . J Infect Dis . 2008 ; 198 (7 ): 1037 –43 . Epub 2008/08/16. 10.1086/591504 ; PubMed Central PMCID: PMCPmc2597014.18702608 16 de Jong BC , Hill PC , Brookes RH , Gagneux S , Jeffries DJ , Otu JK , et al Mycobacterium africanum elicits an attenuated T cell response to early secreted antigenic target, 6 kDa, in patients with tuberculosis and their household contacts . J Infect Dis . 2006 ; 193 (9 ):1279 –86 . Epub 2006/04/06. 10.1086/502977 .16586366 17 UNICEF. Generation 2030 Africa: child demographics in Africa. Division of Data, Research Policy. 2014. Available: http://www.unicef.org/media/files/UNICEF_Africa_Generation_2030_en_11Aug.pdf 18 Kallenius G , Koivula T , Ghebremichael S , Hoffner SE , Norberg R , Svensson E , et al Evolution and clonal traits of Mycobacterium tuberculosis complex in Guinea-Bissau . J Clin Microbiol . 1999 ; 37 (12 ):3872 –8 . Epub 1999/11/24. ; PubMed Central PMCID: PMCPmc85833.10565899 19 Dosso M , Bonard D , Msellati P , Bamba A , Doulhourou C , Vincent V , et al Primary resistance to antituberculosis drugs: a national survey conducted in Cote d'Ivoire in 1995–1996. Ivoirian Study Group on Tuberculosis Resistance . Int J Tuberc Lung Dis . 1999 ; 3 (9 ): 805 –9 . Epub 1999/09/17. 10488889 20 Koro Koro F , Kamdem Simo Y , Piam FF , Noeske J , Gutierrez C , Kuaban C , et al Population dynamics of tuberculous Bacilli in Cameroon as assessed by spoligotyping . J Clin Microbiol . 2013 ; 51 (1 ): 299 –302 . Epub 2012/11/02. 10.1128/jcm.01196-12 ; PubMed Central PMCID: PMCPmc3536219.23115266 21 Huet M , Rist N , Boube G , Potier D . Bacteriological study of tuberculosis in Cameroon . Rev Tuberc Pneumol (Paris) . 1971 ; 35 (4 ): 413 –26 . Epub 1971/06/01. .5002285 22 Niobe-Eyangoh SN , Kuaban C , Sorlin P , Cunin P , Thonnon J , Sola C , et al Genetic biodiversity of Mycobacterium tuberculosis complex strains from patients with pulmonary tuberculosis in Cameroon . J Clin Microbiol . 2003 ;41 (6 ): 2547 –53 . Epub 2003/06/07. ; PubMed Central PMCID: PMCPmc156567.12791879 23 Hershberg R , Lipatov M , Small PM , Sheffer H , Niemann S , Homolka S , et al High functional diversity in Mycobacterium tuberculosis driven by genetic drift and human demography . PLOS Biol . 2008 ; 6 (12 ): e311 10.1371/journal.pbio.0060311 19090620 24 Yeboah-Manu D , Asante-Poku A , Bodmer T , Stucki D , Koram K , Bonsu F , et al Genotypic diversity and drug susceptibility patterns among M. tuberculosis complex isolates from South-Western Ghana . PLOS ONE . 2011 ;6 (7 ): e21906 10.1371/journal.pone.0021906 21779354 25 Asante-Poku A , Yeboah-Manu D , Otchere ID , Aboagye SY , Stucki D , Hattendorf J , et al Mycobacterium africanum is associated with patient ethnicity in Ghana . PLoS Negl Trop Dis . 2015 ; 9 (1 ): e3370 10.1371/journal.pntd.0003370 25569290 26 Ghana Statistical Service National Analytical Report. 2010 Population and Housing Census. 2012. Available: http://www.statsghana.gov.gh/publications.html. Accessed 31 January 2016. 27 Ghana Statistical Service Greater Accra Regional Analytical Report, 2010 Population and Housing Census. 2013. Available: http://www.statsghana.gov.gh/publications.html. Accessed 31 January 2016. 28 Ghana Statistical Service Central Region Analytical Report, 2010 Population and Housing Census. 2013. Available: http://www.statsghana.gov.gh/publications.html. Accessed 31 January 2016. 29 Ghana Statistical Service Northern Region Analytical Report, 2010 Population and Housing Census. 2013. Available: http://www.statsghana.gov.gh/publications.html. Accessed 31 January 2016. 30 Ghana Statistical Service District Analytical Report. 2010 Population and Housing Census. Accra Metropolitan. 2014. Available: http://www.statsghana.gov.gh/publications.html. Accessed 31 January 2016. 31 Yeboah-Manu D , Bodmer T , Mensah-Quainoo E , Owusu S , Ofori-Adjei D , Pluschke G . Evaluation of decontamination methods and growth media for primary isolation of Mycobacterium ulcerans from surgical specimens . J Clin Microbiol . 2004 ; 42 (12 ): 5875 –6 . 10.1128/JCM.42.12.5875-5876.2004 15583329 32 Yeboah-Manu D , Yates MD , Wilson SM . Application of a simple multiplex PCR to aid in routine work of the mycobacterium reference laboratory . J Clin Microbiol . 2001 ; 39 (11 ): 4166 –8 . 10.1128/JCM.39.11.4166-4168.2001 11682550 33 Shryock HS , Siegel JS , Larmon EA . The Methods and Materials of Demography : Department of Commerce, Bureau of the Census ; 1980 562 p. 34 ESRI . Environmental Systems Resource Institute. ArcMap 10.0 . Redlands, California 2010 . 35 Kulldorff M. SaTScanTM User Guide for version 9.4 2015. 36 Kammerer JS , Shang N , Althomsons SP , Haddad MB , Grant J , Navin TR . Using statistical methods and genotyping to detect tuberculosis outbreaks . International journal of health geographics . 2013 ; 12 :15 Epub 2013/03/19. 10.1186/1476-072x-12-15 ; PubMed Central PMCID: PMCPmc3608068.23497235 37 Kulldorff M , Nagarwalla N . Spatial disease clusters: Detection and inference . Stat Med . 1995 ; 14 (8 ): 799 –810 . 7644860 38 Dangisso MH , Datiko DG , Lindtjørn B . Spatio-Temporal Analysis of Smear-Positive Tuberculosis in the Sidama Zone, Southern Ethiopia . PLOS ONE . 2015 ; 10 (6 ): e0126369 10.1371/journal.pone.0126369 26030162 39 Nana Yakam A , Noeske J , Dambach P , Bowong S , Fono LA , Ngatchou-Wandji J . Spatial analysis of tuberculosis in Douala, Cameroon: clustering and links with socio-economic status . Int J Tuberc Lung Dis . 2014 ; 18 (3 ): 292 –7 . Epub 2014/03/29. 10.5588/ijtld.13.0573 .24670564 40 Zhao F , Cheng S , He G , Huang F , Zhang H , Xu B , et al Space-Time Clustering Characteristics of Tuberculosis in China, 2005–2011 . PLOS ONE . 2013 ; 8 (12 ): e83605 10.1371/journal.pone.0083605 . PMC3868653. 24367604 41 Kulldorff M . A spatial scan statistic . Communications in Statistics Theory and Methods . 1997 ; 26 (6 ):1481 –96 . 42 Diop S , de Medeiros D , de Medeiros G , Baylet R , Sankale M . [Incidence and geographic distribution of Mycobacterium africanum in Senegal ]. Bull Soc Med Afr Noire Lang Fr . 1976 ;21 (1 ):50 –6 . Epub 1976/01/01. .1000739 43 Michel AL , Muller B , van Helden PD . Mycobacterium bovis at the animal-human interface: a problem, or not? Vet Microbiol . 2010 ;140 (3–4 ):371 –81 . Epub 2009/09/24. 10.1016/j.vetmic.2009.08.029 .19773134 44 Laniado-Laborin R , Muniz-Salazar R , Garcia-Ortiz RA , Vargas-Ojeda AC , Villa-Rosas C , Oceguera-Palao L . Molecular characterization of Mycobacterium bovis isolates from patients with tuberculosis in Baja California, Mexico . Infect Genet Evol . 2014 ; 27 :1 –5 . 10.1016/j.meegid.2014.06.020 .24997332 45 Meyer CG , Scarisbrick G , Niemann S , Browne ENL , Chinbuah MA , Gyapong J , et al Pulmonary tuberculosis: Virulence of Mycobacterium africanum and relevance in HIV co-infection . Tuberculosis . 2008 ; 88 (5 ):482 –9 . 10.1016/j.tube.2008.05.004 18590979 46 Addo K , Owusu-Darko K , Yeboah-Manu D , Caulley P , Minamikawa M , Bonsu F , et al Mycobacterial species causing pulmonary tuberculosis at the Korle Bu teaching hospital, Accra, Ghana . Ghana Med J . 2007 ; 41 (2 ): 52 –7 . 17925842 47 Goyal M , Lawn S , Afful B , Acheampong JW , Griffin G , Shaw R . Spoligotyping in molecular epidemiology of tuberculosis in Ghana . J Infect . 1999 ; 38 (3 ): 171 –5 . Epub 1999/07/29. .10424797

PMC005xxxxxx/PMC5001707.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756455110.1371/journal.pone.0160369PONE-D-16-16339Research ArticleBiology and Life SciencesBiochemistryProteinsCollagensBiology and Life SciencesCell BiologyCellular TypesAnimal CellsConnective Tissue CellsFibroblastsBiology and Life SciencesAnatomyBiological TissueConnective TissueConnective Tissue CellsFibroblastsMedicine and Health SciencesAnatomyBiological TissueConnective TissueConnective Tissue CellsFibroblastsResearch and Analysis MethodsSpectrum Analysis TechniquesSpectrophotometryFluorophotometryFluorescence Resonance Energy TransferBiology and Life SciencesDevelopmental BiologyMorphogenesisBiology and Life SciencesCell BiologyCell MotilityCell MigrationBiology and Life SciencesDevelopmental BiologyCell MigrationBiology and Life SciencesCell BiologyCell AdhesionIntegrinsBiology and Life SciencesCell BiologyCellular Structures and OrganellesExtracellular MatrixIntegrinsBiology and Life SciencesBiochemistryTissue DistributionMedicine and Health SciencesPharmacologyPharmacokineticsTissue DistributionPhysical SciencesMaterials ScienceMaterials by AttributeCoatingsEngineering and TechnologyManufacturing ProcessesSurface TreatmentsCoatingsFull-Length Fibronectin Drives Fibroblast Accumulation at the Surface of Collagen Microtissues during Cell-Induced Tissue Morphogenesis Fibronectin Drives Tissue MorphogenesisFoolen Jasper 1*Shiu Jau-Ye 1Mitsi Maria 1¤Zhang Yang 1Chen Christopher S. 23Vogel Viola 11 Laboratory of Applied Mechanobiology, Department of Health Sciences and Technology, ETH Zurich, Vladimir-Prelog-Weg 4, Zurich, Switzerland2 Department of Biomedical Engineering, Boston University, Boston, Massachusetts, United States of America3 Wyss Institute for Biologically Inspired Engineering, Harvard University, Boston, Massachusetts, United States of AmericaKumar Sanjay EditorUniversity of California Berkeley, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: JF. Data curation: JF. Formal analysis: JF JYS MM. Methodology: JF JYS MM. Resources: CSC. Supervision: VV. Visualization: JF. Writing – original draft: JF. Writing – review & editing: JF JYS MM YZ CSC VV. ¤ Current address: Paul Scherrer Institute, Villigen, Switzerland * E-mail: jasper.foolen@hest.ethz.ch26 8 2016 2016 11 8 e016036922 4 2016 18 7 2016 © 2016 Foolen et al2016Foolen et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Generating and maintaining gradients of cell density and extracellular matrix (ECM) components is a prerequisite for the development of functionality of healthy tissue. Therefore, gaining insights into the drivers of spatial organization of cells and the role of ECM during tissue morphogenesis is vital. In a 3D model system of tissue morphogenesis, a fibronectin-FRET sensor recently revealed the existence of two separate fibronectin populations with different conformations in microtissues, i.e. ‘compact and adsorbed to collagen’ versus ‘extended and fibrillar’ fibronectin that does not colocalize with the collagen scaffold. Here we asked how the presence of fibronectin might drive this cell-induced tissue morphogenesis, more specifically the formation of gradients in cell density and ECM composition. Microtissues were engineered in a high-throughput model system containing rectangular microarrays of 12 posts, which constrained fibroblast-populated collagen gels, remodeled by the contractile cells into trampoline-shaped microtissues. Fibronectin’s contribution during the tissue maturation process was assessed using fibronectin-knockout mouse embryonic fibroblasts (Fn-/- MEFs) and floxed equivalents (Fnf/f MEFs), in fibronectin-depleted growth medium with and without exogenously added plasma fibronectin (full-length, or various fragments). In the absence of full-length fibronectin, Fn-/- MEFs remained homogenously distributed throughout the cell-contracted collagen gels. In contrast, in the presence of full-length fibronectin, both cell types produced shell-like tissues with a predominantly cell-free compacted collagen core and a peripheral surface layer rich in cells. Single cell assays then revealed that Fn-/- MEFs applied lower total strain energy on nanopillar arrays coated with either fibronectin or vitronectin when compared to Fnf/f MEFs, but that the presence of exogenously added plasma fibronectin rescued their contractility. While collagen decoration of single fibronectin fibers enhanced the non-persistent migration of both Fnf/f and Fn-/- MEFs, the migration speed was increased for Fn-/- MEFs on plasma fibronectin fibers compared to Fnf/f MEFs. In contrast, the average speed was the same for all cells on collagen-coated Fn fibers. A Fn-FRET sensor revealed that fibronectin on average was more extended on the microtissue surface compared to fibronectin in the core. Gradients of collagen-to-fibronectin ratios and of the fraction of collagen-adsorbed to stretched fibrillar fibronectin conformations might thereby provide critical cell migration cues. This study highlights a dominant role for fibronectin in tissue morphogenesis and the development of tissue heterogeneities. The research performed by Jasper Foolen leading to these results has received funding from the European Union Seventh Framework Programme (FP7/2007-2013)under grant agreement n°. PIEF-GA-2013-628585Foolen Jasper Competence Center for Materials Science and Technology CCMX0-21108-11Mitsi Maria NIH grantGM74048Chen Christopher S RESBIO Technology Resource for Polymeric BiomaterialsChen Christopher S SwissTransMed grant33/2013 LifeMatrixVogel Viola NCCR Molecular Systems EngineeringVogel Viola European Research Council ERC advanced Grant233157Vogel Viola Swiss National Science Foundation, Grant SNF310030B_133122Vogel Viola The following funding sources are furthermore acknowledged: The research performed by Jasper Foolen leading to these results has received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement n°. PIEF-GA-2013-628585 (JF); the Competence Center for Materials Science and Technology CCMX, Grant 0-21108-11 (MM); NIH grant GM74048 and the RESBIO Technology Resource for Polymeric Biomaterials (CC); SwissTransMed grant 33/2013 LifeMatrix (VV); NCCR Molecular Systems Engineering (VV); European Research Council ERC advanced Grant 233157 (VV); Swiss National Science Foundation, Grant SNF-310030B_133122 (VV). Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Healthy connective tissue maintains a function-specific hierarchical organization of cells and ECM. Formation of tissue anisotropy is thereby driven by biochemical gradients and physical factors, including the need to bear the imposed loads while ascertaining in vivo tissue functionality and durability [1,2,3,4]. However in pathological conditions, external overloading or inflammation of fibrous tissue can cause a progressive increase in cell and matrix reorganization, as in atherosclerosis [5], ventricular overloading [6], cardiomyopathy [7] and tendon overload [8]. In addition, replicating and maintaining this native-like architecture of ECM and cells when engineering tissues in vitro, still remains a challenge. In tissue engineering, where homogeneous polymeric scaffolds are typically presented to cells, the density of cells and associated ECM are found to progressively increase at tissue surfaces [9,10,11,12,13], and in thin tissues this cannot solely be attributed to diffusion limitations of the supplied nutrients. For example, oxygen limitations only start at a tissue depth of 200μm [14]. Recently, the increase in fibroblast density towards tissue surfaces in reconstituted collagenous microtissues, approximately 40μm thick, was correlated with an increasingly anisotropic distribution of fibronectin (Fn) [10,13,15]. The emerging shell-like tissues contained a collagen-rich tissue core that was sparsely populated by cells, but progressively covered by fibronectin-rich surface layers containing high cell densities [10]. However, insight is still missing regarding the underlying mechanisms that drive tissue morphogenesis to generate such cell density gradients. In particular, it is unknown what the role of ECM is and how it affects spatial gradients of the associated processes, such as cell migration and force generation. In the current study we asked how fibronectin in the context of collagen gels impacts cell-induced tissue morphogenesis, specifically the depth-dependent distribution of cells and their ECM. Fibronectin is a dimeric glycoprotein that consists of two almost identical monomers [16,17,18]. Cellular fibronectin, as produced by fibroblasts and many other cells, contains additional segments compared to plasma fibronectin, notably, the EDA and EDB domains [16,17,18,19,20]. Plasma fibronectin is produced solely by hepatocytes [16] and circulates in the blood plasma, as well as in other body fluids. Both plasma and cellular fibronectin interact with cells mostly via the classic integrins αvβ3 and α5β1, but plasma and cellular fibronectin show distinct downstream effects on cell signaling [19,20,21,22,23,24,25,26], and were thus shown to have distinct and independent functions during tissue repair [22]. For cellular fibronectin, the EDA domain contains binding sites for α4β1 and α9β1, while no binding sites were identified so far on the EDB domain [16]. While cells probe their environment, they pull, stretch and deflect the ECM fibers to which they are attached until homeostasis is reached, or until they start to migrate. Fibroblasts are unable to migrate into plasma clots when the clots are depleted of fibronectin [27]. A similar phenomenon is observed for sprouting microvessels from rat aorta explants that are embedded in rat tail collagen gels [28]. The length of these sprouting microvessels was observed to increase with increasing concentrations of plasma fibronectin added to the collagen gel [28]. The addition of fibronectin was also observed to promote migration and matrigel invasion by ovarian cancer cells, in a focal adhesion kinase (FAK)-dependent manner [29]. Since mechanical forces acting on ECM fibers result in conformational changes of fibronectin through stretching and partial unfolding of its domains which all carry specific binding sites [30,31,32,33], fibronectin can be regarded as a mechanotransducer [34]. Stretch-induced conformational changes of the protein exposes cryptic binding sites for fibronectin self-association [32,33,35], reduces the affinity to bacterial adhesins [36,37] and promotes the osteogenic differentiation of human mesenchymal stem cells [38]. Furthermore, the otherwise hidden synergy site, which enhances α5β1-integrin binding to fibronectin, gets exposed when cells pull on fibronectin anchored to rigid but not to soft polyacrylamide substrates, and the exposure of the synergy site correlates with fibronectin-mediated FAK activation in fibrocarcinoma cells [39]. Major synergies exist between fibronectin and other ECM components during tissue remodeling and morphogenesis [20]. While it was initially reported that cellular but not plasma fibronectin is required to induce collagen gel contraction [40], it was later shown that fibronectin-knockout mouse embryonic fibroblasts (Fn-/- MEFs) are still able but limited in their ability to contract type I collagen gels prepared in fibronectin-depleted growth medium (only a ~30% reduction in gel volume was observed) [41], whereby gel contractions up to ~80% is observed with the addition of exogenous plasma fibronectin [41]. Also inhibiting the interaction between collagen and fibronectin through the addition of a fibronectin-to-collagen binding inhibitor (R1R2), reduces gel contraction by Fn-/- MEFs in a dose-dependent manner [42] and severely affects tissue morphogenesis of 3D collagen-fibronectin tissue [43]. Additionally, contraction of a gel mixture of Fn-/- MEFs, type I collagen and fibronectin was significantly reduced upon treatment of the gel with anti-β1 or -α5 integrin antibodies [41]. Fibronectin can thus promote cell-induced collagen gel contraction via the most dominant cell-fibronectin binding integrin, α5β1 [44]. Not surprisingly, fibronectin provides a template for type I and III collagen assembly in tissue repair and development [45,46]. There is a specific role for fibronectin in regulating tissue morphogenesis, as already well established in early development, wound healing and in cancer progression. Model systems using cell-contracted collagen gels gave further insights how fibronectin impacts cell-induced tissue morphogenesis and the resulting heterogeneous distribution of cells and ECM components. By using FRET-labeled fibronectin to sense the conformation of fibronectin in contracted collagen microtissues, Legant et al. [10] showed that two different populations of fibronectin exist, i.e. collagen-absorbed, more compact fibronectin, as well as cell-assembled fibrillar fibronectin that was stretched by cell-generated forces [10]. Interestingly, cells were increasingly located at tissue surfaces with progression of culture time, correlating with the highest densities of fibrillar fibronectin, but the reason why the cells accumulated on the surface of the contracted gels remained unclear. Here, the contribution of fibronectin to the collagen gel microtissue morphogenesis was studied using fibronectin-knockout mouse embryonic fibroblasts (Fn-/- MEFs) and floxed equivalents (Fnf/f MEFs) [47] in combination with a FRET-fibronectin sensor that can directly bind to collagen, as unlabeled fibronectin does as well, and gets incorporate by fibroblasts into their ECM fibrils [30,31]. Since we found here that the presence of plasma fibronectin in the medium, but not of its fragments, has a major effect on tissue morphogenesis by Fn-/- MEFs, cellular traction forces were measured using nanopillar arrays in a secondary series of experiments, while migration properties were determined by tracking migrating single cells on single manually pulled fibronectin fibers. Results Fibronectin is required for the accumulation of cells at the microtissue periphery To address the question of how fibronectin impacts cell-induced microtissue morphogenesis, Fn-/- MEFs and Fnf/f MEFs were mixed with collagen type I and seeded in microwells containing tissue-constraining posts (Fig 1A–1C). Fibronectin-depleted FBS-rich (10%) growth medium was added in all experiments with or without supplementation of exogenous plasma fibronectin. Microtissues produced by Fnf/f MEFs developed into shell-like tissues (Fig 2A and 2B and S1A, S1 and S2 Movies). The core of the collagen scaffold was sparsely populated by cells whereas the periphery contained a high cell density (Figs 2B and S2B and S2D; S2 Movie). Contrary to the cells, plasma fibronectin was found to be more evenly distributed throughout the tissue (Figs 2B and S2B and S2D) as it can directly bind to collagen gel fibers independent of the presence of cells [10]. This confirms previous findings that 3T3 fibroblasts initially immersed in isotropic collagen gels drive the formation of a shell-like tissue morphology [10,15]. In contrast, microtissues contracted by Fn-/- MEFs displayed a strikingly different depth-dependent distribution of the cells 72h after seeding. In the absence of fibronectin, the overall distribution of Fn-/- MEFs in the collagen scaffold remained more homogenous (Figs 2C and S2B; S3 Movie), i.e. distributions of nuclei and collagen highly overlapped (S2B Fig). Adding exogenous plasma fibronectin at the start of the Fn-/- MEFs culture, restored the shell-like composition of their floxed equivalents (Fig 2B), with plasma fibronectin being present throughout the complete tissue and also covering the collagen core (Figs 2D and S2B and S2E; S4 Movie). Interestingly, a much thicker fibrillar layer of plasma fibronectin was assembled by Fn-/- MEFs at the tissue top surface, different from Fnf/f MEFs equivalents, also visible from overall tissue thickness (Fig 2B vs 2D). The formation of shell-like tissues in the presence of plasma fibronectin for both cell types mostly established between 48 and 72h (S1A and S1B Fig; for floxed cells S5, S6 and S1 Movies (24h, 48h and 72h, respectively) and for knockout cells S7, S8 and S3 Movies (24h, 48h and 72h, respectively)). 10.1371/journal.pone.0160369.g001Fig 1 Model platform of tissue morphogenesis to generate fibroblast-contracted collagen microtissues. (A) An 8x8 microwell PDMS template was adhered to a petri dish containing a glass bottom. The white square indicates one single well in which a microtissue was engineered, as depicted in (B): Top view of a biaxially constrained microtissue in a single well of 1125×1125μm, composed of rat tail collagen type I mixed with MEFs. After contraction, tissues measure approximately 800×800×50μm. The white square indicates the area scanned with confocal microscopy (246×246μm) (C) Side view of 4 posts from one microwell. Post diameter was 75μm at the base and 125μm at the top, with a total height of 125μm. 10.1371/journal.pone.0160369.g002Fig 2 Fibronectin contributes to the depth-dependent separation of cells and collagen scaffold in microtissues. Microtissues, cultured for 72h, were either engineered from Fnf/f MEFs in the absence (A) or presence of plasma fibronectin (B), or from Fn-/- MEFs in the absence (C) or presence of plasma fibronectin (D). The depth-dependent distributions of cells and ECM components in a single tissue are represented by volume rendering of a cropped section of the Z-stack using Imaris software in a bottom and top view (1st and 2nd column) and a maximum projection cross-section of a cropped section of the tissue Z-stack with associated depth-dependent intensity histogram from the full Z-stack (3rd and 4th column, nuclei in blue; collagen in green and plasma fibronectin in red)). Tissue depth in the histograms is represented as the distance from the location in the tissue with maximum collagen intensity (0μm). The depth-dependent intensity histogram of all tissues combined is depicted in the 5th column, which represents the raw data. Fnf/f MEFs produce shell like tissues, where cells reside on both sides of the collagen core (A and B). However, in absence of fibronectin, Fn-/- MEFs predominantly reside in the collagen core (C). Fn-/- MEFs supplemented with fibronectin not only migrate towards tissue surfaces, but also assemble fibrillar fibronectin, primarily at the top surface (D). Percentage overlap of collagen and nuclei at the tissue bottom, core and top are quantified in S2 Fig. Exposure to fibronectin enhances Fn-/- MEF migration speed To further understand why Fn-/- MEFs continued to reside within the collagen scaffold in the absence of fibronectin, we asked whether exposure to fibronectin versus collagen directly affects cell migration speed and persistence. Therefore, cell migration speed and persistence was quantified in single cell assays where fibroblast-populated beads were seeded on manually pulled plasma fibronectin fibers [32] (Fig 3A), since we know that the conformation of fibrillar fibronectin, which is often stretched by the cells, is far more extended compared to that of fibronectin adsorbed to flat glass substrates [48] or directly to collagen gels [10] and that the fibronectin conformational distribution within these single fibers closely resembled those observed in a native 3D extracellular environment [32]. The fibers were either left uncoated or were decorated with rat tail collagen type I (matching that of the 3D scaffolds in this study). The aim here was thus to construct fibers that are similar in their physical properties (diameter, rigidity, etc.), but present either fibrillar fibronectin versus collagen 1-decorated fibronectin fibers. To assess solely how cell migration is affected by the presence of different ECM components presented by single fibers, the migration speed and persistence of individual MEFs was tracked using fibronectin-depleted growth medium. Two major observations can be made: First, while the Fnf/f MEFs and the Fn-/- MEFs were highly persistent walkers on the single plasma fibronectin fibers, collagen I decoration increased the turn-around events for both (Fig 3B and 3D). Almost half of both Fnf/f and Fn-/- MEFs changed migration direction over the course of the observation (Fig 3B and 3C, magenta-colored circles). Second and perhaps surprising, the migration speed of Fn-/- MEFs on plasma fibronectin fibers was significantly higher compared to Fnf/f MEFs (Fig 3B and 3D), but decorating the plasma fibronectin fibers with collagen significantly reduced the migration speed of the Fn-/- MEFs to levels approaching those of Fnf/f MEFs (Fig 3B), while collagen coating had little effect on the migration speed of the Fnf/f MEFs. However, persistently migrating cells on collagen decorated fibers did not show a significantly different average speed when compared to non-persistent cells (assessed for both Fnf/f and Fn-/- MEFs individually using Student’s T tests). The efficacy of the collagen coating covering the Fn-fibers was verified by staining for the different components after having tracked the cells in real-time (Fig 3C). 10.1371/journal.pone.0160369.g003Fig 3 Average cell migration speed and persistence along fibers depends on the type of ECM coating. (A) Experimental setup for assessing migration speed and persistence of cells on manually pulled plasma fibronectin fibers. MEFs migrated from microcarrier beads onto fibronectin fibers, either left untreated, or decorated with rat tail collagen type I monomers. (B) Migration speed of single cells was tracked every 15min during 24h and averaged; speed was measured only for single cells (collectively migrating cells were excluded) using the fiber as their substrate. Non-persistently migration cells, i.e. cells that changed migration direction at least once, are marked by a magenta-colored circle. Fn-/- MEFs, compared to Fnf/f MEFs, displayed a significant increase in migration speed on pFn fibers (* p<0.05). Decorating pFn fibers with collagen however significantly reduced migration speed of Fn-/- MEFs, to attain values that lacked a significant difference with Fnf/f MEFs on uncoated fibers. Migration speed of Fnf/f MEFs was not affected by the fiber coating. Migration persistence was negatively affected by collagen decoration, i.e. both Fn-/- and Fnf/f MEFs changed migration direction significantly more often, when compared to their counterparts migrating on pFn fibers. A significant difference in migration speed between non-persistent and persistent cells was not detected. (C) Confirmation of the presence of collagen coating after termination of the migration assay (at 24h). The stained collagen is unlikely to originate from the migrating Fn-/- MEFs, since a specific rat-antibody for collagen is used, the medium does not contain ascorbic acid, and performed experiments were short-term (24h). (D) Raw data of cellular migration speed on manually pulled plasma fibronectin (pFn) fibers, untreated or decorated with collagen I monomers. For each condition, a number of 22 cells (y-axis) were tracked every 15min in at least 2 independent experiments. The color of each dot represents the migration speed (corresponding to the speed, indicated in the bar at the right) of a single cell at a specific time point (x-axis). Although manual tracking of the cells from one time point to the next is susceptible to the introduction of human errors by misinterpretation of the cell location, average cell speed as represented in panel B, is hardly affected. Cells were tracked for 24h maximum, however when encountering other cells, or when leaving the fibronectin fiber, cell tracking was terminated. Since Fn-/- MEFs had a high migration speed, they more quickly encountered other cells, and were therefore often tracked for a shorter time. Non-persistent cells are indicated by a magenta-colored circle just next to the y-axis. Fibronectin enhances the generation of cell traction forces Since cell-generated forces drive tissue morphogenesis [2,3,49,50,51,52], we next asked how the presence of plasma fibronectin in the medium might affect the generation of cell traction forces. Fnf/f or Fn-/- MEFs were seeded on nanopillar substrates coated with fibronectin, or vitronectin (which binds αv, but not α5 integrins), with or without exogenous plasma fibronectin (full-length or fragments) in the medium (Fig 4A). Total strain energy [53,54] was calculated from the pillar deflections measured 2h and 24h after cell seeding (Fig 4B; S3 Fig for average strain energy per pillar). A slight increase was observed for the total strain energy between 2h and 24h for almost all groups, associated with an increased spreading area (as visible from the decreasing average strain energy in time per pillar, S3 Fig), in agreement with previous literature [55]. On fibronectin-coated nanopillars, Fn-/- MEFs generated significantly lower (~45%) strain energy in the absence of exogenously added fibronectin to the growth medium for both time points when compared to Fnf/f MEFs. The addition of exogenous plasma fibronectin to the medium fully rescued total strain energy by Fn-/- MEFs to match their floxed equivalents. This is surprising since in both cases, cells are exposed to plasma fibronectin present as coating on the nanopillars. 10.1371/journal.pone.0160369.g004Fig 4 Total strain energy of fibroblasts on SU-8 nanopillar arrays. (A) Experimental setup for measuring cell-induced nanopillar displacement: Pre-labeled MEFs (red) were seeded on plasma fibronectin (pFn) or vitronectin-coated nanopillars for 30min in Fn-depleted FBS-rich (10%) media (SEM image added at the right, showing how fibroblasts deflect the posts). Pillars have a diameter of 250nm, a height of 1.5μm and center to center distance was 800nm. Medium was supplemented without or with 45nM fibronectin (full-length, or the 40k, 70k or 120kDa fragments), during cell seeding, and pillar displacements were measured 2 and 24 hours after cell seeding. (B) Total strain energy per cell, 2h and 24h after seeding, for different pillar coatings (fibronectin versus vitronectin) and in the presence of exogenous pFn and of its fragments. For pillars coated either with fibronectin or vitronectin, pFn in the medium upregulates total strain energy generated by Fn-/- MEFs approaching those of Fnf/f MEFs. On pFn-coated nanopillars, addition of pFn in the medium significantly increased total strain energy per Fn-/- MEF attaining values that equal Fnf/f MEFs. Vitronectin coating significantly decreased the strain energy by Fn-/- MEFs (indicated by #). On vitronectin coated pillars, the 70kDa fragment significantly increased strain energy of Fn-/- MEFs (indicated by £), which is likely to be caused by a possible contamination of full length fibronectin in this particular fragment, S4 Fig. However, only full length exogenously added pFn rescues total strain energy by Fn-/- MEFs on fibronectin or vitronectin-coated pillars to meet their floxed (Fnf/f MEFs) counterparts. For a representation of the average strain energy per pillar, see S3 Fig. To detect the extent to which fibronectin-coating alone affected the total strain energy, the experiment was repeated for Fn-/- MEFs on vitronectin coated nanopillars, again for both 2h and 24h after seeding. Vitronectin coating significantly reduced total strain energy (~35%), compared to fibronectin coating (Fig 4B; S3 Fig for average strain energy per pillar). On vitronectin-coated pillars, the addition of exogenous plasma fibronectin to the medium again fully rescued the strain energy generated by Fn-/- MEFs (Fig 4B; S3 Fig for average strain energy per pillar). Hence, fibronectin, both as substrate coating and to a far greater extend when present in the growth medium, increased the total cellular strain energy. Since fibronectin contains binding sites for various integrins [16], we next asked whether exogenously added full-length plasma fibronectin was required to upregulate cell traction forces, or whether certain fibronectin fragments, none of which can undergo fibrillogenesis, are equally efficient. Fn-/- MEFs were therefore seeded on vitronectin coated nanopillars and exposed to fibronectin fragments supplemented to the medium. The 70kDa N-terminal fibronectin fragment (repeats FnI1-9 and FnII1-2) [56] provides integrin binding sites for α5β1 and contains the collagen/gelatin binding region [16]. The 40kDa C-terminal fibronectin fragment (repeats FnIII12-14) [56] contains the heparin binding domain II, while the 120kDa central fragment (repeats FnIII1-11) [56] provides many integrin binding sites, among others the RGD epitope, the α5β1 epitope and the synergy site [16,41]. Although addition of the 70kDa fibronectin fragment significantly increased the total cellular strain energy (~50%) at both time points, it did not reach the strain energy levels generated in the presence of full-length fibronectin (Fig 4B). This 70kDa-mediated increase is likely to be caused by a modest contamination of what could be full length fibronectin, as detected by Western blotting of the 70k fragment using a polyclonal fibronectin antibody (S4 Fig). Also the other fibronectin fragments used (40kDa and 120kDa) did not increase total strain energy, irrespective of the time point. Thus, the presence of full-length fibronectin in the medium, and not of its fragments, is required to fully rescue the cell traction forces of Fn-/- MEFs. Full-length fibronectin is required for the development of steep cell density gradients Since the magnitude of traction forces and developing tissue heterogeneity may be linked, Fn-/- MEFs were mixed with collagen I to produce microtissues, while being exposed to the different fibronectin fragments (40, 70 or 120kDa) that were supplemented to the growth medium from the start of culture. Even though all of these fragments were found to bind to the collagen gel, cell nuclei remained homogeneously distributed throughout the collagen gel (Figs 5 and S2C; S9–S11 Movies). Tissue contraction appeared to be further limited by the addition of the fragments, since the tissues were generally observed to be thicker than Fn-/- MEFs in collagen only (Fig 2C). Thus, cells require full-length fibronectin for tissue morphogenesis, i.e. the accumulation of fibroblasts at the surface of the contracted collagen gels. 10.1371/journal.pone.0160369.g005Fig 5 Fibronectin fragments (40, 70, 120kDa) do not contribute to the depth-dependent separation of cells from the collagen core. Microtissues, cultured for 72h, were engineered from Fn-/- MEFs mixed with rat tail collagen type I in the presence of fibronectin fragments, either 40kDa (A), 70kDa (B) or 120kDa (C). In the presence of the fibronectin fragments, Fn-/- MEFs predominantly resided in the collagen core. The depth-dependent distributions of cells and ECM components in a single tissue are represented by volume rendering using Imaris software in a bottom and top view (1st and 2nd column) and a maximum projection cross-section of the tissue with the associated depth-dependent intensity histogram (3rd and 4th column, nuclei in blue; collagen in green and fibronectin fragments in red, all stained using a polyclonal antibody). Tissue depth in the histograms is represented as the distance from the location in the tissue with maximum collagen intensity (0μm). The depth-dependent intensity histogram of all tissues combined is depicted in the 5th column, which represents the raw data. Percentage overlap of collagen and nuclei at the tissue bottom, core and top are quantified in S2 Fig. Gradient of fibronectin conformations, from the core to the microtissue periphery Since a previous study showed that FRET-labeled plasma fibronectin is assembled by 3T3 cells into fibrillar ECM at the microtissue periphery, much more than in the core where it was mostly observed to adsorb to collagen fibers [10], FRET-labeled plasma fibronectin was added to the growth medium of Fn-/- versus Fnf/f MEFs-populated microtissues to assess its depth-dependent conformational gradients. The FRET signal (calibrated using previously designed methods [31], S5 Fig) was detected after 24, 48, 72 and for Fn-/- MEFs also 96h after the start of culturing. Fibronectin was observed to be highly stretched at the tissue surfaces (corresponding to low FRET) that gradually decreased (corresponding to high FRET) towards the core (Figs 6A, S12–S18 hin for Fn-/- MEFs, tissues grew thicker over time (distributions widened in time, similar to the results shown in S1B Fig). For both Fn-/- and Fnf/f MEFs, the complete FRET histograms of fibronectin moved over time to lower FRET values, while roughly maintaining their shape (Fig 6A, graphs at the right). This decrease in average Fn-FRET with culture time happens as the cells contract the collagen gels, coinciding with a possible change in the ratio of collagen-absorbed, more compact fibronectin vs extended fibrillar fibronectin assembled from the growth medium by fibroblasts [10]. Remarkably, from the 48h time point onwards, the FRET-ratios in the core of the collagen gels contracted by the Fn-/- MEFs lagged behind in time approximately 24h, compared to Fnf/f MEFs. 10.1371/journal.pone.0160369.g006Fig 6 The Fn-FRET probe reveals a steep gradient of fibronectin conformations. (A) MEFs embedded in collagen type I gels were incubated up to 96h in Fn-containing growth medium supplemented with trace amounts of FRET-Fn. The left panel displays representative cropped average intensity projection images showing the depth-dependent FRET signal in microtissue cross sections at the indicated time points. The right panel shows the distribution of the FRET ratio throughout all tissue Z-stacks for each measured tissue (images were taken every 1μm in the Z-direction). The maximum FRET ratio was used to set tissue depth to 0μm. The colored bar in between both graphs represents the color-coded pFn-FRET ratio, corresponding to the left panel. Values correspond to FRET ratios for fibronectin in solution progressively denatured by GdnHCl, as depicted in S5 Fig. For both the left and right panel, the average pFn-FRET-ratios decreased from the tissue core towards the outer surfaces. In addition, although the shape of the fibronectin conformational distribution remains unchanged with increasing culture time, average FRET ratios drop with time, whereby Fn-/- MEFs lag behind on Fnf/f MEFs approximately 24h (see also panel B). (B) Comparison of maximum pFn-FRET values between groups (representing the average value at 0μm tissue depth), confirming that maximum values for Fn-/- MEFs at time point 48h and 72h are significantly higher compared to Fnf/f MEFs. The * represents a significant difference (p<0.05) between time-matching groups, assessed with student t-test. Discussion Since the underpinning mechanisms that drive the formation of cell gradients and gradients of ECM composition in tissues and engineered tissue scaffolds are not yet resolved [9,10,11,12,20,21,22,27,28,41,42,43,45,57], we exploited 3D collagen-derived microtissues as model system to investigate the role of fibronectin in such processes. During contraction of the collagen gels by fibroblasts, cells accumulate at the microtissue surface (S1 and S2 Figs), in agreement with previous observations [10,13]. Using fibronectin knockout (Fn-/-) versus floxed mouse fibroblasts (Fnf/f), we found that fibronectin plays a crucial role in directing the accumulation of cells at the tissue surface (Fig 2). The current study revealed that the absence of fibronectin abolished the accumulation of cells at the tissue surfaces (Fig 2C). The microtissue platforms allowed us to show that the presence of exogenous plasma fibronectin, but not of the 40kDa, 70kDa or 120kDa fibronectin fragments, rescued the ability of knockout fibroblasts to accumulate at the tissue surface (Figs 2 and 5). This is especially interesting in the context of the ability for full-length fibronectin to undergo fibrillogenesis, which is lacking for the fibronectin fragments alone [41,58,59]. This observation supports the possibility that gradients in the conformation of fibronectin, as probed here by our Fn-FRET sensor, perhaps in combination with changing fibronectin-to collagen ratios, direct tissue morphogenesis by guiding cells towards the tissue surfaces. Cell migration assays on single fibers revealed that Fn-/- MEFs migrate faster on plasma fibronectin fibers, when compared to collagen-decorated fibronectin fibers (Fig 3). In contrast, Fnf/f MEFs show similar migration speeds, irrespective of the Fn-fiber coating. The positive effect of fibronectin on the migration properties of Fn-/- MEFs was previously shown in a scratch assay, where the addition of plasma or recombinant wild type fibronectin increased wound closure by Fn-/- MEFs on 2D collagen coated surfaces [42,59]. Also, C2C12 cells migrate further on fibronectin-coated surfaces, compared to gelatin [60]. In 3D, fibronectin depletion from plasma clots prevented fibroblast invasion [27], and invasion of collagen constructs by tumor cells is enhanced upon addition of exogenous fibronectin [61].Here we show that collagen decoration of fibronectin fibers enhanced non-persistent migration by both Fnf/f and Fn-/- MEFs (Fig 2). This may explain why Fn-/- do not accumulate at tissue surfaces in the absence of fibronectin. Although Fnf/f MEFs were equally affected by collagen decoration of fibronectin fibers in decreasing persistent migration, our data show that their ability to produce cellular fibronectin allows them to accumulate at microtissue surfaces, even in the absence of plasma fibronectin. A possible explanation lies in the key difference between the manually pulled fibers and 3D: as the Fnf/f cells migrate on collagen-decorated fibers, they produce and assemble cellular fibronectin matrix underneath themselves which might override the collagen “effect” seen in the accelerated movement of the Fn-/- cells. In 3D, the cells can encounter endogenous fibronectin produced by other cells while they migrate through the ECM. While we acknowledge that directly relating observations made from cells migrating on manually pulled fibers to phenomena in 3D cell-assembled ECM is difficult, manually pulled fibers and cell-derived ECM fibers do share similarities. The conformational distribution of Fn is similar on both types of fibers [32]. Another constraint that potentially limits Fn-/- MEFs to accumulate at tissue surfaces on collagen scaffolds is the required presence of MMPs for migration through collagen ECM of small pore sizes [62]. Since fibronectin is known to upregulate the production of various MMPs [26], this could be a means by which Fn-/- MEFs travel through the collagenous rich ECM to accumulate at the tissue surface. Additionally, fibronectin was shown to induce cell proliferation of Fn-/- MEFs in 3D collagen constructs [43,63]. The addition of fibronectin to Fn-/- MEFs in our setup very likely also resulted in increased proliferation, as visible from the volume rendered images, compare Fig 2C with 2D. However, since the tissue core is completely deprived of cells with the addition of fibronectin after 72h, it seems unlikely that local proliferation at tissue surfaces has been the sole factor to drive the special separation of cells and ECM, i.e. cell migration through the ECM is proposed to have played a dominant role. Taken together with our finding that Fn-/- MEFs embedded in a 3D collagen scaffold resided in the collagen core (Fig 2), and the above-mentioned literature, we conclude that the presence of fibronectin is essential and drives the formation of a microtissue with well-defined gradients in cell density and ECM composition. We furthermore observed that plasma fibronectin added to the medium upregulated the generation of traction forces by Fn-/- MEFs pulling on nanopillars coated with plasma fibronectin compared to vitronectin as probed 2h and 24h after cell seeding (Fig 4). In agreement with these results, previous work showed that type I collagen gel contraction by Fn-/- MEFs is limited when fibronectin fragments or vitronectin was added to the gel, when compared to adding full-length fibronectin [41]. Additionally, fibronectin upregulates 3T3 fibroblast contractility and migration compared to RGD-only modified substrata [64], and that vitronectin and fibronectin, although they share binding sites for several integrins, namely αvβ1 [65], αIIbβ3 (platelet integrin), α8β1 and αvβ3 [44], are distinct in αvβ5 (vitronectin) versus α5β1 and αvβ6 integrins (fibronectin) [44]. Our finding that fibroblasts apply enhanced traction forces, giving rise to an enhanced total strain energy to fibronectin versus vitronectin coated pillars is in agreement with the literature, i.e. integrin α5β1 but not αvβ3 was shown to increase the ability of fibroblasts to sustain force [66], and quantitative proteomics linked αv-class integrins to a GEF-H1-RhoA pathway coupled to the formin mDia1, but not myosin II, and α5β1 integrins to a RhoA-Rock-myosin II pathway [67]. Strikingly, only the presence of full-length plasma fibronectin in the medium, and not of its fragments, could recover the forces by which Fn-/- MEFs pulled on either vitronectin or fibronectin-coated nanopillars. The 70kDa fragment partially rescued the forces; however this is very likely the result of a minor contamination with possible full length fibronectin as shown by Western blot (S4 Fig). Consequently, not just the integrin junctions formed with surface-bound vitronectin, but the presence of fibronectin in the medium positively contributes to cellular traction force generation, suggesting activation of additional integrins through adsorption of fibronectin from solution to the cell membrane surface. This mode of integrin activation is supported by previous studies performed on fibroblasts adhering to RGD-functionalized supported lipid membranes and showing that integrin activation occurs before the cells apply tensile forces to those integrins [68]. Such effect, i.e. the effect of a soluble protein in the growth medium on cell behavior was also recently demonstrated for FAK-activation in freely suspended cells in growth medium by adding soluble collagen [39]. Although a direct extrapolation of results from nanopillars to 3D constructs is difficult to make, lower force generation in the presence of fibronectin fragments, compared to full-length fibronectin, corroborates with our observation in the 3D constructs that tissue compaction in the Z-direction appeared to be limited in the presence of fibronectin fragments, compared to all other groups. The latter agrees with a study by Hocking et al. [41], who also observed limited gel contraction in the presence of the same fragments. The high cell densities at the microtissue surfaces correlated with fibronectin that showed significantly lower FRET than regions in the core of the microtissue (Fig 6). Since we show that the fibronectin FRET-profile changes significantly at the length scale of a single cell, a depth-dependent gradient of fibronectin conformations in these microtissues exists, from more compact in the core of the contracted collagen to more extended and partially stretched fibrillar fibronectin at the tissue periphery. Since it was shown recently [39] that fibronectin-mediated FAK activation is dependent on the mechanical tension acting on fibronectin, which may expose its otherwise hidden synergy site to integrin α5, the accumulation of fibroblasts at the microtissue surfaces may thus originate from increasing the fraction of collagen-adsorbed to stretched fibrillar fibronectin conformations, from the tissue core to the surface, which might thereby provide conformational-dependent guidance cues. Different fibronectin conformations could also indirectly act as a cell guiding cue by differential exposure of binding sites for chemoattractant growth factors (e.g. PDGF, VEGF, TGFβ and FGF) [19,69,70,71]. The preferred accumulation of cells at the top surface of microtissues, compared to the bottom, is a feature that has been observed before as well [10], but the causative factors for this phenomenon are still unknown. Microtissues have previously been observed to contract around posts within 12h [15,72] and are thus suspended up to 75μm above the substrate at a stage well before asymmetries in cellular distributions from apical to basal appear. Although oxygen limitations were shown to only be a limiting factor in tissues that exceed 200μm [14], and our microtissues are approximately 4 times thinner (~50μm), accessibility of larger molecules by cells at the basal side may be limited especially in the first phase of gel contraction when compared to the apical side. We thus acknowledge that the difference in cell density from apical to basal may be driven by chemotaxis and thereby presents an artifact of the model system. The central part of microtissues contains high collagen densities (Fig 2) and high FRET-values (Fig 6), while moving towards the tissue surface, collagen density and FRET progressively decrease. The fibronectin FRET-ratios in microtissues populated by Fn-/- MEFs were shifted to significantly higher values at the time points 24h, 48h and 72h (Fig 6B), respectively, indicating that Fn-/- MEFs may contract the tissues at a slower rate. Locations in the tissue that contain high collagen densities may correlate with a high degree of fibronectin that is absorbed to the collagen, as shown earlier for collagen-adsorbed fibronectin compared to fibrillar fibronectin [10]. The FRET decrease over time in the core may indicate that the initially adsorbed fibronectin may be gradually released from the collagen over time, thus changing the ratio between absorbed and fibrillar fibronectin. In conclusion, fibroblasts accumulated at microtissue surfaces in the presence of full-length fibronectin. Building order from randomly assembled cells requires cell migration and force generation, and both were shown here to be promoted by full-length fibronectin on manually assembled fibers and nanopillars, respectively. Unraveling fibronectin-induced mechanisms that drive spatial organization of cells and ECM, and thus tissue morphogenesis, is of vital interest in development, as well as to build and maintain engineered tissue functionality and durability. Materials and Methods Microtissue fabrication A previously developed model system was used [73,74], composed of an array of 8 by 8 microwells, with each well containing 12 posts with a narrow base and wide top (total height 125μm) in a square setup (Fig 1). Before use, the model systems were made hydrophilic by plasma treatment (1 min at 100W) to facilitate gel seeding. Subsequently, they were sterilized by treatment with 70% alcohol for 30min and UV exposure for 15min, and finally treatment with 0.2% Pluronic F127 (BASF) for 2 min to reduce cell adhesion. Fibronectin-knockout (Fn-/-) or fibronectin floxed (Fnf/f) mouse embryonic fibroblasts (MEFs) [25,47,75] were obtained from Professor Rainhard Faessler (Max-Planck Institute for Biochemistry, Martinsried, Germany). Cells were added to a gel mixture of growth medium (DMEM low glucose with glutamax, Gibco; 10% Fn-depleted FBS, Biowest; 1% penicillin/streptomycin, Gibco), collagen type I (final concentration 1mg/mL, rat tail, BD Bioscience) and sodium hydroxide to neutralize the pH of the acidic collagen. Gels were mixed with cells to attain a final density of 106 MEFs per mL gel. All MEFs used were below passage 10. To each microwell, approximately 0.16μL of the cell-gel mixture was pipetted (the volume of a single well). During this procedure, microwells were kept on ice to prevent evaporation of the gel. Gels subsequently polymerized in an incubator at 37°C, 5% CO2 for 10min. To prevent dehydration of the gel during polymerization, the petri dish was inverted and sterile water was added to the lid. After polymerization, growth medium was added to the gels, which contracted around the posts in approximately 6h [73,74]. Where indicated, the growth medium was supplemented with 40nM fibronectin (unlabeled plasma fibronectin, single or double (FRET) labeled plasma fibronectin, or fibronectin fragments, i.e. 40kDa, 70kDa or 120kDa). Supplemented concentrations of full-length fibronectin, as indicated in this manuscript, are derived from the molecular weight of the monomer and thus theoretically contain twice the amount of binding sites compared to added fragments. Fibronectin fragments are purchased from Sigma and obtained through proteolysis of human plasma fibronectin. FRET-labeled fibronectin was always mixed with unlabeled fibronectin at a ratio of 1:10 to avoid signal from intermolecular FRET. Fibronectin was added just after tissue seeding and remained in the culture throughout the complete experiment. Engineered microtissues were cultured up to 96h. Visualization and distribution analysis After the designated culture period, microtissues were fixed in 10% formalin for 30min and permeabilized for 30min in 0.5% Triton-X in PBS (visualization of FRET-fibronectin was performed in absence of triton-X treatment). The following antibodies were used for specific staining of ECM components and cell nuclei: 1F10C2 (Chondrex) for rat tail collagen type I, ab23750 (Abcam) for fibronectin, and Dapi (Fluka) for cell nuclei. Exogenously added plasma fibronectin was pre-labeled with Alexa488 (A20000, Invitrogen). Z-stack fluorescent images were taken with a confocal microscope (Leica SPX5). The pinhole of the photo-multiplier was set to the optimal width for the used 63x, 1.2N.A. water immersion lens (Zeiss Plan Apo CS). Photo-multipliers accepted wavelength regions of 418–461nm for Dapi, 498-544nm for fibronectin (alexa488), or 643-739nm for collagen (alexa633). Microtissues anchored to the posts were visualized through the glass bottom of the petri dish. Individual images were taken every 1μm through the complete tissue depth, which measured 246 x 246μm each (as shown in Fig 1B). No additional image processing was performed. A representation of the distribution of cells, collagen and fibronectin throughout the tissue depth was obtained by processing maximum projection images from the Z-stacks using ImageJ, which were cropped in the x-direction to 75μm for a clear representation of the cell and protein distributions in the Z-direction. Emission intensity profiles throughout the tissue depth were obtained using ImageJ for each tissue, using the complete stack, not the cropped version. For each condition, intensity profiles were normalized, i.e. maximum emission for each channel was set to 1, after which profiles within the same condition were averaged. Averaging was performed after shifting the intensity profiles over the x-axis to have the maximum collagen intensity slice at the exact same location (set at a tissue depth of Z equals 0μm), explaining the lack of a standard deviation at this specific depth for collagen. Tissue depth is represented as distance from the location in the tissue with maximum collagen intensity, which therefore equals the arbitrary value of 0μm. Imaris software was used to produce volume rendered representative parts (cropped to 100x100μm) of the 3D stacks from single tissues. Contrast and brightness of individual channels (nuclei, collagen and fibronectin) were visually optimized for each tissue. Direct comparison between samples in terms of intensity is not valid. The volume rendered images merely aim to give the reader a 3D view of the differences between cellular distributions throughout tissue depth. The distribution of cells and ECM (collagen and fibronectin) throughout microtissue thickness was quantified as follows. For each tissue, the tissue core was defined by all Z-stack slices from the collagen channel that had a normalized fluorescent intensity of 0.5 or higher (S2A Fig, vertical dashed line intersecting with the green line: representing collagen), thereby defining 3 regions in the tissue, i.e. bottom, core and top. Intensity values belonging to the three different regions, were summed and normalized to the total sum of intensity values, thus normalizing to 100%. Since high intensity values for the collagen channel are taken as a reference, the tissue core contains the majority of the collagen. Tissue bottom and top of the microtissues thus have limited amount of collagen and represent the basal and apical side, respectively. Subsequently, the sum of intensity values for the Dapi and fibronectin channels were calculated in the same way, but always based on the ‘bottom, core, top’-distribution of the collagen channel. The visual representation of these distributions thus represent the fraction of cells or fibronectin that overlap with the collagen core, and the top and bottom surface of the microtissues. To determine significant differences between bottom, core and top, One-Way-ANOVA, Bonferroni post hoc test was applied to the Dapi and fibronectin distributions. Fibronectin isolation, fluorescent labeling, and FRET imaging and analysis Fibronectin was isolated, (FRET-)labeled and analyzed, including a characterization via chemical denaturation curves, as previously described [31]. In short, fibronectin was purified from human plasma (Swiss Red Cross). The cysteines on the fibronectin modules FnIII7 and FnIII15 were labeled with Alexa Fluor 546 as acceptor fluorophores (A, maleimide, Molecular Probes), whereas amines were randomly labeled with Alexa Fluor 488 as donor fluorophores (D, succinimidyl ester, Molecular Probes). The labeling ratio of Fn-DA was determined by measuring the absorbance of at 280, 498 and 556nm. On average, 13.5 donors and 4.0 acceptors were labeled on each fibronectin dimer. Upon stretching fibronectin, the distance between the multiple acceptors and donors increases resulting in a reduction in the acceptor to donor intensity ratios, here referred to as FRET (IA/ID). The use of multiple donors and acceptors prevents the calculation of FRET efficiency, but is sensitive to a large range of conformations. FRET–labeled fibronectin in PBS was stored at -20°C and added to cultures immediately after thawing. All images that involve FRET measurements were acquired with an Olympus FV-1000 confocal microscope with a 40×0.9 N.A. water immersion objective, with a set pinhole diameter of 200μm. Slices from each Z-stack were acquired at 318x318μm with a slice spacing of 1μm. All images were acquired with a 3xKalman line averaging. Tissues were excited with a 488nm laser, while emission of the donor and acceptor were detected using photomultiplier tubes (PMTs) with detection ranges set at 514–526nm (donor channel) and 566–578nm (acceptor channel). All images were acquired with the exact same settings, i.e. PMT voltage, laser power and pixel dwell time, to prevent hardware related misinterpretation of the FRET-values. Donor bleed into the acceptor channel was determined by imaging the emission of 488-labeled Fn in both channels. Donor bleed was constant throughout tissue depth and averaged 0.207 ± 0.001, meaning that approximately 21% of the donor channel was present in the acceptor channel. Because of the linear relation, i.e. donor bleed being constant throughout tissue depth, the shape of the FRET-curves were not affected. Obtained Z-stack images were analyzed using a modified version (automation of FRET measurements for Z-stacks was implemented) of a previously designed script in Matlab [10,31]. FRET-distributions were obtained by calculating the average FRET value for each image in the Z-stack throughout the complete tissue depth. Previous work [10] showed that the FRET-signal itself is not affected by imaging through collagen-based microtissues and thus truly represents a conformation change in the molecule. The batch calibration of fibronectin FRET-ratios was performed using the technique as previously described [31] by dissolving FRET-labeled fibronectin at different concentrations of the denaturant GdnHCl. The settings for visualization of the calibration curve using the Olympus microscope were maintained when imaging FRET in 3D constructs. Migration assay Migration speed and persistence of MEFs was assessed on manually pulled plasma fibronectin fibers, based on a previously developed system [32,76]. Fibronectin fibers were pulled from solution (0.4 mg/mL plasma fibronectin in PBS) and deposited on silicone sheets (SMI .005” NRV G/G 40D, Specialty Manufacturing, Saginaw, MI) that were cut to fit a 4-well LabTek chamber (#155382, Thermo Scientific). Fibronectin fibers were washed gently with PBS and treated with 4% BSA in PBS for 60min. After washing with PBS, fibers were left untreated or were decorated for 1h with rat tail collagen I (10μg/mL in DMEM at 37°C, BD Bioscience). Subsequently, all fiber-containing silicone sheets were treated overnight with 20 mg/mL BSA, 10% PenStrep in PBS. After the migration assay, fibers were imaged for the presence of collagen decoration for validation purposes. Cytodex microcarrier beads (C3275, Sigma) coated with MEFs to near confluence were used as sources of cells that could migrate away onto the fibronectin fibers. Briefly, MEFs (0.3million) were suspended in 1mL growth medium and added to untreated 24-well plates (351147, Falcon). Subsequently, 200μL of Cytodex beads (104 beads per mL of PBS) were added to the wells. MEFs attached to the beads in the subsequent 72h of culturing. Cell-populated beads were transferred to fibronectin fibers (uncoated or coated with collagen) in fibronectin-depleted FBS-rich (10%) growth medium. After 6h, MEFs started to migrate from the beads onto the fibers. Migration was imaged time-lapsed using an Axiovert 200 M inverted microscope (Carl Zeiss). Images were taken every 15min for 24h at multiple locations per condition. Migration speed was quantified using ImageJ software by selecting the position of individual cells at every time point. Only those MEFs were tracked that migrated as a single cell and used the fiber as guidance. Non-persistent cells were marked when changing direction at least once during tracking. Nanopillar experiments SU-8 nanopillar arrays (SU-8 2000.5, Microchem) were fabricated by a combination of nanosphere lithography and plasma etching, attached to glass coverslips, as previously described [77]. Nanopillars were coated with vitronectin (10nM) or plasma fibronectin (10nM) via incubation for 30min. MEFs, pre-labeled with membrane dye DiI (Invitrogen), were allowed to spread on coated nanopillars at 7,500cells/cm2 in fibronectin-depleted growth medium. Growth medium was, where indicated, supplemented with plasma fibronectin or the fibronectin fragments; 40kDa, 70kDa and 120kDa (final concentration 45nM). After 2h or 24h, cells were fixed in 10% formalin for 30min after which pillar deflection and cell morphology was imaged using a Leica confocal microscope SP5 with a 63x, oil immersion, 1.4 N.A. objective. Pillar displacements were measured using Diatrack 3.03 (Powerful Particle Tracking, Semasopht). Atomic force microscopy was used to determine the spring constant of single nanopillars, resulting in an average value of 79 ± 3 nN/μm. To calculate the total strain energy for each cell, the displacement of each post was squared and multiplied by half the spring constant of a nanopillar and summed for all pillars that connected to that cell, as described previously [53,54]. Statistics For the migration assay, One-Way ANOVA was used to determine the effect of the coating and its interaction with migration speed. For cellular force measurements on the nanopillars, One-Way ANOVA was used to determine the effect of the combination of cell type and growth medium content (using the same coating, either fibronectin or vitronectin) and its interaction with total strain energy. For comparing the distributions of nuclei and fibronectin relative to the collagen core (S2 Fig), One-Way ANOVA was used to detect significant differences in the percentage of cells or Fn between bottom, core and top of tissues from different conditions. For ANOVA, all p-values were corrected with the Bonferroni criterion. Student’s t-tests were performed to check for significant differences between MEFs (analysis performed for Fnf/f and Fn-/-, separately) that showed persistent migration vs non persistent migration on collagen decorated pFn fibers. Student T-tests were also performed to test for significant differences between maximum fibronectin FRET-value and cell type (Fn-/- versus Fnf/f MEFs) for each time point (24h, 48h and 72h) and for total strain energy of Fn-/- MEFs on fibronectin vs vitronectin coated nanopillars. P-values <0.05 were considered statistically significant. The indicated sample size originates from at least 2 different wells, in which differences between the independent experiments were not observed. Supporting Information S1 Fig The presence of fibronectin contributes to microtissue morphogenesis in a time-dependent manner. Tissues either contained Fnf/f MEFs (A) or Fn-/- MEFs (B) in rat tail collagen gels, supplemented with plasma fibronectin in the growth medium. MEFs produced shell like tissues from 48-72h, where at 72h cells reside primarily apically from the collagen core, with sparse cells at the basal side of the collagen core. The depth-dependent distributions of cells and ECM components in a single tissue are represented by volume rendering using Imaris software in a bottom and top view (1st and 2nd column) and a maximum projection cross-section of the tissue with associated depth-dependent intensity histogram (3rd and 4th column, nuclei in blue; collagen in green and plasma fibronectin in red)). Tissue depth in the histograms is represented as the distance from the location in the tissue with maximum collagen intensity (0μm). The depth-dependent intensity histogram of all tissues combined is depicted in the 5th column, which represents the raw data. Strikingly, Fn-/- MEFs assemble more fibronectin at the tissue surface, visible at 72h, compared to their floxed counterparts. To note, the data presented for Fnf/f MEFs at 72h resemble Fig 2A, while data for Fn-/- MEFs at 72h resemble Fig 2C. Percentage overlap of collagen and nuclei at the tissue bottom, core and top are quantified in S2 Fig. (TIF) Click here for additional data file. S2 Fig Quantification of the depth-dependent collagen-nuclei overlap in microtissues. (A) Method of quantifying the distributions of cells and ECM (collagen and fibronectin) throughout microtissue depth, as explained in the materials and methods section. The collagen core is defined as the sum intensities from an arbitrary threshold of 0.5, resulting in three regions, i.e. bottom, core and top. Subsequently, percentages based on sum intensity for the three different tissue sections are calculated for nuclei (blue curve), collagen (green curve) and fibronectin (red curve). (B-D) Distributions of nuclei, collagen and fibronectin in the different tissue zones (bottom, core and top), in which error bars represent standard deviations. Analyses represent histograms presented in Fig 2 (S2B Fig), Fig 5 (S2C Fig: fragments) and S1 Fig A (S2D Fig: time-course of Fnf/f MEFs) and S1B Fig (S2E Fig: time-course Fn-/- MEFs). Analysis of statistical differences between the percentages in ‘bottom’, ‘core’ and ‘top’ are performed using One-Way-ANOVA with a Bonferroni post hoc test. (EPS) Click here for additional data file. S3 Fig Cellular traction forces (2 and 24h after seeding), represented by average strain energy per pillar, for different pillar coatings (fibronectin versus vitronectin) and in the presence of exogenously added pFn and of its fragments. Compared to total strain energy (Fig 4), similar trends are visible, however values at 24h for average strain energy are slightly lower, resulting from an increased spreading area between 2h and 24h after cell seeding. (EPS) Click here for additional data file. S4 Fig Western blot analysis of the 70k fibronectin fragment shows a possible contamination with full length fibronectin. 1μg of the 70k fibronectin fragment was loaded and probed with a rabbit polyclonal antibody against fibronectin (ab23750, Abcam). Although the majority of the protein consists of the 70k fragment, a possible contamination of fibronectin monomer can be seen at band size 250. This may explain the slightly elevated cellular traction forces measured using the nanopillar assay. (EPS) Click here for additional data file. S5 Fig Calibration of fibronectin FRET-ratios in solution upon progressive denaturation. FRET-labeled fibronectin was dissolved in different concentrations of the denaturant GndHCl. The loss of secondary structure of the fibronectin protein starts beyond the concentration of 1M GndHCl [31,48] (corresponding to acceptor-versus donor intensities, Ia/Id = 0.63 and higher). The protein is completely denatured at GndHCl concentrations of 4M (la/ld = 0.37). Left: Probability density distributions of FRET-fibronectin in solutions at different GndHCl concentrations. Right: The denaturation curve containing average values from 3 individual measurements of the probability density distributions are presented in combination with the standard deviation. (EPS) Click here for additional data file. S1 Movie Representative Z-Stack of a tissue from Fig 2A: Fnf/f MEFs in a collagen gel at 72h of culturing. (AVI) Click here for additional data file. S2 Movie Representative Z-Stack of a tissue from Fig 2B: Fnf/f MEFs in a collagen gel at 72h of culturing with exogenously added plasma fibronectin. (AVI) Click here for additional data file. S3 Movie Representative Z-Stack of a tissue from Fig 2C: Fn-/- MEFs in a collagen gel at 72h of culturing. (AVI) Click here for additional data file. S4 Movie Representative Z-Stack of a tissue from Fig 2D: Fn-/- MEFs in a collagen gel at 72h of culturing with exogenously added plasma fibronectin. (AVI) Click here for additional data file. S5 Movie Representative Z-Stack of a tissue from S1 Fig A 24h: Fnf/f MEFs in a collagen gel at 24h of culturing with exogenously added plasma fibronectin. (AVI) Click here for additional data file. S6 Movie Representative Z-Stack of a tissue from S1A Fig 48h: Fnf/f MEFs in a collagen gel at 48h of culturing with exogenously added plasma fibronectin. (AVI) Click here for additional data file. S7 Movie Representative Z-Stack of a tissue from S1B Fig 24h: Fn-/- MEFs in a collagen gel at 24h of culturing with exogenously added plasma fibronectin. (AVI) Click here for additional data file. S8 Movie Representative Z-Stack of a tissue from S1B Fig 48h: Fn-/- MEFs in a collagen gel at 48h of culturing with exogenously added plasma fibronectin. (AVI) Click here for additional data file. S9 Movie Representative Z-Stack of a tissue from Fig 5A: Fn-/- MEFs in a collagen gel at 72h of culturing supplemented with the 40kDa fibronectin fragment. (AVI) Click here for additional data file. S10 Movie Representative Z-Stack of a tissue from Fig 5B: Fn-/- MEFs in a collagen gel at 72h of culturing supplemented with the 70kDa fibronectin fragment. (AVI) Click here for additional data file. S11 Movie Representative Z-Stack of a tissue from Fig 5C: Fn-/- MEFs in a collagen gel at 72h of culturing supplemented with the 120kDa fibronectin fragment. (AVI) Click here for additional data file. S12 Movie Representative FRET-distribution in a tissue from Fig 6A 24h: Fnf/f MEFs in a collagen gel at 24h of culturing supplemented with FRET-labeled fibronectin. (AVI) Click here for additional data file. S13 Movie Representative FRET-distribution in a tissue from Fig 6A 48h: Fnf/f MEFs in a collagen gel at 48h of culturing supplemented with FRET-labeled fibronectin. (AVI) Click here for additional data file. S14 Movie Representative FRET-distribution in a tissue from Fig 6A 72h: Fnf/f MEFs in a collagen gel at 72h of culturing supplemented with FRET-labeled fibronectin. (AVI) Click here for additional data file. S15 Movie Representative FRET-distribution in a tissue from Fig 6B 24h: Fn-/- MEFs in a collagen gel at 24h of culturing supplemented with FRET-labeled fibronectin. (AVI) Click here for additional data file. S16 Movie Representative FRET-distribution in a tissue from Fig 6B 48h: Fn-/- MEFs in a collagen gel at 48h of culturing supplemented with FRET-labeled fibronectin. (AVI) Click here for additional data file. S17 Movie Representative FRET-distribution in a tissue from Fig 6B 72h: Fn-/- MEFs in a collagen gel at 72h of culturing supplemented with FRET-labeled fibronectin. (AVI) Click here for additional data file. S18 Movie Representative FRET-distribution in a tissue from Fig 6B 96h: Fn-/- MEFs in a collagen gel at 96h of culturing supplemented with FRET-labeled fibronectin. (AVI) Click here for additional data file. We gratefully acknowledge Prof. Reinhard Fassler for providing us with the mouse embryonic fibroblasts (Fn-/- MEFs) and floxed equivalents (Fnf/f MEFs), Dr. Philip Kollmannsberger for writing the Matlab script, able to quantify fibronectin FRET ratios in 3D constructs; and Michael Borochin for assistance in fabricating the PDMS model system. ==== Refs References 1 Heisenberg CP , Bellaiche Y . Forces in tissue morphogenesis and patterning . Cell 2013 ;153 (5 ):948 –962 . 10.1016/j.cell.2013.05.008 23706734 2 Fernandez-Sanchez ME , Brunet T , Roper JC , Farge E . Mechanotransduction's impact on animal development, evolution, and tumorigenesis . Annu.Rev.Cell Dev.Biol . 2015 ;31373 –397 . 3 Bonnans C , Chou J , Werb Z . Remodelling the extracellular matrix in development and disease . Nat.Rev.Mol.Cell Biol . 2014 ;15 (12 ):786 –801 . 10.1038/nrm3904 25415508 4 Bellas E , Chen CS . Forms, forces, and stem cell fate . Curr.Opin.Cell Biol . 2014 ;3192 –97 . 5 Doras C , Taupier G , Barsella A , Mager L , Boeglin A , Bulou H , et al Polarization state studies in second harmonic generation signals to trace atherosclerosis lesions . Opt.Express 2011 ;19 (16 ):15062 –15068 . 10.1364/OE.19.015062 21934867 6 Weber KT , Pick R , Silver MA , Moe GW , Janicki JS , Zucker IH , et al Fibrillar collagen and remodeling of dilated canine left ventricle . Circulation 1990 ;82 (4 ):1387 –1401 . 2401072 7 Pujadas S , Bordes R , Bayes-Genis A . Ventricular non-compaction cardiomyopathy: CMR and pathology findings . Heart 2005 ;91 (5 ):582 – 15831637 8 Cho NS , Hwang JH , Lee YT , Chae SW . Tendinosis-like histologic and molecular changes of the Achilles tendon to repetitive stress: a pilot study in rats . Clin.Orthop.Relat Res . 2011 ;469 (11 ):3172 –3180 . 10.1007/s11999-011-2000-1 21800208 9 Foolen J , Deshpande VS , Kanters FM , Baaijens FP . The influence of matrix integrity on stress-fiber remodeling in 3D . Biomaterials 2012 ;33 (30 ):7508 –7518 . 10.1016/j.biomaterials.2012.06.103 22818650 10 Legant WR , Chen CS , Vogel V . Force-induced fibronectin assembly and matrix remodeling in a 3D microtissue model of tissue morphogenesis . Integr.Biol.(Camb.) 2012 ;4 (10 ):1164 –1174 .22961409 11 Dijkman PE , Driessen-Mol A , Frese L , Hoerstrup SP , Baaijens FP . Decellularized homologous tissue-engineered heart valves as off-the-shelf alternatives to xeno- and homografts . Biomaterials 2012 ;33 (18 ):4545 –4554 . 10.1016/j.biomaterials.2012.03.015 22465337 12 Mol A , van Lieshout MI , Dam-de Veen CG , Neuenschwander S , Hoerstrup SP , Baaijens FP , et al Fibrin as a cell carrier in cardiovascular tissue engineering applications . Biomaterials 2005 ;26 (16 ):3113 –3121 . 15603806 13 Sakar MS , Eyckmans J , Pieters R , Eberli D , Nelson BJ , Chen CS . Cellular forces and matrix assembly coordinate fibrous tissue repair . Nat.Commun . 2016 ;711036 – 14 Liu J , Hilderink J , Groothuis TA , Otto C , van Blitterswijk CA , de BJ . Monitoring nutrient transport in tissue–engineered grafts . J.Tissue Eng Regen.Med . 2013 ; 10.1002/term 15 Legant WR , Pathak A , Yang MT , Deshpande VS , McMeeking RM , Chen CS . Microfabricated tissue gauges to measure and manipulate forces from 3D microtissues . Proc.Natl.Acad.Sci.U.S.A 2009 ;106 (25 ):10097 –10102 . 10.1073/pnas.0900174106 19541627 16 Pankov R , Yamada KM . Fibronectin at a glance . J.Cell Sci . 2002 ;115 (Pt 20 ):3861 –3863 . 12244123 17 Kosmehl H , Berndt A , Katenkamp D . Molecular variants of fibronectin and laminin: structure, physiological occurrence and histopathological aspects . Virchows Arch . 1996 ;429 (6 ):311 –322 . 8982375 18 Mao Y , Schwarzbauer JE . Fibronectin fibrillogenesis, a cell-mediated matrix assembly process . Matrix Biol . 2005 ;24 (6 ):389 –399 . 16061370 19 Hynes RO . The extracellular matrix: not just pretty fibrils . Science 2009 ;326 (5957 ):1216 –1219 . 10.1126/science.1176009 19965464 20 Van Obberghen-Schilling E , Tucker RP , Saupe F , Gasser I , Cseh B , Orend G . Fibronectin and tenascin-C: accomplices in vascular morphogenesis during development and tumor growth . Int.J.Dev.Biol . 2011 ;55 (4-5 ):511 –525 . 10.1387/ijdb.103243eo 21769776 21 White ES , Baralle FE , Muro AF . New insights into form and function of fibronectin splice variants . J.Pathol . 2008 ;216 (1 ):1 –14 . 10.1002/path.2388 18680111 22 To WS , Midwood KS . Plasma and cellular fibronectin: distinct and independent functions during tissue repair. Fibrogenesis .Tissue Repair 2011 ;421 – 23 Zhang Z , Vuori K , Reed JC , Ruoslahti E . The alpha 5 beta 1 integrin supports survival of cells on fibronectin and up-regulates Bcl-2 expression . Proc.Natl.Acad.Sci.U.S.A 1995 ;92 (13 ):6161 –6165 . 7541142 24 Giancotti FG , Ruoslahti E . Integrin signaling . Science 1999 ;285 (5430 ):1028 –1032 . 10446041 25 Lutz R , Sakai T , Chiquet M . Pericellular fibronectin is required for RhoA-dependent responses to cyclic strain in fibroblasts . J.Cell Sci . 2010 ;123 (Pt 9 ):1511 –1521 . 10.1242/jcs.060905 20375066 26 Zhang Y , Lin Z , Foolen J , Schoen I , Santoro A , Zenobi-Wong M , et al Disentangling the multifactorial contributions of fibronectin, collagen and cyclic strain on MMP expression and extracellular matrix remodeling by fibroblasts . Matrix Biol . 2014 ;4062 –72 . 27 Knox P , Crooks S , Rimmer CS . Role of fibronectin in the migration of fibroblasts into plasma clots . J.Cell Biol . 1986 ;102 (6 ):2318 –2323 . 3711148 28 Nicosia RF , Bonanno E , Smith M . Fibronectin promotes the elongation of microvessels during angiogenesis in vitro . J.Cell Physiol 1993 ;154 (3 ):654 –661 . 7679680 29 Yousif NG . Fibronectin promotes migration and invasion of ovarian cancer cells through up-regulation of FAK-PI3K/Akt pathway . Cell Biol.Int . 2014 ;38 (1 ):85 –91 . 10.1002/cbin.10184 24115647 30 Baneyx G , Baugh L , Vogel V . Fibronectin extension and unfolding within cell matrix fibrils controlled by cytoskeletal tension . Proc.Natl.Acad.Sci.U.S.A 2002 ;99 (8 ):5139 –5143 . 11959962 31 Smith ML , Gourdon D , Little WC , Kubow KE , Eguiluz RA , Luna-Morris S , et al Force-induced unfolding of fibronectin in the extracellular matrix of living cells . PLOS.Biol . 2007 ;5 (10 ):e268 – 17914904 32 Little WC , Smith ML , Ebneter U , Vogel V . Assay to mechanically tune and optically probe fibrillar fibronectin conformations from fully relaxed to breakage . Matrix Biol . 2008 ;27 (5 ):451 –461 . 10.1016/j.matbio.2008.02.003 18417335 33 Kubow KE , Klotzsch E , Smith ML , Gourdon D , Little WC , Vogel V . Crosslinking of cell-derived 3D scaffolds up-regulates the stretching and unfolding of new extracellular matrix assembled by reseeded cells . Integr.Biol.(Camb.) 2009 ;1 (11-12 ):635 –648 .20027372 34 Vogel V . Mechanotransduction involving multimodular proteins: converting force into biochemical signals . Annu.Rev.Biophys.Biomol.Struct . 2006 ;35459 –488 . 35 Zhong C , Chrzanowska-Wodnicka M , Brown J , Shaub A , Belkin AM , Burridge K . Rho-mediated contractility exposes a cryptic site in fibronectin and induces fibronectin matrix assembly . J.Cell Biol . 1998 ;141 (2 ):539 –551 . 9548730 36 Chabria M , Hertig S , Smith ML , Vogel V . Stretching fibronectin fibres disrupts binding of bacterial adhesins by physically destroying an epitope . Nat.Commun . 2010 ;1135 – 37 Hertig S , Chabria M , Vogel V . Engineering mechanosensitive multivalent receptor-ligand interactions: why the nanolinker regions of bacterial adhesins matter . Nano.Lett . 2012 ;12 (10 ):5162 –5168 . 10.1021/nl302153h 22938173 38 Li B , Moshfegh C , Lin Z , Albuschies J , Vogel V . Mesenchymal stem cells exploit extracellular matrix as mechanotransducer . Sci.Rep . 2013 ;32425 – 39 Seong J , Tajik A , Sun J , Guan JL , Humphries MJ , Craig SE , et al Distinct biophysical mechanisms of focal adhesion kinase mechanoactivation by different extracellular matrix proteins . Proc.Natl.Acad.Sci.U.S.A 2013 ;110 (48 ):19372 –19377 . 10.1073/pnas.1307405110 24222685 40 Asaga H , Kikuchi S , Yoshizato K . Collagen gel contraction by fibroblasts requires cellular fibronectin but not plasma fibronectin . Exp.Cell Res . 1991 ;193 (1 ):167 –174 . 1995291 41 Hocking DC , Sottile J , Langenbach KJ . Stimulation of integrin-mediated cell contractility by fibronectin polymerization . J.Biol.Chem . 2000 ;275 (14 ):10673 –10682 . 10744764 42 Sottile J , Shi F , Rublyevska I , Chiang HY , Lust J , Chandler J . Fibronectin-dependent collagen I deposition modulates the cell response to fibronectin . Am.J.Physiol Cell Physiol 2007 ;293 (6 ):C1934 –C1946 . 17928541 43 Sevilla CA , Dalecki D , Hocking DC . Regional fibronectin and collagen fibril co-assembly directs cell proliferation and microtissue morphology . PLOS.One . 2013 ;8 (10 ):e77316 – 10.1371/journal.pone.0077316 24116223 44 Humphries JD , Byron A , Humphries MJ . Integrin ligands at a glance . J.Cell Sci . 2006 ;119 (Pt 19 ):3901 –3903 . 16988024 45 Reinke JM , Sorg H . Wound repair and regeneration . Eur.Surg.Res . 2012 ;49 (1 ):35 –43 . 10.1159/000339613 22797712 46 Kubow KE , Vukmirovic R , Zhe L , Klotzsch E , Smith ML , Gourdon D , et al Mechanical forces regulate the interactions of fibronectin and collagen I in extracellular matrix . Nat.Commun . 2015 ;68026 – 47 Fontana L , Chen Y , Prijatelj P , Sakai T , Fassler R , Sakai LY , et al Fibronectin is required for integrin alphavbeta6-mediated activation of latent TGF-beta complexes containing LTBP-1 . FASEB J . 2005 ;19 (13 ):1798 –1808 . 16260650 48 Baugh L , Vogel V . Structural changes of fibronectin adsorbed to model surfaces probed by fluorescence resonance energy transfer . J.Biomed.Mater.Res.A 2004 ;69 (3 ):525 –534 . 15127399 49 Rozario T , DeSimone DW . The extracellular matrix in development and morphogenesis: a dynamic view . Dev.Biol . 2010 ;341 (1 ):126 –140 . 10.1016/j.ydbio.2009.10.026 19854168 50 Grinnell F , Rocha LB , Iucu C , Rhee S , Jiang H . Nested collagen matrices: a new model to study migration of human fibroblast populations in three dimensions . Exp.Cell Res . 2006 ;312 (1 ):86 –94 . 16256985 51 Brown RA , Talas G , Porter RA , McGrouther DA , Eastwood M . Balanced mechanical forces and microtubule contribution to fibroblast contraction . J.Cell Physiol 1996 ;169 (3 ):439 –447 . 8952693 52 Murrell M , Oakes PW , Lenz M , Gardel ML . Forcing cells into shape: the mechanics of actomyosin contractility . Nat.Rev.Mol.Cell Biol . 2015 ;16 (8 ):486 –498 . 10.1038/nrm4012 26130009 53 Sniadecki NJ , Anguelouch A , Yang MT , Lamb CM , Liu Z , Kirschner SB , et al Magnetic microposts as an approach to apply forces to living cells . Proc.Natl.Acad.Sci.U.S.A 2007 ;104 (37 ):14553 –14558 . 17804810 54 Yang MT , Reich DH , Chen CS . Measurement and analysis of traction force dynamics in response to vasoactive agonists . Integr.Biol.(Camb.) 2011 ;3 (6 ):663 –674 .21445393 55 Lemmon CA , Chen CS , Romer LH . Cell traction forces direct fibronectin matrix assembly . Biophys.J . 2009 ;96 (2 ):729 –738 . 10.1016/j.bpj.2008.10.009 19167317 56 Ruoslahti E , Hayman EG , Engvall E , Cothran WC , Butler WT . Alignment of biologically active domains in the fibronectin molecule . J.Biol.Chem . 1981 ;256 (14 ):7277 –7281 . 6788765 57 da Rocha-Azevedo B , Grinnell F . Fibroblast morphogenesis on 3D collagen matrices: the balance between cell clustering and cell migration . Exp.Cell Res . 2013 ;319 (16 ):2440 –2446 . 10.1016/j.yexcr.2013.05.003 23664837 58 Schwarzbauer JE . Identification of the fibronectin sequences required for assembly of a fibrillar matrix . J.Cell Biol . 1991 ;113 (6 ):1463 –1473 . 2045422 59 Hocking DC , Chang CH . Fibronectin matrix polymerization regulates small airway epithelial cell migration . Am.J.Physiol Lung Cell Mol.Physiol 2003 ;285 (1 ):L169 –L179 . 12639845 60 Vaz R , Martins GG , Thorsteinsdottir S , Rodrigues G . Fibronectin promotes migration, alignment and fusion in an in vitro myoblast cell model . Cell Tissue Res . 2012 ;348 (3 ):569 –578 . 10.1007/s00441-012-1364-1 22427060 61 Brekhman V , Neufeld G . A novel asymmetric 3D in-vitro assay for the study of tumor cell invasion . BMC.Cancer 2009 ;9415 – 62 Wolf K , Te LM , Krause M , Alexander S , Te RJ , Willis AL , et al Physical limits of cell migration: control by ECM space and nuclear deformation and tuning by proteolysis and traction force . J.Cell Biol . 2013 ;201 (7 ):1069 –1084 . 10.1083/jcb.201210152 23798731 63 Sevilla CA , Dalecki D , Hocking DC . Extracellular matrix fibronectin stimulates the self-assembly of microtissues on native collagen gels . Tissue Eng Part A 2010 ;16 (12 ):3805 –3819 . 10.1089/ten.TEA.2010.0316 20673131 64 Rajagopalan P , Marganski WA , Brown XQ , Wong JY . Direct comparison of the spread area, contractility, and migration of balb/c 3T3 fibroblasts adhered to fibronectin- and RGD-modified substrata . Biophys.J . 2004 ;87 (4 ):2818 –2827 . 15454473 65 Marshall JF , Rutherford DC , McCartney AC , Mitjans F , Goodman SL , Hart IR . Alpha v beta 1 is a receptor for vitronectin and fibrinogen, and acts with alpha 5 beta 1 to mediate spreading on fibronectin . J.Cell Sci . 1995 ;108 (Pt 3 )1227 –1238 . 7542669 66 Roca-Cusachs P , Gauthier NC , Del RA , Sheetz MP . Clustering of alpha(5)beta(1) integrins determines adhesion strength whereas alpha(v)beta(3) and talin enable mechanotransduction . Proc.Natl.Acad.Sci.U.S.A 2009 ;106 (38 ):16245 –16250 . 10.1073/pnas.0902818106 19805288 67 Schiller HB , Hermann MR , Polleux J , Vignaud T , Zanivan S , Friedel CC , et al beta1- and alphav-class integrins cooperate to regulate myosin II during rigidity sensing of fibronectin-based microenvironments . Nat.Cell Biol . 2013 ;15 (6 ):625 –636 . 10.1038/ncb2747 23708002 68 Yu CH , Law JB , Suryana M , Low HY , Sheetz MP . Early integrin binding to Arg-Gly-Asp peptide activates actin polymerization and contractile movement that stimulates outward translocation . Proc.Natl.Acad.Sci.U.S.A 2011 ;108 (51 ):20585 –20590 . 10.1073/pnas.1109485108 22139375 69 Martino MM , Hubbell JA . The 12th-14th type III repeats of fibronectin function as a highly promiscuous growth factor-binding domain . FASEB J . 2010 ;24 (12 ):4711 –4721 . 10.1096/fj.09-151282 20671107 70 Zhu J , Clark RA . Fibronectin at select sites binds multiple growth factors and enhances their activity: expansion of the collaborative ECM-GF paradigm . J.Invest Dermatol . 2014 ;134 (4 ):895 –901 . 10.1038/jid.2013.484 24335899 71 Mitsi M , Hong Z , Costello CE , Nugent MA . Heparin-mediated conformational changes in fibronectin expose vascular endothelial growth factor binding sites . Biochemistry 2006 ;45 (34 ):10319 –10328 . 16922507 72 Obbink-Huizer C , Foolen J , Oomens CW , Borochin M , Chen CS , Bouten CV , et al Computational and experimental investigation of local stress fiber orientation in uniaxially and biaxially constrained microtissues . Biomech.Model.Mechanobiol . 2014 ;13 (5 ):1053 –1063 . 10.1007/s10237-014-0554-z 24458588 73 van Spreeuwel AC , Bax NA , Bastiaens AJ , Foolen J , Loerakker S , Borochin M , et al The influence of matrix (an)isotropy on cardiomyocyte contraction in engineered cardiac microtissues . Integr.Biol.(Camb.) 2014 ;6 (4 ):422 –429 .24549279 74 Obbink-Huizer C , Foolen J , Oomens CW , Borochin M , Chen CS , Bouten CV , et al Computational and experimental investigation of local stress fiber orientation in uniaxially and biaxially constrained microtissues . Biomech.Model.Mechanobiol . 2014 ;131053 –63 . 75 Sakai T , Johnson KJ , Murozono M , Sakai K , Magnuson MA , Wieloch T , et al Plasma fibronectin supports neuronal survival and reduces brain injury following transient focal cerebral ischemia but is not essential for skin-wound healing and hemostasis . Nat.Med . 2001 ;7 (3 ):324 –330 . 11231631 76 Mitsi M , Schulz MM , Gousopoulos E , Ochsenbein AM , Detmar M , Vogel V . Walking the Line: A Fibronectin Fiber-Guided Assay to Probe Early Steps of (Lymph)angiogenesis . PLOS ONE . 2015 ;10 (12 ):e0145210 – 10.1371/journal.pone.0145210 26689200 77 Kuo CW , Shiu JY , Chien FC , Tsai SM , Chueh DY , Chen P . Polymeric nanopillar arrays for cell traction force measurements . Electrophoresis 2010 ;31 (18 ):3152 –3158 . 10.1002/elps.201000212 20803755

PMC005xxxxxx/PMC5001708.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756470710.1371/journal.pone.0160851PONE-D-16-09402Research ArticleBiology and Life SciencesCell BiologyCellular TypesAnimal CellsNeuronsBiology and Life SciencesNeuroscienceCellular NeuroscienceNeuronsBiology and Life SciencesOrganismsAnimalsVertebratesAmniotesMammalsPrimatesMonkeysPhysical SciencesPhysicsClassical MechanicsKinematicsBiology and Life SciencesAnatomyMusculoskeletal SystemMedicine and Health SciencesAnatomyMusculoskeletal SystemBiology and Life SciencesAnatomyBrainMotor CortexMedicine and Health SciencesAnatomyBrainMotor CortexBiology and Life SciencesNeuroscienceCognitive ScienceCognitive PsychologyLearningBiology and Life SciencesPsychologyCognitive PsychologyLearningSocial SciencesPsychologyCognitive PsychologyLearningBiology and Life SciencesNeuroscienceLearning and MemoryLearningBiology and Life SciencesCell BiologyCellular TypesAnimal CellsNeuronsMotor NeuronsBiology and Life SciencesNeuroscienceCellular NeuroscienceNeuronsMotor NeuronsBiology and Life SciencesComputational BiologyComputational NeuroscienceSingle Neuron FunctionBiology and Life SciencesNeuroscienceComputational NeuroscienceSingle Neuron FunctionPremotor and Motor Cortices Encode Reward Premotor and Motor Cortices Encode RewardRamkumar Pavan 12*Dekleva Brian 34Cooler Sam 5Miller Lee 4Kording Konrad 121 Sensorimotor Performance Program, Rehabilitation Institute of Chicago, Illinois, 60611, United States of America2 Department of Physical Medicine and Rehabilitation, Northwestern University, Chicago, United States of America3 Department of Biomedical Engineering, Northwestern University, Chicago, United States of America4 Department of Physiology, Feinberg School of Medicine, Northwestern University, Chicago, United States of America5 Northwestern University Interdepartmental Neuroscience Program, Chicago, United States of AmericaLytton William W EditorSUNY Downstate MC, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Conceived and designed the experiments: PR BD LM KK. Performed the experiments: BD. Analyzed the data: PR SC. Contributed reagents/materials/analysis tools: PR SC. Wrote the paper: PR BD LM KK. * E-mail: pavan.ramkumar@northwestern.edu26 8 2016 2016 11 8 e016085112 3 2016 26 7 2016 © 2016 Ramkumar et al2016Ramkumar et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Rewards associated with actions are critical for motivation and learning about the consequences of one’s actions on the world. The motor cortices are involved in planning and executing movements, but it is unclear whether they encode reward over and above limb kinematics and dynamics. Here, we report a categorical reward signal in dorsal premotor (PMd) and primary motor (M1) neurons that corresponds to an increase in firing rates when a trial was not rewarded regardless of whether or not a reward was expected. We show that this signal is unrelated to error magnitude, reward prediction error, or other task confounds such as reward consumption, return reach plan, or kinematic differences across rewarded and unrewarded trials. The availability of reward information in motor cortex is crucial for theories of reward-based learning and motivational influences on actions. http://dx.doi.org/10.13039/100000065National Institute of Neurological Disorders and Stroke5R01NS074044-05Kording Konrad This work was supported by the National Institute of Neurological Disorders and Stroke (NINDS). Data AvailabilityAll data are freely available here: (https://figshare.com/articles/Ramkumar_et_al_2016_Premotor_and_motor_cortices_encode_reward/3573447).Data Availability All data are freely available here: (https://figshare.com/articles/Ramkumar_et_al_2016_Premotor_and_motor_cortices_encode_reward/3573447). ==== Body Introduction How the brain learns based on action outcomes is a central question in neuroscience. Theories of motor learning have usually focused on rapid, error-based learning mediated by the cerebellum, and slower, reward-based learning mediated by the basal ganglia (for a review, see [1]). Different combinations of reward and sensory feedback result in different learning rates. For instance, positive and negative rewards influence motor learning differently [2, 3]. When reward is combined with sensory feedback, it can accelerate motor learning [4]. Reward is thus a fundamental aspect of learning [5, 6, 7, 8]. Various reward signals have been characterized in the midbrain, prefrontal and limbic cortices [9, 10, 11, 12, 13]. Yet, we do not know how neurons in the motor system obtain the reward information that could be useful for planning subsequent movements. The dorsal premotor cortex (PMd) and the primary motor cortex (M1) are known to be involved in planning and executing movements. We know this because movement goals (e.g., direction of upcoming movement), kinematics (e.g., position, velocity and acceleration) and dynamics (e.g. forces, torques, and muscle activity) are reflected in the firing rates of motor cortical neurons [14, 15, 16, 17, 18, 19, 20]. If movement plans need to be modified based on previous actions, then information about their outcomes must reach motor cortices. In many real world settings, task outcomes typically manifest in the form of reward. Recently, Marsh et al. [21] have shown a robust modulation of M1 activity by reward expectation both during movement and observation of movement. To further investigate the nature of this potential reward signal, we trained monkeys to reach to targets based on noisy spatial cues and rewarded them for correct reaches. We induced different reward expectation on a trial-by-trial basis and quantified the representation of reward in PMd and M1. We observed that ~28% of PMd neurons and ~12% of M1 neurons significantly modulated their firing rates following trials that were not rewarded. The effect could not be explained simply by kinematic variables such as velocity or acceleration, reward consumption behavior, or upcoming movement plans, nor by task variables that may bias successful task performance, such as the noise in the target cue, the reward history, or the precision of the reach. This effect might constitute an important piece in the larger puzzle of how motor plans are modified based on reward. Results Our goal in this study was to investigate whether the motor system—in addition to planning and executing actions—also encodes responses to reward, which are key for learning about the environment and modifying motor plans. To this end, we trained two macaque monkeys to make center–out reaches to uncertain targets and rewarded them for successful reaches. The monkeys made reaching movements while grasping the handle of a planar manipulandum, their hand position represented by an on-screen cursor (Fig 1A). During this task, we recorded from two 96-channel microelectrode arrays (Fig 1B, Blackrock Microsystems), chronically implanted in the primary motor cortex (M1) and the dorsal premotor cortex (PMd). 10.1371/journal.pone.0160851.g001Fig 1 Reaching task to uncertain targets. (A) Monkeys made center–out reaches using a planar manipulandum that controlled an on-screen cursor. (B) We recorded from chronic microelectrode arrays implanted in dorsal premotor cortex (PMd) and primary motor cortex (M1). CS: central sulcus, AS: arcuate sulcus, PCD: precentral dimple (C) Monkeys reached towards a target that was cued using a set of 5 line segments, whose dispersion varied from trial to trial. In each trial, the true target location was sampled from a von Mises “prior” distribution centered at 90° (clockwise from the rightmost point on the annulus) with one of two concentration parameters specifying a broad or narrow prior. The line segments making up the cue were then sampled from a “likelihood” von Mises distribution centered on the target location, with one of two concentration parameters (κ = 5 or 50 for Monkey M and κ = 1 or 100 for Monkey T) specifying a broad or narrow spread. Adapted from Dekleva et al., 2016. A trial began when the monkey moved the cursor to the central target (Fig 1C). The true location of the target was not shown. Instead, a noisy cue was presented at a 7-cm radial distance from the center to indicate the approximate location of the outer target. The cue comprised a cluster of line segments generated from a distribution centered on the true target. Monkeys were trained to reach only after a combined visual/auditory go cue, which was delivered after a variable (0.8–1.0 s) delay period following cue appearance. At the end of the reach, the actual circular target (15° diameter) was displayed. If the monkey had successfully reached the target, an appetitive auditory cue announced the subsequent delivery of a juice reward. If the reach ended outside the target, an aversive auditory cue announced the failure of the reach and no reward was delivered. After the end of the trial, the monkey was cued to return to the center target in order to begin the next trial. The median inter-trial interval across animals and sessions was 2.79 ± 0.25 seconds. On any given trial, since the actual target was not shown, the monkey had to infer its location, potentially by combining the information in the noisy target cue with prior knowledge accumulated about the target location in previous trials (for details, see [22]). We assume that the monkeys calibrated their expected reward based on the cue uncertainty. To manipulate their reward expectation, we varied cue uncertainty from trial to trial. Specifically, we determined the dispersion of the line segment cluster on each trial by drawing the location of each line segment from either a narrow or a broad distribution (see Fig 1 for details). A narrower spread of line segments indicated the target location with lesser uncertainty than a broader spread. To verify that animals indeed change their reward expectation, we looked at the latency of movement onset after the go cue. We found that animals indeed started their reach later on average when they were more uncertain about the target location (35 ± 28 ms for narrow spreads; 111 ± 16 ms for broad spreads; mean ± SEM across animals and sessions). Thus, manipulating the dispersion on each trial is likely to have induced trial-by-trial changes in reward expectation. Neural coding of reward We asked if the firing rates of PMd and M1 neurons indicated whether a reward was obtained in the trial by comparing the peristimulus time histograms (PSTHs) aligned to the end of the trial timestamp (corresponding to the auditory cue that indicated whether a reward will be delivered) for rewarded and unrewarded trials. We matched the kinematics of the trials across the two conditions to control for trivial firing rate consequences of behavioral differences (see Fig A and Table A in S1 File for details about matching kinematics). We found that ~28% of PMd neurons and ~12% of M1 neurons modulated their firing rates in response to reward or lack thereof. Nearly 25% of all PMd neurons recorded increased their firing rates following unrewarded trials compared with rewarded trials (Fig 2A and 2B show example PMd and M1 neurons from one session). In comparison, only ~3% of PMd neurons increased their firing rates after rewarded trials. For M1, these numbers were ~8% and ~4%, respectively. 10.1371/journal.pone.0160851.g002Fig 2 Neural coding of reward. (A, B) Example neurons from PMd and M1 showing reward modulation that persisted after controlling for kinematic differences between rewarded and unrewarded trials (see Fig A in S1 File). Error bars show standard errors (SEMs) across trials. Vertical dashed line at zero indicates the time of reward. (C, D) Trial-averaged, population-averaged normalized firing rates (mean ± SEMs across neurons) for PMd and M1 from two monkeys. For each session, significant differences between rewarded and unrewarded peak PSTH amplitudes are indicated using an asterisk. PMd shows a clear increase after unrewarded trials compared to kinematically-matched rewarded trials. We then determined the extent to which this effect was visible across the entire population. To do so, we normalized single neuron PSTHs computed from kinematically-matched trials by setting the peak of each PSTH to 1, and computing separate PTSHs for M1 and PMd. There was a significant increase in population-wide normalized firing rate following unrewarded trials in both monkeys (Fig 2C and 2D). PMd had a significant effect in all 9 sessions (6 from Monkey M and 3 from Monkey T), whereas the effect in M1 was significant in only 2 out of 9 sessions. The firing rate effect of unrewarded trials in PMd and to a lesser extent, M1, is completely confounded by the fact that only successful trials were rewarded. Thus, increased firing rate for unrewarded trials could potentially be an intrinsic signal of success or failure, or might indicate some other correlate of the outcome of a goal-directed movement. To eliminate this confound, we ran a separate experiment on one monkey (Monkey M) in which we withheld reward in a subset of successful reaches. We found that firing rates increased even for these successful but unrewarded trials (Fig B in S1 File), suggesting that the increased activity following unsuccessful trials is related to lack of extrinsic reward, not an intrinsic measure of task outcome. Putative reward signal is not explained away by task confounds Several other variables could potentially confound this putative reward signal as well. As we did for kinematic differences (Fig A in S1 File), examining groups of rewarded and unrewarded trials that are matched for these confounding variables is a potential means of disambiguating the source of the effect. Yet, adequately matching all possible confounding variables is impossible because of the trade-off between precise matching of numerous potential confounds, and adequate remaining sample size. Instead, we controlled for potential confounds by using multiple linear regression models of trial-by-trial firing rates. Specifically, we modeled single neuron spike trains using Poisson generalized linear models (GLMs) (see e.g., [23, 24, 25], and Methods for details). To construct the GLM, we modeled neural spike counts during a 2-second epoch (–0.5 to 1.5 seconds, in 10-ms bins) around the reward onset. Spike counts were modeled as a function of the reward, which we represented as a binary variable (+1 for rewarded trials, –1 for unrewarded trials), aligned to the reward onset. In addition, we included the following confounding variables in the multiple regression. Kinematics. PMd and M1 neurons are known to encode kinematic variables during movement planning and execution [16, 17, 26]. Therefore, we included instantaneous velocity and acceleration time series, binned in 10-ms time bins. Uncertainty. Previous work has suggested that PMd can encode plans for more than one potential target [27] and we have recently shown that PMd encodes uncertainty about the reach target location [22]. Further, cue uncertainty influences the likelihood of a successful outcome, since monkeys are more successful in low-uncertainty trials. Although we did not find an effect of uncertainty on PSTHs aligned to reward time (Fig C in S1 File), we included a measure of trial-specific uncertainty as a confounding variable. Specifically, we used the dispersion of the target cue line segments, where dispersion is the largest circular distance between all possible pairs of line segments. Reward history. The outcome of the previous trial (and more generally, the history of reward) can influence the level of satiety, and thus the motivation and perceived value of a potential reward [12]. To control for this possibility, we included the previous trial’s outcome as a binary covariate (+1 for success, –1 for failure). Error. The reward-related signal might be useful for reinforcement learning (temporal difference learning) if it encoded some information about the discrepancy between the reach direction and the true target direction (presented visually at the end of the trial). To test whether PMd/M1 neurons encode error magnitude, we included the unsigned reaching error (reach precision) as a covariate. Return goal. Another potential confound is that the movement plan for the return reach to the center target may be modulated by recently obtained reward. Although a separate control analysis (Fig D in S1 File) suggested there were no systematic differences between return reach planning for rewarded and unrewarded trials, here we controlled for this possibility by including a covariate that specified the return reach direction; in particular, we used two covariates specifying the direction cosines (cosine and sine) of the return reach direction. Environmental covariates—uncertainty, error, and reward—that are potential causes of firing rate changes, invariably lead spikes in M1 and PMd. By contrast, movements are the consequence of motor cortical activity. Therefore, kinematic variables are likely to lag spikes. To model these latency differences for the different covariates, we used temporal basis functions (see Methods for details). Our model accurately captured the reward-related activity of many neurons. Comparing the data and cross-validated fit (see Methods) panels of Fig 3, we see that the trial-averaged data PSTHs and trial-averaged model-predicted firing rates are extremely similar. Across 70 PMd and 191 M1 neurons in a representative session (Monkey M, session 4), the model explained almost all the variance in the trial-averaged data (mean ± standard deviation of R2 = 0.96 ± 0.05 and 0.93 ± 0.08 for PSTHs averaged across successful and unsuccessful trials, respectively). These high R2s suggest that the model includes almost all potential sources of predictable variance. Therefore, if the reward covariate cannot be explained away by the confounding covariates, it is likely that the neurons represent reward. 10.1371/journal.pone.0160851.g003Fig 3 Generalized linear modeling of reward coding. Model predictions for two example neurons are shown. Left: PSTHs for rewarded (blue) and unrewarded (red) trial subsets are shown for the test set, along with corresponding single-trial rasters for both data and model predictions on the test set. Right: Component predictions corresponding to each covariate. To understand whether the reward covariate explains a significant fraction of the variance, we visualized the predictions of individual model covariates (reward, kinematics, uncertainty, reward history, error, and return goal), for rewarded and unrewarded trial subsets. Among all model covariates, only the reward covariate made different predictions for firing rates in rewarded and unrewarded trials (Fig 3, right panels); the predictions of other covariates were similar across these conditions. This preliminary analysis of example neurons seemed to suggest that reward was indeed the predominant driver of firing rate variance. To quantify the marginal effects of reward, kinematics and other confounding variables, we also built partial models leaving each covariate out and then comparing these respective partial models against the full model using the relative pseudo-R2 metric as a measure of effect size (see Methods for definitions and details). Briefly, we used two-fold cross-validation to quantify error estimates on pseudo-R2s. We used each half of the data as a training set to fit the full and partial models and computed the pseudo-R2s on the other half (the test set). We obtained 95% confidence intervals (CIs) on the cross-validated test set pseudo-R2 by bootstrapping on the test sets and used these to determine which neurons were significantly predicted by the covariate of interest at the 2σ and 5σ significance levels (see Methods). Three of the covariates accounted for a large fraction of the variance in PMd and M1 firing rates. As expected, a large number of neurons (41/70 in PMd, and 179/191 in M1; 2σ significance criterion; Fig 4A, left panel) were significantly modulated by reach kinematics (instantaneous velocity and acceleration) in a representative session (Monkey M, session 4). Further, many neurons (20/70 PMd and 67/191 M1; Fig 4A, middle panel) encoded the direction of the upcoming return reach. However, a large fraction of PMd (48/70), and M1 (75/191) neurons also encoded reward (Fig 4A, right panel). By comparison, a negligible number of neurons in either PMd or M1 encoded cue uncertainty (1), error magnitude (3) or reward history (15)—these were likely false positives that did not survive a multiple-comparison correction. These results were very similar across multiple sessions in both monkeys (Fig 4B). For Monkey T, the quality of the M1 array had degraded at the time of these experiments, and the spike-sorted neurons had extremely low firing rates, insufficient to fit reliable multivariate GLMs. Therefore we were only able to quantify the effects in PMd. 10.1371/journal.pone.0160851.g004Fig 4 Across-session summary of reward encoding. (A) The distribution of effect sizes across the population from one session in each monkey (Monkey M, Session 4 and Monkey T, Session 1) as measured by mean relative pseudo-R2s are shown for kinematics, return reach goal, and reward. A large fraction of neurons in both PMd and M1 remained statistically significantly modulated by reward after controlling for cue uncertainty, reward in the previous trial, error magnitude, instantaneous kinematics, and planning of the return reach. (B) A stacked bar shows the number of significant neurons at the 2-sigma level for each session. The representative session for which histograms of effect sizes are shown above is indicated with a black border. In the generalized linear models, we could not directly control for actual reward consumption, which involves mouth and neck movements, and might therefore affect firing rates in PMd and M1. However, we mainly observed increases in neural activity when reward was not received. This would only be possible if the monkey had made more vigorous mouth or neck movements in the absence of reward, than while actually consuming reward (e.g. by potentially sucking harder on the tube when no reward was delivered). We could not quantitatively control for this possibility because we did not measure kinematics or EMG signals from the face. However, we made doubly sure that the monkey correctly interpreted the auditory cue signaling lack of reward. To do this, we filmed one monkey during a separate session (Monkey M; see Fig E in S1 File) and observed that it simply sat still during trials when no reward was delivered, without making any oral contact with the reward delivery tube. Therefore, the robust reward signal cannot be explained by reward consumption. Taken together, our results suggest that a large fraction of PMd and M1 neurons encode the outcome of the task independently of uncertainty, error magnitude, kinematics, reward history, reward consumption, and the return reach plan. Discussion We asked if the premotor and motor cortices, implicated in planning and executing movements, might also represent the reward associated with those movements. We found a strong representation of reward in PMd firing rates, with a lesser effect in M1. The increase in firing rates was observed in response to the absence, and to a lesser extent, the occurrence of extrinsic reward, but not the intrinsic success or failure of the trial. We then asked if the reward signal encoded motivation or satiety (modeled by reward history), prediction error (modeled by cue uncertainty), or movement precision (modeled by error magnitude), but found no evidence for any such signals. We also confirmed that although kinematics and return movement planning could explain firing rate variance, neither of them could explain away the reward signal. Although the motor cortex has traditionally been thought of as a brain area that sends control signals to the spinal cord and muscles, recent studies (such as [21]) including ours, establish the important additional effect of reward on motor cortex activity. One weakness of this experiment is the lack of statistical power to ask if there were trial-by-trial reward dependent learning effects. The target location and the dispersion of the cue lines were drawn at random on each trial; hence, there was very little opportunity to transfer knowledge from one trial to the next. As trial-by-trial learning is generally relatively slow (and further slowed by high feedback uncertainty [28] we expected only a small trial-by-trial effect. Not surprisingly then, we did not find that the reward signal was directly tied to the behavioral performance of subsequent trials. If and how the reward signal does influence trial-by-trial learning should be investigated with further experiments. A second weakness of our design, which is typically common across many animal experiments, is that it does not rule out the possibility of covert motor rehearsal following error or lack of reward. Such rehearsal might activate premotor and motor cortices without resulting in overt behavior. Although we rule out any direction-specific effects of the reported reward outcome signal, it is impossible to definitively rule out the influence of covert rehearsal. This is an important constraint that future experiments must contend with. Reward is a central feedback mechanism that regulates motivation, valuation, and learning [8, 9]. Existing computational theories of this phenomenon, such as reinforcement learning and temporal difference learning [5, 7], have been successful in explaining the dopaminergic prediction error signal, but reward coding in the brain is far more heterogeneous and pervasive than just that [6, 10]. The large majority of brain areas implicated in reward processing, such as basal ganglia, ventral striatum, ventral tegmental area, and the orbitofrontal cortex, are predictive in nature, with predictions including reward probability, reward expectation, and expected time of future reward (for a review see [9]). Dopaminergic neurons in the ventral striatum also encode the mismatch between predicted and obtained rewards, combining reward prediction with reward feedback. Thus far, only the lateral prefrontal cortex has been shown to encode reward feedback without any predictive component. To our knowledge, the previous studies examining reward-related signaling in the premotor and motor cortices [11, 12, 21] reported a predictive code for reward magnitude and reward expectation but not for reward outcome or feedback. Previous studies have implicated the motor cortex in error-related signaling [29, 30]. Here, we show for the first time that single neurons in premotor and motor cortices encode reward-related feedback. Our finding adds another piece to the heterogeneity of reward representation in the midbrain and cortex, which will help extend future theories of reward-based learning. The latency of the reward signal in PMd and M1 is on the order of 400–600 milliseconds. This latency is much slower than the rapid (~100 ms) reward prediction error signal observed in dopaminergic neurons in the midbrain [31, 32]. Thus, the pathway to reward outcome representation in the motor cortex is likely to be mediated by the basal ganglia-thalamo-cortical loop. In particular, we know the striatum, which receives projections from reward-sensitive dopaminergic neurons, feeds back to the cortex through other basal ganglia structures and the thalamus [33]. The anterior cingulate cortex is also implicated in decision-making based on past actions and outcomes [34]. This is an alternate possibility for the origin of the signal that we observe in premotor and motor cortices. Thus, it is likely that the motor cortex, along with prefrontal cortex and other areas, reflects rather than generates the reward signal. At present, the function of this reward outcome signal in the motor cortices is unclear. A recent EEG-fMRI study [35] suggests that two distinct value systems shape reward-related learning in the brain. In particular, they found that an earlier system responding preferentially to negative outcomes engaged the arousal-related and motor-preparatory brain structures, which could be useful for switching actions if needed. Therefore, the reward signal in PMd and M1 could potentially induce the cortical connectivity changes required for correcting subsequent motor plans based on mistakes. Further investigations of our finding might thus potentially reveal the mechanisms by which the brain acquires new motor skills. Behavioral studies of motor control are at the advanced stage of describing trial-to-trial learning and generalization to novel contexts using sophisticated Bayesian decision theory and optimal control models [3, 36, 37, 38, 39]. Yet, we are only beginning to understand how different neural systems work together to achieve these behaviors. We have shown a robust reward signal in premotor and motor cortex that is not simply the result of movement kinematics or planning. Establishing a link between this reward signal and motor learning could potentially open up a new area of research within computational motor control. Methods Surgical procedures and animal welfare We surgically implanted two chronic 96-channel microelectrode arrays in dorsal premotor cortex (PMd) and primary motor cortex (M1) of two macaque monkeys. For further surgical details including array locations on the cortex, please see Dekleva et al., (2016). All surgical and experimental procedures were fully consistent with the guide for the care and use of laboratory animals and approved by the institutional animal care and use committee of Northwestern University. Monkeys received appropriate pre- and post-operative antibiotics and analgesics. The monkeys are pair-housed in standard size quad cages at the Feinberg School of Medicine. They receive a standard ration of Purina monkey chow biscuits twice a day. They are provided with hammocks and numerous puzzle and foraging toys in their cages. After completing a full series of recording experiments spanning six years, we placed monkey T under deep, surgical anesthesia with an intraveneous IV injection of Euthasol (25mg/kg), prior to transcardiac perfusion with saline followed by 4% formaldehyde solution. Experiments continue with monkey M. Single-neuron and population PSTHs We calculated peri-stimulus time histograms (PSTHs) of firing rates (spikes/s), in 25-ms windows aligned to the reward timestamp and averaged them across trials. Error bars were computed as standard errors of mean across trials. To test whether neurons were significantly modulated by reward, we compared mean firing rate in a [0, 1.5] second interval after the reward timestamp across rewarded and unrewarded trials using a one-sided t-test, with a significance level of α = 0.05, Bonferroni-corrected for the number of neurons recorded in a single session. To calculate population-averaged PSTHs, we took the mean trial-averaged PSTHs, normalized them to have a peak firing-rate of 1, and then averaged these across neurons. Error bars were computed as standard errors of mean across trials. Generalized Linear Modeling Temporal basis functions We used raised-cosine temporal basis functions to model the latencies between environmental events, firing rates, and kinematics. We used 4 basis functions with equal widths of 400 ms, and equispaced from each other with centers separated by 200 ms. We convolved each covariate time series with its respective basis set and then used these to predict firing rates. To prevent discontinuities between trial epochs, we zero-padded each trial with 500 ms (i.e., 50 time bins of 10 milliseconds, each), concatenated them, convolved the zero-padded time series, and then removed the zero-padding. Model fitting We fit models using the Matlab glmnet package which solves the convex maximum-likelihood optimization problem using coordinate descent (Hastie et al., 2009; Friedman et al., 2010). To prevent overfitting, we regularized model fits using elastic-net regularization [40]. We did not optimize the hyperparameters, but we found that a choice of λ = 0.1 (which determines the weight of the regularization term) and α = 0.1 (which weights the relative extent of L1 and L2 regularization) resulted in comparable training and test-set errors, and therefore did not inordinately over-fit or under-fit the data. We also cross-validated the model by fitting it to one random half of the trials and evaluating it on the other half. To evaluate model goodness of fit, we computed the pseudo-R2, which is related to the likelihood ratio. The idea of the pseudo-R2 metric is to map the likelihood ratio into a [0, 1] range, thus extending the idea of the linear R2 metric to non-Gaussian target variables. We used McFadden’s definition of pseudo-R2 [41, 23, 25]. For each neuron, we computed bootstrapped 95% confidence intervals of the pseudo-R2s. Model comparison To quantify whether individual covariates explain unique firing rate variance, we used partial models, leaving out the covariate of interest and comparing this partial model against the full model. To quantify this nested model comparison, we also used the relative pseudo-R2 metric. We obtained 95% confidence intervals on this metric using bootstrapping, for each cross-validation fold. We then treated the minimum of the lower bounds and the maximum of the upper bounds across cross-validation folds as confidence intervals. From these CIs, we approximated 2σ and 5σ significance levels, by calculating appropriate lower bounds for each significance level and comparing these lower bounds against zero. Supporting Information S1 File Supporting Information file containing 5 supporting figures (A–E) and one supporting table (A). (DOCX) Click here for additional data file. ==== Refs References 1 Shmuelof L , Krakauer JW . Are we ready for a natural history of motor learning? Neuron . 2011 ; 72 , 469 –476 . 10.1016/j.neuron.2011.10.017 22078506 2 Abe M , Schambra H , Wassermann EM , Luckenbaugh D , Schweighofer N , Cohen LG . Reward improves long-term retention of a motor memory through induction of offline memory gains . Current Biology . 2011 : 21 , 557 –562 . 10.1016/j.cub.2011.02.030 21419628 3 Galea JM , Mallia E , Rothwell J , Diedrichsen J . The dissociable effects of punishment and reward on motor learning . Nature Neuroscience . 2015 :18 , 597 –602 . 10.1038/nn.3956 25706473 4 Nikooyan AA , Ahmed AA . Reward feedback accelerates motor learning . Journal of Neurophysiology . 2015 : 113 , 633 –646 . 10.1152/jn.00032.2014 25355957 5 Hollerman JR , Schultz W . Dopamine neurons report an error in the temporal prediction of reward during learning . Nature Neuroscience . 1998 : 1 , 304 –309 . 10.1038/1124 10195164 6 Dayan P , Balleine BW . Reward, motivation, and reinforcement learning . Neuron . 2002 : 36 , 285 –298 . 10.1016/S0896-6273(02)00963-7 12383782 7 O'Doherty JP , Dayan P , Friston K , Critchley H , Dolan RJ . Temporal difference models and reward-related learning in the human brain . Neuron . 2003 : 38 , 329 –337 . 10.1016/S0896-6273(03)00169-7 12718865 8 Schultz W . Behavioral theories and the neurophysiology of reward . Annual Reviews in Psychology . 2006 : 57 , 87 –115 . 10.1146/annurev.psych.56.091103.070229 9 Schultz W . Multiple reward signals in the brain . Nature reviews neuroscience . 2000 : 1 , 199 –207 . 10.1038/35044563 11257908 10 Wallis JD , Kennerley SW . Heterogeneous reward signals in prefrontal cortex . Current opinion in neurobiology . 2010 : 20 , 191 –198 . 10.1016/j.conb.2010.02.009 20303739 11 Roesch MR , Olson CR . Impact of expected reward on neuronal activity in prefrontal cortex, frontal and supplementary eye fields and premotor cortex . Journal of Neurophysiology . 2003 : 90 , 1766 –1789 . 10.1152/jn.00019.2003 12801905 12 Roesch MR , Olson CR . Neuronal activity related to reward value and motivation in primate frontal cortex . Science . 2004 : 304 , 307 –310 . 15073380 13 Roesch MR , Olson CR . Neuronal activity in primate orbitofrontal cortex reflects the value of time . Journal of Neurophysiology . 2005 : 94 , 2457 –2471 . 15958600 14 Evarts EV . Relation of pyramidal tract activity to force exerted during voluntary movement . Journal of Neurophysiology . 1968 : 31 , 14 –27 . 4966614 15 Cheney PD , Fetz EE . Functional classes of primate corticomotorneuronal cells and their relation to active force . Journal of Neurophysiology . 1980 : 44 , 773 –791 . 6253605 16 Georgopoulos AP , Ashe J , Smyrnis N , Taira M . The motor cortex and the coding of force . Science . 1992 : 256 , 1692 –1695 . 10.1126/science.256.5064.1692 1609282 17 Ashe J , Georgopoulos AP . Movement parameters and neural activity in motor cortex and area 5 . Cerebral Cortex . 1994 : 4 , 590 –600 . 10.1093/cercor/4.6.590 7703686 18 Scott SH , Kalaska JF . Changes in motor cortex activity during reaching movements with similar hand paths but different arm postures . Journal of Neurophysiology 1995 : 73 , 2563 –2567 . 7666162 19 Holdefer RN , Miller LE . Primary motor cortical neurons encode functional muscle synergies . Experimental Brain Research . 2002 : 146 , 233 –243 . 12195525 20 Morrow MM , Jordan LR , Miller LE . Direct comparison of the task-dependent discharge of M1 in hand space and muscle space . Journal of Neurophysiology . 2007 : 97 , 1786 –1798 . 17122326 21 Marsh BT , Tarigoppula VSA , Chen C , Francis JT . Toward an Autonomous Brain Machine Interface: Integrating Sensorimotor Reward Modulation and Reinforcement Learning . Journal of Neuroscience . 2015 : 35 , 7374 –7387 . 10.1523/JNEUROSCI.1802-14.2015 25972167 22 Dekleva BM , Ramkumar P , Wanda PA , Kording KP , Miller LE . The neural representation of likelihood uncertainty in the motor system . eLife . 2016 : 5 , e14316 .27420609 23 Fernandes HL , Stevenson IH , Phillips AN , Segraves MA , Kording KP . Saliency and saccade encoding in the frontal eye field during natural scene search . Cerebral Cortex . 2014 : 24 , 3232 –3245 . 10.1093/cercor/bht179 23863686 24 Park IM , Meister ML , Huk AC , Pillow JW . Encoding and decoding in parietal cortex during sensorimotor decision-making . Nature Neuroscience . 2014 : 17 , 1395 10.1038/nn.3800 25174005 25 Ramkumar P , Lawlor PN , Wood DW , Glaser JI , Segraves MA , Körding KP . Task-relevant features predict gaze behavior but not neural activity in FEF during natural scene search . Journal of Neurophysiology . 2016 . EPub Ahead of Print. 26 Churchland MM , Santhanam G , Shenoy KV . Preparatory activity in premotor and motor cortex reflects the speed of the upcoming reach . Journal of Neurophysiology . 2006 : 96 , 3130 –3146 . 10.1152/jn.00307.2006 16855111 27 Cisek P , Kalaska JF . Neural correlates of reaching decisions in dorsal premotor cortex: specification of multiple direction choices and final selection of action . Neuron . 2005 : 45 , 801 –814 . 10.1016/j.neuron.2005.01.027 15748854 28 Wei K , Körding KP . Uncertainty of feedback and state estimation determines the speed of motor adaptation . Frontiers in computational neuroscience . 2010 : 4 10.3389/fncom.2010.00011 29 Diedrichsen J , Hashambhoy Y , Rane T , Shadmehr R . Neural correlates of reach errors . The Journal of Neuroscience . 2005 : 25 , 9919 –9931 .16251440 30 Even-Chen N, Stavisky SD, Kao JC, Ryu SI, Shenoy KV. Auto-deleting brain machine interface: Error detection using spiking neural activity in the motor cortex. In Engineering in Medicine and Biology Society (EMBC). 2015: 37th Annual International Conference of the IEEE (p71–75). IEEE. 10.1109/EMBC.2015.7318303. 31 Schultz W . Predictive reward signal of dopamine neurons . Journal of Neurophysiology . 1998 : 80 , 1 –27 . 9658025 32 Dommett E , Coizet V , Blaha CD , Martindale J , Lefebvre V , Walton N , et al How visual stimuli activate dopaminergic neurons at short latency . Science . 2005 : 307 , 1476 –1479 . 10.1126/science.1107026 15746431 33 Haber SN . Neuroanatomy of reward: A view from the ventral striatum In: Gottfried JA , editor. Neurobiology of Sensation and Reward . CRC Press/ Taylor & Francis , Boca Raton, FL ; 1991 . 34 Kennerley SW , Walton ME , Behrens TE , Buckley MJ , Rushworth MF . Optimal decision making and the anterior cingulate cortex . Nature Neuroscience . 2006 : 9 , 940 –947 . 16783368 35 Fouragnan E , Retzler C , Mullinger K , Philiastides MG . Two spatiotemporally distinct value systems shape reward-based learning in the human brain . Nature Communications . 2015 : 6 , 8107 10.1038/ncomms9107 26348160 36 Wolpert DM , Ghahramani Z . Computational motor control . Science . 2004 : 269 , 1880 –1882 . 37 Körding KP , Wolpert DM . Bayesian decision theory in sensorimotor control . Trends in cognitive sciences . 2006 : 10 , 319 –326 . 10.1016/j.tics.2006.05.003 16807063 38 Krakauer JW , Mazzoni P . Human sensorimotor learning: adaptation, skill, and beyond . Current opinion in neurobiology . 2011 : 21 , 636 –644 . 10.1016/j.conb.2011.06.012 21764294 39 Shadmehr R , Smith MA , Krakauer JW . Error correction, sensory prediction, and adaptation in motor control . Annual review of neuroscience . 2010 : 33 , 89 –108 . 10.1146/annurev-neuro-060909-153135 20367317 40 Qian J, Hastie T, Friedman J, Tibshirani R, Simon N. Glmnet for Matlab. 2013. Available: http://www.stanford.edu/~hastie/glmnet_matlab/. 41 McFadden D . Conditional logit analysis of qualitative choice behavior . Zarembka P , editor. Taylor & Francis ; 1974 .

PMC005xxxxxx/PMC5001709.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756371910.1371/journal.pone.0161876PONE-D-16-12556Research ArticlePhysical SciencesMaterials ScienceMaterials by AttributePigmentsOrganic PigmentsMelaninBiology and Life SciencesGeneticsGene ExpressionBiology and life sciencesMolecular biologyMolecular biology techniquesSequencing techniquesRNA sequencingResearch and analysis methodsMolecular biology techniquesSequencing techniquesRNA sequencingBiology and Life SciencesOrganismsAnimalsInvertebratesMolluscsBiology and Life SciencesOrganismsAnimalsInvertebratesMolluscsBivalvesBiology and Life SciencesBiochemistryNeurochemistryNeurochemicalsNitric OxideBiology and Life SciencesNeuroscienceNeurochemistryNeurochemicalsNitric OxideBiology and Life SciencesGeneticsGene ExpressionGene RegulationBiology and Life SciencesBiochemistryEnzymologyEnzyme ChemistryEnzyme RegulationIntegration of Next Generation Sequencing and EPR Analysis to Uncover Molecular Mechanism Underlying Shell Color Variation in Scallops Molecular Mechanism Underlying Shell Color Variation in ScallopsSun Xiujun 1Liu Zhihong 1Zhou Liqing 1Wu Biao 1Dong Yinghui 2Yang Aiguo 1*1 Yellow Sea Fisheries Research Institute, Chinese Academy of Fishery Sciences, Qingdao, China2 Zhejiang Provincial Top Key Discipline of Biological Engineering, Zhejiang Wanli University, Ningbo, 315100, ChinaSlominski Andrzej T EditorUniversity of Alabama at Birmingham, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: XS AY. Data curation: XS. Formal analysis: XS ZL. Funding acquisition: AY ZL. Investigation: XS ZL. Methodology: ZL BW. Project administration: AY. Resources: AY YD. Software: LZ YD. Supervision: LZ YD. Validation: XS. Visualization: LZ YD. Writing – original draft: XS. Writing – review & editing: ZL BW. * E-mail: yangag@ysfri.ac.cn26 8 2016 2016 11 8 e016187628 3 2016 12 8 2016 © 2016 Sun et al2016Sun et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.The Yesso scallop Patinopecten yessoensis displays polymorphism in shell colors, which is of great interest for the scallop industry. To identify genes involved in the shell coloration, in the present study, we investigate the transcriptome differences by Illumina digital gene expression (DGE) analysis in two extreme color phenotypes, Red and White. Illumina sequencing yields a total of 62,715,364 clean sequence reads, and more than 85% reads are mapped into our previously sequenced transcriptome. There are 25 significantly differentially expressed genes between Red and White scallops. EPR (Electron paramagnetic resonance) analysis has identified EPR spectra of pheomelanin and eumelanin in the red shells, but not in the white shells. Compared to the Red scallops, the White scallops have relatively higher mRNA expression in tyrosinase genes, but lower expression in other melanogensis-associated genes. Meantime, the relatively lower tyrosinase protein and decreased tyrosinase activity in White scallops are suggested to be associated with the lack of melanin in the white shells. Our findings highlight the functional roles of melanogensis-associated genes in the melanization process of scallop shells, and shed new lights on the transcriptional and post-transcriptional mechanisms in the regulation of tyrosinase activity during the process of melanin synthesis. The present results will assist our molecular understanding of melanin synthesis underlying shell color polymorphism in scallops, as well as other bivalves, and also help the color-based breeding in shellfish aquaculture. The Basic Scientific Research Fund of YSFRI2060302201516054Sun Xiujun Zhejiang Provincial Top Key Discipline of Biological EngineeringKF2015005Sun Xiujun This work was supported by the grants from the Basic Scientific Research Fund of YSFRI (No. 2060302201516054), Zhejiang Provincial Top Key Discipline of Biological Engineering (KF2015005). Data AvailabilityAll relevant data are within the paper and all sequencing reads were deposited in the National Center for Biotechnology Information (NCBI) center under the SRA accession of SRP065869.Data Availability All relevant data are within the paper and all sequencing reads were deposited in the National Center for Biotechnology Information (NCBI) center under the SRA accession of SRP065869. ==== Body Introduction Yesso scallop (Patinopecten yessoensis) is an economically important marine bivalve species in aquaculture and fishery in Asian countries, due to its large and edible adductor muscle [1]. The colors of the left and right valves are obviously distinct, typically having reddish-brown for the left and white for the right. In the cultured populations, we found the rare occurrence of albinism in the left valve, displaying two-side white phenotypes (Fig 1). Due to the higher muscle production of the two-side white individuals, they have been under mass selection since 2004 to produce the new color strain for aquaculture. In land animals, albinism is thought to be caused by the reduction or absence in the synthesis of melanin [2, 3]. However, to date, the cause of the albino phenotype in the scallop P. yessoensis is still unknown. The availability of albino specimens and normal color specimens is the ideal material to uncover the molecular mechanisms directly linked to shell color variation in this species. In our previous study, the deep sequencing of mantle transcriptome for the scallop has identified some dual functioning proteins involved in both biomineralization and melanogenesis, including proteins related to calcium metabolism, and putative tyrosinase-like protein [4]. Furthermore, we have extracted melanin extract from the pigmented shells in our previous study which allows us to propose the hypothesis that the occurrence of albino specimens is probably due to lack of melanin. 10.1371/journal.pone.0161876.g001Fig 1 The external view for left and right valves of Yesso scallop Patinopecten yessoensis. The predominant color of left valve is reddish-brown or white, while the color of right valve is white. Color polymorphism, where two or more distinct color morphs occur within species, is responsible for recognition, adaption, and camouflage, and has been documented in a variety of taxa in animal kingdom [5–7]. In both land and aquatic animals, color variation has been of great interests to the scientific and breeding communities [8–11]. Molluscs, a large phylum of invertebrate animals, usually show fabulous and diverse colors and patterns in their shells, which have been appreciated for hundreds of years by collectors and scientists [12–14]. Shell color in molluscs is most commonly due to the presence of melanins, indigoids, quinones and flavones [13,15]. Following the pioneering work of Comfort, carotenoids or polyenes were also identified as shell pigments in molluscs by Resonance Raman microspectrometry in more recent studies [16–18]. However, most of the biological pigments in molluscan shells have not been well studied due to difficulties in extraction and highly complex chemistry [15]. As one of those shell pigments, melanin is ubiquitous in nature and produced by a variety of organisms, including bacteria, fungi, plants, and animals [9, 19–21]. In land animals, the colors in feathers, fur and skin are largely determined by melanocytes, which are responsible for melanin production, and melanin is the main contributor to their pigmentation [22]. The quantity, quality, and distribution of melanin contents in animal tissues are usually correlated with the visual phenotypes of their pigmentation [7, 11, 23]. In contrast, the nature of melanin is far from being well understood in molluscs, such as bivalves, although melanin is almost certainly common in their shells [15]. Due to the difficulty in extracting the pigments, melanin in shells has been rarely identified and characterized by chemical methods. For molluscs, regulatory mechanism for melanin synthesis is poorly understood in bivalves, but it is well known in cephalopods. Cephalopod molluscs have the ink gland to produce melanin, mainly eumelanin, which underscores the complex interplay of melanogenic enzymes and regulatory factors [24, 25]. Studies on melanogenesis in the ink gland of cephalopods have revealed ink production are affected and modulated by neuronal NOS (nitric oxide synthase) and cyclic GMP (cGMP) in certain pathways of central nervous system, which control the activation of tyrosinase and increased melanin synthesis in the gland [25, 26]. Additionally, according to the comprehensive melanin studies in mammals and birds, two main types of melanin (pheomelanin and eumelanin) are both regulated by a common tyrosinase-dependent pathway with the same precursor, tyrosine [7, 9, 22]. All these studies shed lights on the functional role of tyrosinase on melanin production in the ink gland, feather, skin, and hair. However, in bivalve molluscs, mantle tissue is the key organ that secretes proteins responsible for calcification of shells and shell pigments, which seems to have a different melanin-producing system from that in mammals, birds, and cephalopods [15]. For bivalves, the biological pigments probably formed in the secretory cells in the mantle and incorporated into the shells along the growing edge [15, 27–29]. To date, with the rapid development of next generation sequencing (NGS) technology, the transcriptome characterization in the mantle tissue and has enabled us to identify dozens of genes potentially related to shell formation and pigmentation at the transcriptome level [4, 30–33]. The identified proteins and genes will greatly help the understanding of molecular progress for pigment synthesis in mantle and distribution in shell of molluscs. Despite molecular studies related to shell pigmentation are burgeoning, most of these studies do not have the pigments identified biochemically prior to molecular analyses, which may cause some problems in identification of potential genes related to color variation. Therefore, the integration of NGS technology and biochemical studies to explore the potential link between gene expression and color phenotypes is essential to understand shell color polymorphism in molluscs. To uncover molecular mechanism underlying shell color variation in scallops, in the present study we performed digital gene expression (DGE) analysis to identify differentially expressed genes in mantle of two extreme color phenotypes of P. yessoensis. Furthermore, EPR (Electron paramagnetic resonance) measurement was performed to identify melanin content in the red and white shells, and explore the relationship between gene expression and melanin content. With the integration of NGS technology, EPR, qPCR, ELISA and enzyme activity, the present findings highlight the functional roles of melanogensis-associated genes in the melanization process of scallop shells, which could help better understand the molecular mechanism underlying shell color variation in scallops, as well as other molluscs. Methods Ethics Statement The scallops used in the current study are marine-cultured animals, and all the experiments on scallops were conducted following the institutional and national guidelines. No endangered or protected species is involved in the present study, and no specific permission is required for the location of the culture experiment. Animal and tissue collection We randomly collected two-year-old live individuals of P. yessoensis, six from the two-side white strain (the fourth selected generation), termed as “White”, and six from normal color specimens, termed as “Red”. To obtain high quality of gene expression data, all of these scallops were cultured at the same condition, in sand-filtered sea water at 14 ± 2°C with the salinity of 32 ± 2 psu for two weeks, in a commercial hatchery in Jiaonan, China. Scallops were fed with Isochrysis galbana and two-thirds of the culture water was exchanged for each day. The mantle tissue of left valves dissected from each scallop was stored in RNAlater individually (Ambion). Total RNA was isolated from collected mantle tissue with Trizol Reagent (Invitrogen) following the manufacturer’s instruction. RNA purity and quality of the samples were checked using the NanoPhotometer™ spectrophotometer (Implen, CA, USA). RNA integrity was assessed using the RNA Nano 6000 Assay Kit. NGS library construction, sequencing and quality control RNA samples of three individuals from the same color type were pooled in equal amounts of 3 μg RNA to generate the mixed sample used for the construction of sequencing library. Two biological replicates were used for the Illumina sequencing. The sequencing libraries were generated using NEBNext Ultra™ RNA Library Prep Kit for Illumina (NEB, USA) following manufacturer’s recommendations. After mRNA being purified, the first strand cDNA was synthesized using random hexamer primer, and second strand cDNA synthesis was subsequently performed using DNA Polymerase I and RNase H. The detailed method for library construction was according to the previous study [4]. In brief, after the ligation of adaptors, the fragments were purified with AMPure XP system (Beckman Coulter, Beverly, USA). The adaptor-ligated cDNA was used for PCR amplification, and PCR products were purified and library quality was assessed. After cluster generation, the library sequencing was carried out in Illumina HiSeq 2000 platfrom using the 100 bp paired-end strategy. To ensure the data quality used for the downstream analyses, reads containing ambiguous ‘N’ nucleotides and with quality score of less than 5, was removed from further analysis. After the quality filtering, Q20, Q30 and GC-content of the clean data were calculated for estimating the quality. Aligning clean reads and differentially expressed genes analysis The clean data were then mapped to the previously assembled transcriptome due to lack of genome resources in the studied species [4], using RSEM [34]. The FPKM (reads per kilobase per million reads) was applied to measure the gene expression levels, which were estimated by read counts obtained from the mapping results for each gene. Differential gene expression analysis was performed to uncover the gene expression profiles of the mantle samples from White and Red scallops using the DESeq R package [35]. To minimize the false discovery rates, P values were adjusted by the Benjamini and Hochberg’s approach, with a significance of P < 0.05. To verify the data accuracy of NGS sequencing, quantitative real-time PCR (qPCR) was used to compare the relative mRNA expression of significantly expressed genes between the mantle of White and Red scallops. Total RNAs were extracted from the mantle of left valves using Trizol Reagent (Invitrogen). The quality and quantity of total RNA were estimated by ultraviolet spectroscopy using a NanoDrop 2000 spectrophotometer (Thermo Scientific, USA). The degradation and contamination of RNA samples was monitored on 1% agarose gels. Three biological replicates were used for qPCR analysis on Applied Biosystems 7500 system. The comparative Ct method (2-ΔΔCt method) was used to calculate the relative gene expression of the samples, which was normalized to β-actin mRNA level [36]. The expression data were subsequently subjected to one-way ANOVA or independent T-test in SPSS 17.0 to determine whether there was any significant difference with P < 0.05. Analysis of tyrosinase activities and tyrosinase protein level Melanin biosynthesis can be initiated from L-tyrosine, which is hydroxylated to L-dihydroxyphenylalanine (L-DOPA) and followed by oxidation of L-DOPA to dopaquinone, showing different absorbance spectra [37]. According to this, the enzyme activity of tyrosinase in the left mantle of Red and White scallops was measured the 475 nm absorbance using Animal tissue tyrosinase activity assay Kit (Haling, Shanghai). The measurements were performed at 25°C according to the manufacturer’s instruction. One unit of enzyme activity was defined as the amount of enzyme required to oxidize 1 μmol substrate within one minute under standard assay conditions. Protein concentration was determined by the Bradford’s method using Bradford Protein Assay Kit (Sangon Biotech). Furthermore, enzyme-linked immunosorbent assay (ELISA) was used for quantitative analysis of tyrosinase in the mantle tissues, which was measured by Fish Tyrosinase ELISA Kit (Jianglaibio, Shanghai) following the manufacturer’s instruction. All the absorbance measurements were performed using three biological replicates on a Multiskan GO Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Melanin extraction from shells and quantification analysis of melanin The red and white shells were carefully cleaned and dried completely in air. The dried shell samples were grinded into shell powder, which were used as the raw material for melanin extraction. The dry weight of 7 g shell powder was selected and treated by 6 M hydrochloric acid (HCl), and melanin was extracted by hot reflux method according to the previous study [38]. The crude products of melanin were collected by filtration on a Buchner funnel, and purified in a Soxhlet extractor with petroleum ether (60–80°C for 4 h). The melanin was washed and dried until constant weight, and then left in air to equilibrate with moisture for 24 h. The melanin extracts from Red and White shells were determined using Electron paramagnetic resonance (EPR) measurements [39]. To increase the sensitivity of EPR determination, measurements were carried out at 100K in liquid nitrogen in the presence of 50 mM zinc acetate using Bruker EPR A300 spectrometer [40]. The parameters of EPR measurement were set as follows: microwave power 20 mW, microwave frequency 9 GHz, modulation amplitude 2.0 G, scan range 50 G. The zinc acetate suspension of 10 mg eumelanin (melanin from Sepia officinalis, Sigma) and 10 mg synthetic pheomelanin were served as the EPR standards. The synthetic pheomelanin was prepared from L-cysteine (1.5 mmol) and L-Dopa (1.0 mmol) according to the previous study [41] with minor modification. Briefly, L-Dopa and L-cysteine were dissolved in 100 ml of 0.05 M sodium phosphate buffer (pH 6.8). The mixture was incubated at 37°C under oxygen current for 4 h in the presence of 20 mg of mushroom tyrosinase (>500 U/mg, Worthington Biochemical Corporation). The mixture was acidified to pH 3 with acetic acid and kept at 4°C for 1 h. The extract was collected by centrifugation, washed with acetic acid and acetone, and then dried in a vacuum freeze dryer for 8 h. The EPR measurement of melanin samples and standards were operated under identical experimental and apparatus conditions. Estimation of melanin content in terms of eumelanin to pheomelanin (a/b) is based on the comparison of major and additional peak height [42]. Results Overall statistics and reads There are a total of 63,311,680 raw reads generated by Illumina sequencing. The raw reads have been deposited in the NCBI SRA database (accession number: SRP065869). After quality control (remove reads containing adapters, ambiguous nucleotides and low quality reads), all sequencing data yields a total of 6.27 G clean bases. There are a total of 1.00 G, 1.92 G, 1.68 G, and 1.67 G clean bases generated in the sequencing libraries of Red_123, Red_456, White_123, and White_456, respectively (Table 1). The clean reads are further used for mapping with the reference mantle transcriptome of P. yessoensis [4]. The sequence alignment indicates that more than 85% reads are mapped into the reference transcriptome. In the four sequencing libraries, the similar values of Q20 percentage are detected among them, all around 98%. The same error rates are observed at 0.03% in the four libraries, and the percentage of GC content varies from 41% to 43%. 10.1371/journal.pone.0161876.t001Table 1 Summary statistics for sequencing and data quality of RNA-Seq. Sample Raw Reads Clean Reads Clean Bases Error (%) Q20 (%) Q30 (%) GC Content (%) C0_123 10,061,486 9,952,872 1.00G 0.03 98.1 93.56 42.17 C0_456 19,330,562 19,176,083 1.92G 0.03 98.22 93.88 41.16 T1_123 16,998,373 16,843,330 1.68G 0.03 98.12 93.56 41.43 T1_456 16,921,259 16,743,079 1.67G 0.03 98.16 93.66 41.93 Quality control of gene expression analysis Quality control of gene expression analysis is determined by saturation curves and correlation analysis. Saturation curves display the number of genes detected by uniquely mapped reads with different FPKM values as a function of the sequencing depth. For the blue line, it tracks the number of genes for transcripts with FPKM of 60–150, while the green line tracks the performance for transcripts with FPKM of 3–15. It is indicated that the fraction of transcript increases with additional sequencing data for the high expressed transcripts (FPKM > 3), even at low mapping rates (less than 20%). The expression data is significantly linear correlated between the two replicates of White scallops, as well as Red scallops, both having the squared Pearson correlation coefficient (R2) > 0.7 (R2 = 0.724 for Red_123 and Red_456; R2 = 0.792 for White_123 and White_456). Identification of differentially expressed genes and functional annotation Identification of different levels of expression for each library sample has pointed out that there are 25 significantly differentially expressed genes between Red and White scallops (Fig 2). Most of the detected genes are not annotated by the NR description, except for comp91333_c0, comp54314_c0, and comp94117_c0, which are identified as Poly [ADP-ribose] polymerase 14, hypothetical protein LOC100494154, and putative tyrosinase-like protein TYR-3, respectively. The unigene of comp94117_c0 is also annotated by PFAM domain description, which has the common central domain of tyrosinase. Additionally, the functional domain of four other unigenes are also detected by PFAM, including SRI (Set2 Rpb1 interacting) domain, Conotoxin, nucleic acid binding, and DNA polymerase (viral) C-terminal domain. Among the 25 screened genes, 17 genes (Red dots) are observed to have significantly higher expression level in White scallops than Red ones, while 8 genes (Green dots) exhibit significantly higher expression level in Red scallops than White ones (Fig 2). For comp94117_c0, the expression of TYR-3 gene in White is almost 4-fold higher than that in Red. 10.1371/journal.pone.0161876.g002Fig 2 The volcano plots for differentially expressed genes between Red and White. Green points represent transcripts with significantly lower expression level, while red circles represent transcripts with significantly higher expression level (P < 0.05). Gene expression analysis by qPCR The quantitative RT-PCR (qPCR) method was used to verify the RNA-Seq results, which resulted in the significantly higher expression of TYR-3 in the mantle tissue of White scallops than that of Red scallops (p < 0.01; Fig 3), in accordance with the RNA-Seq results. Furthermore, tissue expression pattern of TYR-3 was determined for both of Red and White scallops, showing the similar pattern of tissue-specific gene expression in the sampled tissues (data not show). The significantly highest expression was detected in the organ of shell formation—mantle tissue, whereas the lowest gene expression was observed in the tissues of foot and adductor muscle (p < 0.01). 10.1371/journal.pone.0161876.g003Fig 3 Quantitative real-time PCR (qPCR) analysis for melanogensis-associated genes expressed in the mantle of White and Red scallops. In addition, other melanogensis-associated genes or pathways, including NOS (nitric oxide synthase), MITF (microphthalmia-associated transcription factor), TYR-1, CREB (cAMP responsive element binding protein), CBP (CREB binding protein), cGMP, sGC (soluble guanylate cyclase) and NMDA (Nmethyl-D-aspartate), were also selected to measure their relative mRNA expression in mantle tissues of Red and White scallops. Independent T-test indicated that there were six in eight genes (NOS, MITF, TYR-1, CREB, cGMP and sGC) detected significantly expressed between Red and White scallops (Fig 3). All of these genes were significantly higher expressed in the mantle of Red scallops than White scallops, except for TYR-1. Tyrosinase activities and tyrosinase protein level The quantification of tyrosinase activities for the mantle of Red and White scallops was summarized in Fig 4. As a result, the value of tyrosinase activities (U/μg protein) in Red scallops (0.0853 ± 0.0065 U/μg protein) was approximately two-fold higher than that in White scallops (0.0427 ± 0.0020 U/μg protein; P < 0.01). 10.1371/journal.pone.0161876.g004Fig 4 The quantification of tyrosinase activities and protein level (ELISA test) in the mantle of Red and White scallops. The concentration of tyrosinase protein in the mantle of Red and White scallops were determined by comparing the O.D. value of the samples to the standard curve made by the ELISA kit. As shown in Fig 4, the significantly higher protein concentration was detected in the Red scallops (36.6856 ± 1.4847 ng/L) than the White scallops (22.8507 ± 3.6868 ng/L; P < 0.05). Melanin determination by EPR measurement EPR measurement of the melanin standards and samples were summarized in Fig 5. The EPR signal of melanin was identified in the melanin extract of red shells, but it was too weak to be detected in the extract of white shells. As shown in Fig 5, the melanin samples from red shells have the similar EPR spectra with the synthetic pheomelanin (cysteinyldopa melanin). By the comparison of major and additional peak height, the ratio of eumelanin to pheomelanin (a/b) was calculated to be 2.1 in the melanin samples of red shells. However, the ratio of a/b could not be determined in the melanin samples of white shells due to their weak EPR signal. 10.1371/journal.pone.0161876.g005Fig 5 EPR spectra of cysteinyldopa melanin (10 mg/ml) and melanin extract from shell power of Yesso scallop (left shell; 7 g). All the samples were suspended in 50 mM zinc acetate and examined at 100K using the same apparatus parameters. a, height of major peak; b, height of the additional peak. Discussion In order to address one of the crucial issues of animal pigmentation that how gene expression correlates with melanin content, we have characterized the gene expression by DGE analysis in mantle tissue and determined the melanin content by EPR measurement in shells. The present results revealed a dozen of differentially expressed transcripts, which are potentially related to the color variation in this species. Compared to the Red scallops, the White scallops have the relatively lower mRNA expression of melanogensis-associated genes (except for TYR-1 and TYR-3), the weak EPR signal, lower concentration of tyrosinase protein, and decreased tyrosinase activity, which are suggested to be associated with the loss of melanin in the white shells. Our present findings shed new light on the role of transcriptional and post-transcriptional mechanisms in the regulation of tyrosinase activity and melanin synthesis in scallops. For vertebrates, melanin is widely distributed in animal feathers, fur and skin, which is the main contributor to their pigmentation, especially in mammals and birds [7, 22, 37]. The two main types of melanin, red/yellow pheomelanin and brown/black eumelanin, are both derived from a common tyrosinase-dependent pathway [22]. Since tyrosinase is the rate-limiting enzyme of the melanin synthesis, the absence or dysfunction of tyrosinase and tyrosinase-related protein may result in the inability of melanocytes to make pigments, which causes preferential dilution of pheomelanin and shows albino phenotypes in mammals [9, 43]. The ratio of eumelanin and pheomelanin can be expressed quantitatively by EPR signal through the a/b [42, 44, 45]. In this study, the ratio of pheomelanin in total melanin (~30%) measured by EPR is similar with that in mammal skin and hair, which have a constant level of about 26% [23, 39, 44, 46]. EPR spectrum of the melanin extract from the Red scallops indicates that they contain a mixed type of pheomelanin and eumelanin in their pigmented shells. In contrast, melanin extract from the White scallops shows no detectable EPR signal, which indicates the occurrence of white shells due to lack of melanin. Melanin synthesis is under complex regulatory control regulated by enzymes, structural proteins, transcriptional regulators, transporters, receptors, and growth factors in animals and microorganisms [37, 47]. Among them, tyrosinase is thought to be the key enzyme in the production of melanin. For bivalve molluscs, tyrosinase has been identified expressed in mantle tissue of the pearl oyster Pinctada fucata [48, 49], pacific oyster Crassostrea gigas [29], as well as the scallop P. yessoensis in our previous study [4]. Furthermore, localization of tyrosinase in pigmented regions has been reported in some bivalve species, such as pearl oyster P. fucata [28, 50], and hard clam Mercenaria mercenaria [51]. In this study, we have not only detected the EPR spectra of melanin from the shell extract of Red scallops, but also found moderately tyrosinase activity and protein level in their mantle tissue. These data provide new evidences for melanogenesis in scallops and consistent with the speculation that tyrosinase is probably secreted from the mantle tissue and transported to the calcified shell layer, where they contribute to melanin synthesis in bivalves [29, 50]. As reported, the process of melanin synthesis requires the control of multiple genes as well as their regulatory and structural products [37]. For tyrosinase, the posttranscriptional processing of pro-tyrosinase mRNA generates several alternatively spliced products, but only one transcript was able to confer tyrosinase enzyme activity [37, 52, 53]. In this study, we found the tyrosinase mRNA (TYR-3 and TYR-1) in both of Red and White scallops, but it showed the negative changes in relative abundances compared with tyrosinase protein and enzyme activity. The present data imply that the stimulation of tyrosinase activity may be caused by other alternatively spliced products, not TYR-3 and TYR-1 in the scallops. Therefore, it is speculated that the regulation of melanin synthesis though tyrosinase appears to be controlled by posttranslational mechanisms [37]. Additionally, for molluscs, hemocyanins and tyrosinase share the similar enzyme activation and catalysis, although their physiological functions differ [54]. In this study, it is therefore suggested that the increase of tyrosinase activity in Red scallops is caused by a hypothetical supplementation with the tyrosinase-like activity of hemocyanin, but not directly induced by the tyrosinase mRNA expression. Besides tyrosinase, other important genes were also reported to be involved in the transcriptional regulation of the complex process of melanin synthesis, such as NOS, MITF, CREB, CBP, cGMP, sGC, NMDA and etc. [37]. For instance in skin pigmentation, ultraviolet B radiation acts through the stimulation of NOS and subsequent release of nitric oxide (NO), and then activation of sGC with subsequent accumulation of cGMP and protein kinase G to stimulate melanogenesis in the melanocytes [37, 55]. In addition, MITF plays a fundamental role in providing positive regulation of transcription of tyrosinase, TyrP1 and TyrP2 genes through interaction with M and E boxes, which may act as a self-regulating switchboard for diverse pathways regulating the activity of the melanogenic apparatus [56, 57]. The mechanism of cAMP regulation of melanogenesis involves the activation of protein kinase A, which involves phosphorylation of CREB and CBP [37]. NMDA receptors can interact with cGMP and MITF, resulting in activation of tyrosinase and increased melanin synthesis [26, 58]. In the present study, NOS, MITF, CREB, cGMP and sGC are significantly higher expressed in Red than White scallops, which are suggested to be responsible for the pigmentation in the red shells. These findings indicate that melanin synthesis in scallops may involve the transcriptional mechanism activated by different signaling systems [37]. Furthermore, the other differently expressed genes also provide some hints for shell color variation in the scallops (Table 2). For example, comp83659_c2 has the domain similarity with SRI (Set2 Rpb1 interacting), which mediates RNA polymerase II interaction and couples histone H3 K36 methylation with transcript elongation [59]. However, the provisional annotation sheds no light on the functional basis of the color variation or shell formation. In contrast, the unigene of comp86031_c0 is annotated related to conotoxins, which are small neurotoxic peptides with disulphide connectivity that modulate the activity of ion channels [60]. Also, the unigene of comp91852_c0 is suggested to be associated with the function of nucleic acid binding and zinc ion binding. These annotations suggest that these genes may be involved in biomineralization processes such as calcium channels and ion exchange during shell formation. All together, these findings are consistent with the current concept proposing that melanogenesis-controlling factors may be not arranged in simple sequences, but instead they usually interact in a complex and multidimensional network, which is determined by the genetic-biochemical-physical context [61]. 10.1371/journal.pone.0161876.t002Table 2 Identification of differentially expressed genes between the red and white of Patinopecten yessoensis. Gene_id T1 reads C0 reads log2FoldChange P-adj Discription comp78910_c0 0.00 95.59 - 8.26E-12 - comp82342_c0 5.41 161.43 -4.90 6.80E-10 - comp86321_c0 11.53 198.74 -4.11 1.05E-08 - comp91673_c0 198.70 28.04 2.83 8.39E-05 - comp96387_c1 159.27 22.84 2.80 2.46E-04 - comp54314_c0 148.00 21.37 2.79 2.97E-04 PREDICTED: hypothetical protein LOC100494154 comp51792_c0 5.71 79.09 -3.79 3.72E-04 - comp94117_c0 9563.20 2405.38 1.99 2.19E-03 Putative tyrosinase-like protein tyr-3 comp83659_c2 15.36 150.75 -3.29 4.88E-03 SRI (Set2 Rpb1 interacting) domain comp86031_c0 384.71 83.10 2.21 6.84E-03 Conotoxin//FHIPEP family comp91852_c0 2531.43 639.72 1.98 9.59E-03 nucleic acid binding//zinc ion binding comp77935_c0 0.44 55.19 -6.97 9.81E-03 - comp78009_c0 67.36 8.00 3.07 9.81E-03 - comp99770_c0 51.43 6.52 2.98 9.81E-03 - comp98835_c0 198.60 42.93 2.21 1.05E-02 DNA polymerase (viral) C-terminal domain comp81979_c0 1.83 44.42 -4.60 1.07E-02 - comp96125_c0 97.58 7.33 3.73 1.18E-02 - comp81217_c0 3.93 48.36 -3.62 1.24E-02 - comp81333_c0 73.60 10.94 2.75 2.43E-02 - comp93541_c0 4117.08 1267.84 1.70 2.49E-02 - comp69580_c0 26.68 0.33 6.35 2.86E-02 - comp53469_c0 170.76 20.77 3.04 3.10E-02 - comp91333_c0 122.04 26.07 2.23 3.10E-02 Poly [ADP-ribose] polymerase 14 comp93725_c0 976.47 221.56 2.14 3.10E-02 - comp87825_c0 88.71 20.06 2.14 3.71E-02 - For bivalve molluscs, melanin pigmentation is almost certainly common in their shells, which probably forms in the secretory cells of the mantle edge and show as part of the general color mosaics in shells and is associated with photo-receptor mechanisms [15, 27, 28]. In cephalopod molluscs, they usually have a vivid, dynamic coloration due to the neurally controlled chromatophores in the body skin, which comprise elastic sacculus containing pigments to make visual signals for concealment and communication [62]. In Octopus vulgaris, there are some black chromatophores in their skin, which are suggested to be composed of melanin (eumelanin) [63]. Although the rapid changes of coloration of the cephalopod skin is neurogenic, the process of the melanin synthesis and its regulation, which undergoes a longer time regime, remains still dependent on the hormonal regulation [37, 62]. It is therefore suggested that the process of melanin synthesis in the shells of bivalves and the skin of cephalopods may have some similarities, since both of them are derived from tyrosinase [15, 63]. Conclusion In conclusion, the Illumina DGE analysis has revealed a dozen of differently expressed genes related to the color variation in P. yessoensis. EPR analysis indicates that EPR spectra of pheomelanin and eumelanin exist in the red shells, but no EPR signal of melanin was identified in white shells. With the integration of gene expression, EPR, ELISA and tyrosinase activity, our findings highlight the functional roles of melanogensis-associated genes in the melanization process of scallop shells, and shed new lights on the transcriptional and post-transcriptional mechanisms in the regulation of tyrosinase activity during the process of melanin synthesis. The present results will assist our molecular understanding of melanin synthesis underlying shell color polymorphism in scallops, as well as other bivalves, and also help the color-based breeding in shellfish aquaculture. We would like to thank the two anonymous reviewers for their extensive and constructive comments on the manuscript. This work was supported by the grants from the Basic Scientific Research Fund of YSFRI (No. 2060302201516054), Zhejiang Provincial Top Key Discipline of Biological Engineering (KF2015005). ==== Refs References 1 Xu K , Li Q , Kong L , Yu R (2008 ) A first-generation genetic map of the Japanese scallop patinopecten yessoensis-based AFLP and microsatellite markers . Aquaculture Research 40 : 35 –43 . 2 Kaelin CB , Barsh GS (2013 ) Genetics of pigmentation in dogs and cats . Annual Review of Animal Biosciences 1 : 125 –156 . 10.1146/annurev-animal-031412-103659 25387014 3 King RA , Summers CG (1988 ) Albinism . Dermatologic Clinics 6 : 217 –228 . 3288382 4 Sun X , Yang A , Wu B , Zhou L , Liu Z (2015 ) Characterization of the mantle transcriptome of Yesso scallop (Patinopecten yessoensis): identification of genes potentially involved in biomineralization and pigmentation . PLoS ONE 10 : e0122967 10.1371/journal.pone.0122967 25856556 5 Jiggins CD , Naisbit RE , Coe RL , Mallet J (2001 ) Reproductive isolation caused by colour pattern mimicry . Nature 411 : 302 –305 . 11357131 6 Mckinnon JS , Pierotti MER (2010 ) Colour polymorphism and correlated characters: genetic mechanisms and evolution . Molecular Ecology 19 : 5101 –5125 . 10.1111/j.1365-294X.2010.04846.x 21040047 7 Roulin A , Ducrest AL (2013 ) Genetics of colouration in birds . In Seminars in cell & developmental biology . Vol 24 : 594 –608 . 8 Colihueque N (2010 ) Genetics of salmonid skin pigmentation: clues and prospects for improving the external appearance of farmed salmonids . Reviews in Fish Biology and Fisheries 20 : 71 –86 . 9 Kaelin CB , Barsh GS (2013 ) Genetics of pigmentation in dogs and cats . Annual Review of Animal Biosciences 1 : 125 –156 . 10.1146/annurev-animal-031412-103659 25387014 10 Takahashi T , Sota T , Hori M (2013 ) Genetic basis of male colour dimorphism in a lake tanganyika cichlid fish . Molecular Ecology 22 : 3049 –3060 . 10.1111/mec.12120 23176589 11 Maan ME , Sefc KM (2013 ) Colour variation in cichlid fish: developmental mechanisms, selective pressures and evolutionary consequences . Seminars in Cell and Developmental Biology 24 : 516 –528 . 10.1016/j.semcdb.2013.05.003 23665150 12 Comfort A (1950 ) Biochemistry of molluscan shell pigments . Journal of Molluscan Studies 28 : 79 –85 . 13 Zheng H , Zhang T , Sun Z , Liu W , Liu H (2013 ) Inheritance of shell colours in the noble scallop chlamys nobilis (bivalve: pectinidae) . Aquaculture Research 44 : 1229 –1235 . 14 Ge J , Li Q , Yu H , Kong L (2015 ) Mendelian inheritance of golden shell color in the Pacific oyster Crassostrea gigas . Aquaculture 441 : 21 –24 . 15 Comfort A (1951 ) The pigmentation of molluscan shells . Biological Reviews 26 : 285 –301 . 16 Hedegaard C , Bardeau JF , Chateigner D (2006 ) Molluscan shell pigments: an in situ resonance Raman study . Journal of Molluscan Studies 72 : 157 –162 . 17 Bergamonti L , Bersani D , Mantovan S , Lottici PP (2013 ) Micro-Raman investigation of pigments and carbonate phases in corals and molluscan shells . European Journal of Mineralogy 25 : 845 –853 . 18 Stemmer K , Nehrke G (2014 ) The distribution of polyenes in the shell of Arctica islandica from North Atlantic localities: a confocal Raman microscopy study . Journal of Molluscan Studies 80 : 365 –370 . 19 Prota G (1995 ) The chemistry of melanins and melanogenesis In Fortschritte der Chemie organischer Naturstoffe/Progress in the Chemistry of Organic Natural Products , 93 –148 (Springer Vienna ). 20 Selvakumar P , Rajasekar S , Periasamy K , Raaman N (2008 ) Isolation and characterization of melanin pigment from Pleurotus cystidiosus (telomorph of Antromycopsis macrocarpa) . World Journal of Microbiology and Biotechnology 24 : 2125 –2131 . 21 Goncalves RCR , Lisboa HCF , Pombeiro-Sponchiado SR (2012 ) Characterization of melanin pigment produced by Aspergillus nidulans . World Journal of Microbiology and Biotechnology 28 : 1467 –1474 . 10.1007/s11274-011-0948-3 22805928 22 Lin JY , Fisher DE (2007 ) Melanocyte biology and skin pigmentation . Nature 445 : 843 –850 . 17314970 23 Bino SD , Ito S , Sok J , Nakanishi Y , Bastien P , Wakamatsu K , et al (2015 ) Chemical analysis of constitutive pigmentation of human epidermis reveals constant eumelanin to pheomelanin ratio . Pigment Cell & Melanoma Research 28 : 707 –717 .26285058 24 Prota G , Ortonne JP , Voulot C , Khatchadourian C , Nardi G , Palumbo A (1981 ) Occurrence and properties of tyrosinase in the ejected ink of cephalopods . Comparative Biochemistry & Physiology Part B Comparative Biochemistry 68 : 415 –419 . 25 Palumbo A (2003 ) Melanogenesis in the ink gland of Sepia officinalis . Pigment Cell Research 16 : 517 –522 . 12950731 26 Palumbo A , Cosmo AD , Poli A , Cristo CD , D'Ischia M (1999 ) A Calcium/Calmodulin-Dependent Nitric Oxide Synthase, NMDAR2/3 Receptor Subunits, and Glutamate in the CNS of the Cuttlefish Sepia officinalis . Journal of Neurochemistry 73 : 1254 –1263 . 10461919 27 Marin F , Luquet G , Marie B , Medakovic D (2007 ) Molluscan shell proteins: primary structure, origin, and evolution . Current Topics in Developmental Biology 80 : 209 –276 . 28 Nagai K , Yano M , Morimoto K , Miyamoto H (2007 ) Tyrosinase localization in mollusc shells . Comparative Biochemistry and Physiology Part B 146 : 207 –214 . 29 Zhang G , Fang X , Guo X , Li L , Luo R , Xu F , et al (2012 ) The oyster genome reveals stress adaptation and complexity of shell formation . Nature 490 : 49 –54 . 10.1038/nature11413 22992520 30 Joubert C , Piquemal D , Marie B , Manchon L , Pierrat F , Zanella-Cléon I , et al (2010 ) Transcriptome and proteome analysis of Pinctada margaritifera calcifying mantle and shell: focus on biomineralization . BMC genomics 11 : 613 10.1186/1471-2164-11-613 21040589 31 Bai Z , Zheng H , Lin J , Wang G , Li J (2013 ) Comparative analysis of the transcriptome in tissues secreting purple and white nacre in the pearl mussel Hyriopsis cumingii . PLoS ONE 8 : e53617 10.1371/journal.pone.0053617 23341956 32 Yue X , Nie Q , Xiao G , Liu B (2015 ) Transcriptome analysis of shell color-related genes in the clam Meretrix meretrix . Marine Biotechnology 17 : 364 –374 . 10.1007/s10126-015-9625-0 25680512 33 Liu H , Zheng H , Zhang H , Deng L , Liu W , Wang S , et al (2015 ) A de novo transcriptome of the noble scallop, Chlamys nobilis, focusing on mining transcripts for carotenoid-based coloration . BMC genomics , 16 : 44 10.1186/s12864-015-1241-x 25651863 34 Li B , Dewey CN (2011 ) RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome . BMC Bioinformatics 12 : 93 –99 .21481263 35 Anders S , Huber W (2010 ) Differential expression analysis for sequence count data . Genome Biology 11 : 79 –82 . 36 Livak KJ , Schmittgen TD (2001 ) Analysis of relative gene expression data using real-time quantitative PCR and the 2-ΔΔCt method . Methods 25 : 402 –408 . 11846609 37 Slominski A , Tobin DJ , Shibahara S , Wortsman J (2004 ) Melanin pigmentation in mammalian skin and its hormonal regulation . Physiological Reviews 84 : 1155 –1228 . 15383650 38 Hao S , Hou X , Wei L , Li J , Li Z , Wang X (2015 ) Extraction and identification of the pigment in the adductor muscle scar of Pacific oyster Crassostrea gigas . PLoS ONE 10 : e0142439 10.1371/journal.pone.0142439 26555720 39 Enochs WS , Nilges MJ , Swartz HM (1993 ) A standardized test for the identification and characterization of melanins using electron paramagnetic resonance (EPR) spectroscopy . Pigment Cell Research 6 : 91 –99 . 8391699 40 d'Ischia M , Wakamatsu K , Napolitano A , Briganti S , Garcia-Borron JC , Kovacs D , et al (2013 ) Melanins and melanogenesis: methods, standards, protocols . Pigment Cell & Melanoma Research 26 : 616 –633 .23710556 41 Ito S (1989 ) Optimization of conditions for preparing synthetic pheomelanin . Pigment Cell Research 2 : 53 –56 . 2497445 42 Vsevolodov EB , Ito S , Wakamatsu K , Kuchina II , Latypov IF (1991 ) Comparative analysis of hair melanins by chemical and electron spin resonance methods . Pigment Cell Research 4 : 30 –34 . 1656423 43 Oetting WS , Fryer JP , Shriram S , King RA (2003 ) Oculocutaneous albinism type 1: the last 100 years . Pigment Cell Research 16 : 307 –311 . 12753405 44 Wolnicka-Glubisz A , Pecio A , Podkowa D , Kolodziejczyk LM , Plonka PM (2012 ) Pheomelanin in the skin of Hymenochirus boettgeri (Amphibia: Anura: Pipidae) . Experimental dermatology 21 : 537 –540 . 10.1111/j.1600-0625.2012.01511.x 22716250 45 Slominski A , Wortsman J , Plonka PM , Schallreuter KU , Paus R , Tobin DJ (2005 ) Hair follicle pigmentation . Journal of Investigative Dermatology 124 : 13 –21 . 15654948 46 Ito S , Wakamatsu K (2011 ) Human hair melanins: what we have learned and have not learned from mouse coat color pigmentation . Pigment Cell & Melanoma Research 24 : 63 –74 .20726950 47 Plonka PM , Grabacka M (2006 ) Melanin synthesis in microorganisms-biotechnological and medical aspects . Acta Biochimica Polonica 53 : 429 –443 . 16951740 48 Miyashita T , Takagi R (2011 ) Tyrosinase causes the blue shade of an abnormal pearl . Journal of Molluscan Studies 77 : 312 –314 . 49 Lemer S , Saulnier D , Gueguen Y , Planes S (2015 ) Identification of genes associated with shell color in the black-lipped pearl oyster, Pinctada margaritifera . BMC Genomics 16 : 568 10.1186/s12864-015-1776-x 26231360 50 Zhang C , Xie L , Huang J , Chen L , Zhang R (2006 ) A novel putative tyrosinase involved in periostracum formation from the pearl oyster (Pinctada fucata) . Biochemical and Biophysical Research Communications 342 : 632 –639 . 16488396 51 Gordon J , Carriker MR (1980 ) Sclerotized protein in the shell matrix of a bivalve mollusc . Marine Biology 57 : 251 –260 . 52 Müller G , Ruppert S , Schmid E , Schütz G (1988 ) Functional analysis of alternatively spliced tyrosinase gene transcripts . Embo Journal 7 : 2723 –30 . 3141148 53 Ruppert S , Muller G , Kwon B , Schutz G (1988 ) Multiple transcripts of the mouse tyrosinase gene are generated by alternative splicing . Embo Journal 7 : 2715 –2722 . 2846281 54 Decker H , Schweikardt T , Nillius D , Salzbrunn U , Jaenicke E , Tuczek F (2007 ) Similar enzyme activation and catalysis in hemocyanins and tyrosinases . Gene 398 : 183 –191 . 17566671 55 Roméro-Graillet C , Aberdam E , Biagoli N , Massabni W , Ortonne JP , Ballotti R (1996 ) Ultraviolet b radiation acts through the nitric oxide and cGMP signal transduction pathway to stimulate melanogenesis in human melanocytes . Journal of Biological Chemistry 271 : 28052 –6 . 8910416 56 Shibahara S , Takeda K , Yasumoto K , Udono T , Watanabe K , Saito H , Takahashi K (2001 ) Microphthalmia-associated transcription factor (MITF): multiplicity in structure, function, and regulation . The Journal of Investigative Dermatology Symposium 6 : 99 –104 . 57 Widlund HR , Fisher DE (2003 ) Microphthalamia-associated transcription factor: a critical regulator of pigment cell development and survival . Oncogene 22 : 3035 –3041 . 12789278 58 Hoogduijn MJ , Hitchcock IS , Smit NPM , Gillbro JM , Schallreuter KU , Genever PG (2006 ) Glutamate receptors on human melanocytes regulate the expression of mitf . Pigment Cell Research 19 : 58 –67 . 16420247 59 Kizer KO , Phatnani HP , Yoichiro S , Hana H , Greenleaf AL , Strahl BD (2005 ) A novel domain in set2 mediates RNA polymerase ii interaction and couples histone h3 k36 methylation with transcript elongation . Molecular & Cellular Biology 25 : 3305 –3316 .15798214 60 Terlau H , Olivera BM (2004 ) Conus venoms: a rich source of novel ion channel-targeted peptides . Physiological Reviews 84 : 41 –68 . 14715910 61 Slominski A , Plonka PM , Pisarchik A , Smart JL , Tolle V , Wortsman J , et al (2005 ) Preservation of eumelanin hair pigmentation in proopiomelanocortin-deficient mice on a nonagouti (a/a) genetic background . Endocrinology 146 : 1245 –1253 . 15564334 62 Messenger JB (2001 ) Cephalopod chromatophores: neurobiology and natural history . Biological Reviews 76 : 473 –528 . 11762491 63 Fox DL , Crane SC (1942 ) Concerning the pigments of the two-spotted octopus and the opalescent squid . Biological Bulletin 82 : 284 –291 .

PMC005xxxxxx/PMC5001710.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756486410.1371/journal.pone.0161729PONE-D-16-18181Research ArticleBiology and life sciencesGeneticsGene expressionGene regulationSmall interfering RNAsBiology and life sciencesBiochemistryNucleic acidsRNANon-coding RNASmall interfering RNAsBiology and Life SciencesOrganismsPlantsSolanumPotatoBiology and Life SciencesAgricultureCrop ScienceCropsVegetablesPotatoBiology and Life SciencesOrganismsPlantsVegetablesPotatoBiology and life sciencesCell biologyChromosome biologyChromatinChromatin modificationDNA methylationBiology and life sciencesGeneticsEpigeneticsChromatinChromatin modificationDNA methylationBiology and life sciencesGeneticsGene expressionChromatinChromatin modificationDNA methylationBiology and life sciencesGeneticsDNADNA modificationDNA methylationBiology and life sciencesBiochemistryNucleic acidsDNADNA modificationDNA methylationBiology and life sciencesGeneticsEpigeneticsDNA modificationDNA methylationBiology and life sciencesGeneticsGene expressionDNA modificationDNA methylationResearch and analysis methodsExtraction techniquesRNA extractionBiology and Life SciencesPlant SciencePlant AnatomyTubersResearch and analysis methodsChemical synthesisBiosynthetic techniquesNucleic acid synthesisRNA synthesisBiology and life sciencesBiochemistryNucleic acidsRNARNA synthesisBiology and Life SciencesMolecular BiologyMolecular Biology TechniquesArtificial Gene Amplification and ExtensionPolymerase Chain ReactionResearch and Analysis MethodsMolecular Biology TechniquesArtificial Gene Amplification and ExtensionPolymerase Chain ReactionResearch and analysis methodsBioassays and physiological analysisBiochemical analysisColorimetric assaysMTS assayResearch and analysis methodsBioassays and physiological analysisBiochemical analysisEnzyme assaysMTS assayEpigenome Editing of Potato by Grafting Using Transgenic Tobacco as siRNA Donor Epigenome Editing of Potato by Grafting Using Transgenic Tobacco as siRNA DonorKasai Atsushi 1*Bai Songling 2Hojo Hatsune 1Harada Takeo 11 Department of Agriculture and Life Science, Hirosaki University, Hirosaki, Japan2 College of Agriculture & Biotechnology, Zhejiang University, Hangzhou, Zhejiang Province, ChinaPradhan Sriharsa EditorInc, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Conceived and designed the experiments: AK SB. Performed the experiments: AK HH. Analyzed the data: AK HH. Contributed reagents/materials/analysis tools: AK SB HH. Wrote the paper: AK TH. * E-mail: kasaia@hirosaki-u.ac.jp26 8 2016 2016 11 8 e01617295 5 2016 10 8 2016 © 2016 Kasai et al2016Kasai et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.In plants, it is possible to induce heritable transcriptional gene silencing (TGS) via RNA-directed DNA methylation (RdDM) using artificially synthesized small RNA (siRNA) homologous to the 5'-flanking region of the target gene. As the siRNA signal with a specific RNA determinant moves through plasmodesmata and sieve elements, we attempted to induce TGS of a transgene and an endogenous gene of potato (Solanum tuberosum) rootstock by grafting using siRNA produced in a tobacco (Nicotiana benthamiana) scion. Our results provide evidence that this system can induce TGS of target genes in tubers formed on potato rootstock. The TGS is maintained in the progeny tubers lacking the transported siRNAs. Our findings reveal that epigenome editing using mobile RNA has the potential to allow breeding of artificial sport cultivars in vegetative propagation crops. http://dx.doi.org/10.13039/501100001691Japan Society for the Promotion of Science22380003Harada Takeo http://dx.doi.org/10.13039/501100001691Japan Society for the Promotion of Science24880007Kasai Atsushi http://dx.doi.org/10.13039/501100001691Japan Society for the Promotion of Science15K18624Kasai Atsushi Health Labor Science Research GrantH24-shokuhin-ippan-007Harada Takeo Health Labor Science Research GrantH27-shokuhin-ippan-007Harada Takeo This work was supported by Japan Society for the Promotion of Science Grants-in-Aid for Scientific Research (KAKENHI) Grant Numbers 22380003, 24880007 and 15K18624 and by Ministry of Health, Labor and Welfare Grants (H24-shokuhin-ippan-007 and H27-shokuhin-ippan-007); Japan Society for the Promotion of Science Grants-in-Aid for Scientific Research (KAKENHI), http://www.jsps.go.jp/english/e-grants/index.html, Ministry of Health, Labor and Welfare Grants, http://mhlw-grants.niph.go.jp/. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction The term epigenetics generally refers to the study of heritable information that is independent of DNA sequence variation. Epigenetic changes occur at the DNA level through DNA methylation of cytosine residues or at the level of histones that influence the accessibility of the DNA to transcription activation [1]. A number of studies have revealed that siRNA initiates gene silencing through the RNA-directed DNA methylation (RdDM) pathway [2, 3]. Indeed, a small RNA (siRNA) homologous to the promoter sequence of a gene can cause transcriptional gene silencing (TGS) of that gene [4]. Epigenetic change is an important mechanism for regulation of genomic integrity in higher eukaryotes. Since this information functions as a transcriptional memory associated with cell fate decisions, developmental switches, or stress responses, the memory needs to be erased during reproduction, as is the case for the vernalization memory in wheat [5, 6]. By contrast, some reports have indicated that transgenerational epigenetic memory can be stably transmitted through meiosis, resulting in inheritance by the subsequent generation [7–10]. Therefore, artificial creation of novel epi-alleles is a promising approach for improvement of crops [11, 12]. Nonetheless, a major concern regarding epimutants is that they may be partially heritable through gametogenesis. On the other hand, a number of commonly cultivated plants, such as potato, sugar cane, and fruit trees, are usually propagated by vegetative means rather than by seeds. Therefore, TGS would be a particularly attractive approach for improvement of such clonally propagated species. Several studies have demonstrated that transgene-derived siRNA moves across the graft union between a scion and a rootstock [13–15]. We have also achieved stable TGS status of a transgene in Nicotiana benthamiana by graft-transported siRNAs derived from the hairpin RNA of the promoter sequence [16]. Here, we attempted to induce TGS in potato tuber through hetero-grafting using tobacco as a siRNA donor. Results and Discussion First, the promoter of the 35S:GFP locus in transgenic potato was targeted by siRNA derived from the TGS starter (Co35SpIR, Fig 1A) to induce trans-locational TGS [16]. In order to enhance the potential siRNA level in the phloem, the transcription of Co35SpIR was controlled by a companion cell-specific promoter (Commelina Yellow Mottle Virus promoter) [17]. A post-transcriptional gene silencing (PTGS) starter (CoGFPIR) for the 35S:GFP transgene was also constructed for comparison with the TGS. The Co35SpIR transgenic N. benthamiana with confirmed siRNA production ability was then grafted onto the 35S:GFP potato rootstock (Fig 1B-1). The hetero-grafted plants were grown aseptically in a tissue culture vessel. Immediately after grafting, the shoot(s) from lateral bud(s) in the potato stock sprouted and grew actively (Fig 1B-2) because of transient loss of apical dominance until secure establishment of the graft union. Meanwhile, the graft union showed strict conglutination, resulting in active growth of the scion. After approximately one month, the grown lateral shoot of the potato rootstock was cut off to increase the sink power for phloem transport of siRNA into the rootstock organ (Fig 1B-3; blue arrow). Unexpectedly, in approximately 20% of cases, an adventitious shoot was regenerated spontaneously from the compact callus formed on the cut surface of the lateral shoot (Fig 1B-4 and 1C and S1 Fig). The resulting plants were then placed on micro-tuber (MT) induction medium [18]. 10.1371/journal.pone.0161729.g001Fig 1 TGS of 35S:GFP transgene in micro tubers formed on hetero-grafted plants. (A) Schematic diagram of the silencing starters. The TGS starter (35SpIR) contained an inverted repeat (IR) of the 35S promoter sequence (-32 ~ -343). The PTGS starter contained an IR of part of GFP (+113 ~ +782). CoYMV is a companion cell-specific promoter. (B) Schematic drawing of grafting and micro-tuber (MT) formation on the grafted plant. The MT (arrow) was generated spontaneously from a callus formed on the cut surface of a lateral potato shoot. Red arrowhead shows the graft union. (C) MT formed on a grafted plant between the N. benthamiana scion and the potato stock. (D) GFP expression and silencing of MTs. The MTs were cut with a razor and the surface was observed under UV light. (E) Levels of methylation of the target region. Asterisks show statistically significant (**; P<0.01 Student's t test) differences relative to 35SGFP. Means and SD of 2 to 4 biological replicates are shown. One and two MTs (Fig 1B and 1C) were formed on two grafted plants, respectively. Although one tuber from grafted plant A exhibited almost the same GFP expression as the 35S:GFP control, two MTs (tubers B1 and B2) from grafted plant B showed decreased expression (Fig 1D). Furthermore, they showed clearly high methylation of the target region (Fig 1E and S2A Fig). In the case of PTGS, a decrease of the GFP transcript was observed, but the methylation level was unchanged (Fig 1E). These methylation levels in the target region were maintained in the shoots that had sprouted from the MTs (S2A Fig). Furthermore, the 2nd progeny MTs on the sprouted shoots derived from tuber B1 exhibited low GFP expression and a low GFP mRNA level, whereas the PTGS case no longer showed the silencing status (S2B and S2C Fig). A second round of experiments showed that one out of four MTs of hetero-grafted plants exhibited clear TGS, which was also supported by the high methylation level of only the target region (S3 Fig). It has been shown that siRNA transported in phloem from the scion induces strong TGS in lateral roots including root apical meristem cells in the rootstock [16], because in angiosperms lateral roots are initiated from only two pericycle cells [19, 20], in which TGS is induced by cell to cell movement of siRNA from sieve elements to phloem cells. Indeed, potato lateral roots harvested from the hetero-grafted plants manifested clear TGS as a result of low levels of GFP expression and hypermethylation of the target region (S4 Fig). Furthermore, the shoot regenerated from the root and the MT on the shoot also exhibited the TGS state (S5 Fig). These results indicated that derivation of a shoot from a TGS lateral root by tissue culture is an infallible method of obtaining a TGS potato plant, despite the long period (approximately 6 months) needed to obtain a sufficiently strong shoot. Next, the TGS induction system using graft-transported siRNA was applied to an endogenous gene in potato. Granule-bound starch synthase I (GBSSI) catalyses the synthesis of amylose in amyloplasts [21]. When the GBSSI was silenced by PTGS or TGS using siRNA, low-amylose and high-amylopectin potato starch was produced in the transgenic potato [22, 23]. The resulting waxy-type potato starch has a smooth pulpy texture, a high-quality taste with high viscosity, and is less retrograde in comparison to regular potato starch. In spite of its high commercial value, this type of improved potato is not cultivated because of the general public distrust of gene-modified (GM) crops [24]. Therefore, using graft-induced epigenetic change, we tried to create a TGS potato for GBSSI lacking the transgene as the TGS starter. The cultivar ‘Waseshiro’ has two allelic promoter sequences of GBSSI, which possesses 3 SNPs and four indel sequences between them (S6 Fig). A transgenic potato harboring 35S:GBpIR (line t33, Fig 2A and S7 Fig), which was constructed from the allele b sequence (S6 Fig) in accordance with a previous study [21], exhibited methylation of the genome target region (Fig 2B and 2C) and a reduced level of GBSSI mRNA (Fig 2D). The tubers harvested from pot cultivation had lower-amylose starch, with approximately 41% less amylose than the wild-type potato (Fig 2E), indicating that the siRNA starter we had designed was fully functional. The CoGBpIR N. benthamiana producing siRNA for the 5' flanking region of GBSSI (allele a, Fig 2A and S6 Fig) was generated, and then the siRNA production ability was confirmed by northern blotting (S8 Fig). These were grafted as a scion onto potato cultivar ‘Waseshiro’ shoots and grown in tissue culture bottles. Fifteen hetero-grafted plants produced MTs on the adventitious shoots (Fig 1B). One of them, the t9 N. benthamiana line (S8 Fig), was used as the scion, formed two MTs exhibiting a high methylation level in the target region, and the methylation status was maintained in the sprouted shoots of the 2nd progeny tubers (Fig 3). The methylation level was almost the same as that of the 35S:GBpIR transgenic potato exhibiting a low amylose-starch ratio compared with the wild-type ‘Waseshiro’. Therefore, we succeeded in generating a TGS potato line for the GBSSI gene. As a matter of course, these TGS potato plants had no fragments corresponding the TGS starter construct in the genome (S9 Fig). The shoot regenerated from the root also exhibited the appropriate GBSSI methylation levels (S10 Fig). We plan to analyze the relationship between methylation levels and TGS using these GBSSI TGS lines. 10.1371/journal.pone.0161729.g002Fig 2 TGS induction of the endogenous gene StGBSSI in 35SGBpIR transgenic potato. (A) Schematic diagram of the GBSSI 5'-flanking region in potato ‘Waseshiro’. Black boxes are untranslated regions, and green boxes are exons. Orange bar shows the target region. Green and red bars indicate regions used for methylation level and transcription level analyses, respectively. The construct of 35SGBpIR and CoGBpIR is drawn schematically. (B) Methylation level of the target region in the potato leaves of in vitro sub-culturing shoots. WT: wild type, t33: 35SGBpIR line 33. Asterisks show statistically significant (**; P<0.01 Student's t test) differences relative to 35SGFP. Means and SD of 3 biological replicates are shown. (C) DNA methylation status in the target (orange) and its flanking regions (black) of GBSSI. The leaves were subjected to bisulphite sequencing. The percentages of methylation at individual cytosines are shown. Upper and lower are WT and t33 results, respectively. 5'-flanking region from -621 to -568 couldn't analyze due to lack of appropriate primer sets. The methylation rates of cytosines with different sequence backgrounds are shown in the inset box. (D) Down-regulation of the GBSSI transcript level in t33. (E) Percentage of amylose in the starch of WT and t33 tubers (n = 5) which were grown in a glasshouse. 10.1371/journal.pone.0161729.g003Fig 3 DNA methylation in the MTs formed trans-grafting plant and its stability in the progenies. Two MTs formed on a grafted plant and their progeny tubers were analyzed. Actually, the sprouts from tubers were used as analyzing materials. Asterisks show statistically significant (*; p<0.05, **; P<0.01 Student's t test) differences relative to WT. Means and SD of 3 biological replicates are shown. To clarify whether the siRNAs transported via the graft union remain in the epigenetically changed potato, we sequenced and analyzed small RNA pools derived from the respective materials. The siRNA profile of 35SGBpIR transgenic potato (t33) showed 2169 reads/million sequence reads (reads/MR) of the target-specific siRNAs (20–24 nt, S2 Table), indicating that the construct we prepared is as efficient at producing siRNAs as that reported previously [15]. While both 3- and 7-wag (weeks after grafting) lateral shoots of rootstock potato exhibited almost the same levels and characteristics in terms of siRNA profiles, the profile of epigenetic potato (Epi-A) showed only 0.07 reads-MR, being lower than that for WT potato (Fig 4 and S2 Table). These results suggest that the siRNAs are transported continuously as long as the graft remains established, but then are decompose quickly. Finally, to clarify whether or not methylation by the graft-transported siRNA occurs at sites other than the intended target, we analyzed off-target effects using our deep seq analysis data. Only one siRNA sequence (0.15 reads/million in the transgenic t33) out of 377 types of 24-nt siRNAs detected 54 putative off-target sites in the potato genome (Solanum tuberosum, PGSC V4.03). All of these sites comprised the 14-bp core sequence and a SSR (AAG / CTT), and adjacent sequences of about 10 bp were also homologous with each other (S11A Fig). The methylation levels of two out of 54 putative off-target sites were analyzed because both were located within the intron of a gene (function unknown) (S11B Fig). Both levels of methylation were not significantly different from that of the WT line (S11C Fig), indicating that both putative off-target sites were not modified at all. As the performance of RdDM requires cooperative work with RNA pol V products and DRM2 (DOMAINS REARRANGED METHYLTRANSFERASE 2), it is considered that the achievement is not merely because of the same sequence with the 24-nt siRNA. 10.1371/journal.pone.0161729.g004Fig 4 Fractions of 20–24 nt siRNA mapped to GBSSI target region in different materials. 3 wag and 7 wag are adventitious shoots (Fig 1B4) in rootstock of trans-grafted plants at 3 weeks and 7 weeks after grafting, respectively. Epi-A is t9A in Fig 3. Obtaining TGS MTs is very convenient for potato propagation because they can be used directly as seed potatoes. In the present study, however, all of the tubers formed from the hetero-graft plants did not manifest the TGS. Probably, when the callus formed on the cut surface of a lateral shoot is derived from distinct TGS cells of vascular bundle tissue, the whole TGS shoot and MT would be obtainable by stable maintenance of the epigenetic change through mitosis. Thus, it is considered that the use of a scion having a high degree of siRNA transport ability and the timing of callus formation on the cut surface are important for obtaining the TGS MTs. In 2012, eight new plant breeding techniques (NPBT) were introduced [25]. Our graft-induced TGS involves a combination of two such technologies: grafting and RdDM (RNA-dependent DNA methylation). Grafting [26] as a NPBT involves culture of a chimeric plant formed between a non-GM scion on a GM root stock [27]. Although some specific RNAs and proteins can move through the graft union [28, 29], there is no evidence that the transgene moves into the graft partner [27]. With our technique, the scion is used as the supplier of a specifically designed siRNA molecule and the root stock provokes epigenetic changes through the natural system of RdDM that already exists in plant cells. When evaluating the resulting product from the viewpoint of NPBT [30], the potato thus created would be exempt from GM restriction. Furthermore, the epigenetically modified potato would not retain the functional siRNA, and the resulting plant would not contain any recombinant DNA. Large-scale cultivation of the epigenetically improved potato is now being planned to verify the stable maintenance of TGS, and another interesting gene in the potato genome is also being targeted for TGS. Thus, we have exploited an epigenome editing technology by mobile RNA silencing for crop improvement [31]. Materials and Methods Plant materials and growth conditions Potato (Solanum tuberosum tetraploid cultivar ‘Waseshiro’) shoots were grown in vitro on MS medium [32] with 30 g/l sucrose and 3 g/l Gellan gum (Wako Pure Chemical, Japan), at 24°C under a 16-h light / 8-h dark cycle with cool fluorescent light at about 100 μmol m–2 s–1. Generation of transgenic plants A mGFP5-ER gene (accession no. U87974) from GFP transgenic Nicotiana benthamiana line 16C [33], amplified by PCR using primer GFP FP and RP, was inserted into the XbaI/SacI sites of pIG121 [34]. mGFP5-ER nearly full length (670 bp out of 792 bp) amplified by PCR using primer GFP S5’FP and S3’RP, was inserted into the pBluescript SKII (+) (STRATAGENE) by TA cloning. The CAT1 intron [34], amplified by PCR using primer intFP and intRP, was inserted into the AatII/HindIII sites as a spacer. For the reverse sequence, an amplification product obtained using primers GFP A3’FP and A5’RP was inserted into the HindIII/ KpnI sites. Finally, CoYMV:GFPIR was completed by inserting GFPIR into the BamHI/KpnI sites of CoYMV-NOSter [16]. The promoter sequence of GBSSI was obtained from the data of Sol genomic network data (http://solgenomics.net/) and a previous paper [23] by PCR using stGBSSIFP and RP. Transformation of potato and the formation of MT was performed essentially as described [18, 35], respectively. Transformation of tobacco and construction of the empty vector was performed as described previously [36]. DNA sequences of each plasmid were confirmed using an ABI 3500 Genetic Analyzer (Life Technologies). All of the primer sequences used in this study are listed in S1 Table. Methylation assays Total genomic DNA was extracted in CTAB buffer (3% CTAB, 0.1 M Tris-HCl pH 8.0, 20 mM EDTA, 1.4 M NaCl) from 50 to 100 mg of plant tissue (leaves or MT). After treatment with RNaseA (Nacalai Tesque, Japan), gDNA was purified using a QuickGene DNA tissue Kit S and QuickGene-Mini80 (KURABO, Japan). The level of DNA methylation was analyzed by treatment with the methylation-dependent restriction enzyme McrBC and qRT-PCR. Bisulfite sequencing and the qPCR analysis were performed essentially as described [14] using a SsoFast EvaGreen Supermix with a Chrome4 real-time PCR detector (Bio-Rad). The primers used for the methylation assays are described in S1 Table. Total RNA extraction and qPCR analysis Total RNA from leaves was extracted using TRizol reagent (Life Technologies). In the case of MT total RNA, the tuber was ground with a mortar and pestle using liquid N2, then the powder was suspended in 300 μl washing buffer (0.1 M Tris-HCl pH 8.0, 0.1% PVP, 4% 2-Me). The supernatant (250 μl) of the centrifuged mixture was treated with 750 μl Trizol LS reagent, and followed by protocols (Life Technologies). In brief, extracted total RNA was treated with TURBO DNA-free. cDNA was synthesized with 1 μg RNA as a template using a Superscript VlLO. The primers used for qPCR are described in S1 Table. Northern blot analysis The initial step for extraction of small RNAs was performed essentially as described previously [36]. The small RNAs obtained were refined by filter purification using mirVana miRNA isolation kit (Life Technologies) in accordance with the manufacturer's instructions. Detection of siRNAs was performed as described previously [36]. The DIG-labeled sense and antisense riboprobe corresponding promoter regions sequence of the target was synthesized using the DIG RNA Labeling kit (Roche). Southern blot analysis Genomic DNA of transformants was extracted from 0.5 to 2.0 g of leaves using the Urea buffer. Fifteen micrograms of genomic DNA was digested with a restriction enzyme and separated by electrophoresis on a 1% agarose gel. The electrophoresed DNA was then blotted onto a nylon membrane (Pall Corp.). Digoxigenin-labeled DNA probes corresponding to 35S promoter were synthesized using a DIG DNA labeling kit (Roche). The DIG DNA probe was hybridized to the DNA at 45°C in DIG Easy Hyb solution. The membrane was washed twice with 2×SSC/0.1% SDS at room temperature and then twice with 0.5×SSC/0.1% SDS at 68°C. Visualization and detection were performed in the same way as that for Northern blot analysis. Micrografting, regeneration from roots, and MT formation The siRNAs donor N. benthamiana plants were germinated on MS medium [32] and the siRNA recipient S. tuberosum shoots were from plant subcultured on MS30 medium. A shoot of N. benthamiana was cut horizontally at approximately 2~3 cm below the tip and its cut side was inserted into a silicone tube (5 mm length, 1 mm internal × 2 mm external diameter) with a vertical slit. Then, the siRNAs recipient potato was also cut horizontally at approximately 2~3 cm above the rooting part including at least one lateral bud and was also inserted into the silicone tube to allow both cut surfaces to cohere. The shoot that sprouted from the lateral bud of the potato stock was cut off using a razor, and the hetero-grafted plants were grown for a further two months. Then, the roots were harvested to analyze their methylation and to obtain the regenerated shoot by culture on MS20, BAP 3 mg/l, TDZ 0.1 mg/l, agar 7 g/l for two weeks, and then on MS20, zeatin 3 mg/l, IAA 10 μg/l, GA3 0.2 mg/l, Gellan gum 3 g/l for further two weeks, To obtain MT, the shoots were placed on MS80 containing 5 μM BAP [18]. GFP imaging GFP fluorescence was photographed using a digital camera (α58, SONY) with a UV-cut filter (Y2 Professional, Kenko·Tokina, Japan) under UV light from a hand-held 100 W long-wavelength UV lamp (B100AP; UVP Ultraviolet Products, Upland, USA). GFP fluorescence on MT samples was monitored with a biological fluorescence microscope (BX61, Olympus, Japan), and their digital images were captured with a digital camera (DP71, Olympus) connected to the microscope. Amylose content The amylose percentage was determined spectrophotometrically in 100 mg of extracted starch according to the method described previously [37]. Extraction of small RNAs and detection of siRNAs The initial step for extraction of small RNAs was performed essentially as described previously [36]. The small RNAs obtained were refined by filter purification using a mirVana miRNA isolation kit (Life Technologies) in accordance with manufacturer’s instructions. Detection of siRNAs was performed essentially as described previously [14]. The DIG-labeled sense and antisense riboprobe corresponding the promoter region sequence of the target was synthesized using the DIG RNA Labeling kit (Roche). Next-generation sequencing Twenty micrograms of total RNA was extracted from shoots of in vitro cultured WT, epigenetically changed potato (Epi-A), t33 and lateral shoots of hetero-grafted plants. Respective samples were composed of at least 15 individuals. Extracted RNA samples were sent to Hokkaido System Science (Sapporo, Japan) for next-generation sequence analysis using an Illumina HiSeq (Illumina, San Diego, CA). After removing the adapter sequences, non-redundant reads (>16 nt) were retained. The filtered reads were analyzed to match against the StGBSS 5' region a or b (S6 Fig), respectively. The resulting reads are summarized in S2 Table. Supporting Information S1 Fig Regeneration from the callus formed on cut surface of potato shoot. After cutting by a razor blade (0 dac, red arrow head), small callus was formed on the surface (11 dac) and then adventitious bud (15 dac) and the shoot (21 dac) grew. To observe clearly the adventitious shoot formation, subcultured shoot was used alone. (TIF) Click here for additional data file. S2 Fig Maintenance of TGS status in the progenies. (A) Methylation level in the target region. Second progeny tubers of the 35S:GFP line (35SGFP), WT/35S:GFP (Cont), CoGFPIR/35S:GFP (PTGS), and Co35SIR/35S:GFP (TGS). (B) GFP expression in the 2nd progeny tubers. The MTs were cut with a razor blade and their surface was observed under UV light. (C) Level of the GFP transcript. Asterisks show statistically significant (*; p<0.05 Student's t test) differences relative to 35SGFP. Means and SD of 3 biological replicates are shown. (TIF) Click here for additional data file. S3 Fig TGS induction in MTs on potato rootstock grafted with Co35SIR N. benthamiana scion. (A) MT formed on a regenerated lateral shoot of root stock potato. (B) GFP expression of MTs. (C) Methylation level of the target region. (D) Methylation level of non-target region (+113~+782 bp of GFP). Asterisks show statistically significant (*; p<0.05 Student's t test) differences relative to 35SGFP. Error bars indicate 95% confidence intervals from 2 to 4 biological replicates. (TIF) Click here for additional data file. S4 Fig TGS manifestation of 35S:GFP in roots of 35SGFP potato grafted with Co35SIR N. benthamiana scion. (A) The roots harvested at two months after grafting. Top shows bright-field images and bottom shows UV fluorescence images. (B) Methylation levels in bulked roots. (C) Levels of GFP transcript in bulked root. Asterisks show statistically significant (*; p<0.05 Student's t test) differences relative to 35SGFP. Means and SD of 3 biological replicates are shown. (TIF) Click here for additional data file. S5 Fig TGS of 35S:GFP in regenerated shoot from the root of the grafted plant between Co35SIR N. benthamiana scion and potato rootstock. (A) Regenerated potato shoots and their GFP expressions. (B) Methylation levels of target and non-target (+113 ~ +782 bp of GFP) regions in regenerated shoots. (C) GFP expression of MTs formed on the regenerated shoots. Asterisks show statistically significant (*; p<0.05 Student's t test) differences relative to 35SGFP. Error bars indicate 95% confidence intervals from 2 to 4 biological replicates. (TIF) Click here for additional data file. S6 Fig GBSSI-a and GBSSI-b of potato cultivar 'Waseshiro'. (A) Schematic presentation of the difference between the 5’flanking regions of a and b allele. (B) Sequence alignment of the 5’flanking regions of a and b allele. (TIF) Click here for additional data file. S7 Fig Characterization of CoGBpIR transgenic potato line t33. (A) Southern hybridization. (B) Northern blot analysis for siRNA of target region. (TIF) Click here for additional data file. S8 Fig Northern blot analysis of siRNAs targeting the GBSSI promoter in the transgenic N. benthamiana. Small RNA enriched nucleic acid (10 μg) was analyzed in 15% polyacrylamide gel and probed with the promoter negative (top) and positive (middle) strand RNA. 5.8S rRNA hybridization (bottom) was used as a loading control. (TIF) Click here for additional data file. S9 Fig Genomic PCR of GBSSI Epi-A potato. (A) Schematic diagrams of the TGS starter CoGBpIR. Orange line indicates the target region. Blue bars (1~7) show the regions where PCR experiments were carried out to know whether the starter CoGBpIR is present in the potato. (B) Genomic PCR products in each region. From left to right, WT potato, A1 and A2 lines of Epi-A potato, and CoGBpIR N. benthamiana. Actin gene of S. tuberosum and ubiquitin gene of N.b. are amplified as controls. The size (bp) of the PCR products are shown at the both sides. (C) Results arranged in each PCR experiment. In region 5, two alleles of GBSSI-a and -b (Fig 3) were amplified. (D) Cycling conditions of the PCR. (TIF) Click here for additional data file. S10 Fig Methylation level of the GBSSI target region in a shoot from the root of a hetero-grafted CoGBpIR N. benthamiana scion and potato rootstock. Three independent TGS lines regenerated from the grafted plant roots were analyzed. Asterisks show statistically significant (*; p<0.05, **; P<0.01 Student's t test) differences relative to the WT. Means and SD of 3 biological replicates are shown. (TIF) Click here for additional data file. S11 Fig Analysis of putative off-target effects on a GBSSI TGS line. (A) Sequence alignment of putative off-target sites on PGSC (Potato Genome Sequencing Consortium) data. Each site is indicated with the chromosome number and start position. Two sequences analyzed for their methylation level are shown against a yellow background. Off-target sequences are shown in red and blue, and black underlining indicates the core sequence. (B) Sequence alignment of two off-target sites between PGSC and ‘Waseshiro’ WT. SSR = simple sequence repeat. (C) Methylation level of the GBSSI target region and two off-target sites. Asterisks show statistically significant (**; P<0.01 Student's t test) differences relative to the WT. Means and SD of 3 biological replicates are shown. (TIF) Click here for additional data file. S1 Table List of oligonucleotides used in this study. (DOCX) Click here for additional data file. S2 Table Summary of siRNAs mapped to the target region of GBSSI. (DOCX) Click here for additional data file. We are grateful to Prof. N. Olszewski (University of Minnesota) for the CoYMV promoter. We thank Matsutani Chemical Industry Co., Ltd. for help with starch analysis, and K. Fujino (Hokkaido University) for providing potato ‘Waseshiro’. ==== Refs References 1 Berger SL , Kouzarides T , Shiekhattar R , Shilarifard A (2009 ) An operational definition of epigenetics . Genes Dev 23 : 781 –783 . 10.1101/gad.1787609 19339683 2 Aufsatz W , Mette MF , van der Winden J , Matzke AJ , Matzke M (2002 ) RNA-directed DNA methylation in Arabidopsis . Proc. Natl. Acad. Sci. USA 99 : 16499 –16506 . 12169664 3 Henderson IR , Jacobsen SE (2007 ) Epigenetic inheritance in plants . Nature 447 : 418 –424 . 17522675 4 Matzke M , Kanno T , Daxinger L , Huettel B , Matzke AJ (2009 ) RNA-mediated chromatin-based silencing in plants . Curr Opin Cell Biol 21 : 367 –376 . 10.1016/j.ceb.2009.01.025 19243928 5 Feng S , Jacobsen SE , Reik W (2010 ) Epigenetic reprogramming in plant and animal development . Science 330 : 622 –627 . 10.1126/science.1190614 21030646 6 Crevillén P . et al (2014 ) Epigenetic reprogramming that prevents transgenerational inheritance of the vernalized state . Nature 515 : 587 –590 . 10.1038/nature13722 25219852 7 Otagaki S , Kawai M , Masuta C , Kanazawa A (2011 ) Size and positional effects of promoter RNA segments on virus-induced RNA-directed DNA methylation and transcrptional gene silencing . Epigenetics 6 : 681 –691 . 21610318 8 Calarco J. P. et al (2012 ) Reprogramming of DNA methylation in pollen guides epigenetic inheritance via small RNA . Cell 151 : 194 –205 . 10.1016/j.cell.2012.09.001 23000270 9 Law JA , Jacobsen SE (2010 ) Establishing, maintaining and modifying DNA methylation patterns in plants and animals . Nat Rev Genet 11 : 204 –220 . 10.1038/nrg2719 20142834 10 Kawashima T , Berger F (2014 ) Epigenetic reprogramming in plant sexual reproduction . Nature Rev Genet 15 : 613 –624 . 10.1038/nrg3685 25048170 11 Springer NM (2013 ) Epigenetics and crop improvement . Trends Genet 29 : 241 –247 . 10.1016/j.tig.2012.10.009 23128009 12 Zhang C , Hsieh T-F (2013 ) Heritable epigenetic variation and its potential applications for crop improvement . Plant Breed Biotech 1 : 307 –319 . 13 Molnar A . et al (2010 ) Small silencing RNAs in plants are mobile and direct epigenetic modification in recipient cells . Science 328 : 872 –875 . 10.1126/science.1187959 20413459 14 Bai S , Kasai A , Yamada K , Li T , Harada T (2011 ) A mobile signal transported over a long distance induces systemic transcriptional gene silencing in a grafted partner . J Exp Bot 62 : 4561 –4570 . 10.1093/jxb/err163 21652532 15 Zhang W . et al (2014 ) Graft-transmissible movement of inverted-repeat-induced siRNA signals into flowers . Plant J 80 : 106 –121 . 10.1111/tpj.12622 25039964 16 Zhao D , Song G (2014 ) Rootstock-to-scion transfer of transgene-derived small interfering RNAs and their effect on virus resistance in nontransgenic sweet cherry . Plant Biotechnol J 12 :1319 –1328 . 10.1111/pbi.12243 25132092 17 Matsuda Y . et al (2002 ) The commelina yellow mottle virus promoter drives companion-cell-specific gene expression in multiple organs of transgenic tobacco . Protoplasma 220 : 51 –58 . 12417936 18 Hendriks T , Vregudenhil D , Stiekema WJ (1991 ) Patatin and for serine proteinase inhibitor genes are differentially expressed during potato tuber development . Plant Mol Biol 17 : 385 –394 . 1715784 19 Goh T , Joi S , Mimura T , Fukaki H (2012 ) The establishment of asymmetry in Arabidopsis lateral root founder cells is regulated by LBD16/ASL18 and related LBD/ASL proteins . Development 139 : 883 –893 . 10.1242/dev.071928 22278921 20 Smet ID (2011 ) Lateral root initiation: one step at a time . New Phytol 193 : 867 –873 . 21 Kuipers AGJ , Jacobsen E , Visser RGF (1994 ) Formation and deposition of amylose in the potato tuber starch granule are affected by the reduction of granule-bound starch synthase gene expression . Plant Cell 6 : 43 –52 . 12244219 22 Heilersig HJB , Loonen A , Bergervoet M , Wolters AMA , Visser RGF (2006 ) Post-transcriptional gene, silencing in potato: effects of size and sequence of the inverted repeats . Plant Mol Biol 60 : 647 –662 . 16649104 23 Heilersig HJB , Loonen A , Janssen FM , Wolters AMA , Visser RGF (2006 ) Efficiency of transcriptional gene silencing of GBSSI in potato depends on the promoter region that is used in an inverted repeat . Mol Gen Genomics 275 : 437 –449 . 24 Hartung F , Schiemann J (2014 ) Precise plant breeding using new genome editing techniques: opportunities, safety and regulation in the EU . Plant J 78 : 742 –752 . 10.1111/tpj.12413 24330272 25 Lusser M , Parisi C , Plan D , Rodríguez-Cerezo E (2012 ) Deployment of new biotechnologies in plant breeding . Nature Genet . 30 : 231 –239 . 26 Mudge K , Janick J , Scofield S , Goldschmidt EE (2009 ) A history of grafting Horticulture Reviews , vol. 35 , eds Jules Janick (Wiley-Blackwell ), pp 437 –493 . 27 Haroldsen VM , Chi-Ham CL , Bennett AB (2012 ) Transgene mobilization and regulatory uncertainty for non-GE fruit products of transgenic rootstocks . J. Biotechnol . 161 : 349 –353 . 10.1016/j.jbiotec.2012.06.017 22749907 28 Han X , Kumar D , Chen H , Wu S , Kim JY (2014 ) Transcription factor-mediated cell-to-cell signaling in plants . J. Exp. Bot . 65 : 1737 –1749 . 10.1093/jxb/ert422 24347464 29 Hannapel DJ , Sharma P , Lin T (2014 ) Phloem-mobile messenger RNAs and root development . Front Plant Sci . 4 : 257 . 30 Heap B (2013 ) Europe should rethink its stance on GM crops . Nature 498 : 409 10.1038/498409a 23803807 31 Gohlke J , Mosher RA (2015 ) Exploiting mobile silencing for crop improvement . Am. J. Bot . 102 : 1399 –1400 . 10.3732/ajb.1500173 26391704 32 Murashige T , Skoog F (1962 ) A revised medium for rapid growth and bio assays with tobacco tissue culture . Physiol. Plant . 15 : 473 –497 . 33 Ruiz MT , Voinnet O , Baulcombe DC (1998 ) Initiation and maintenance of virus-induced gene silencing . Plant Cell 10 : 937 –946 . 9634582 34 Tanaka A , Mita S , Kyozuka J , Shimamoto K , Nakamura K (1990 ) Enhancement of foreign gene expression by a dicot intron in rice but not in tobacco is correlated with an increased level of mRNA and an efficient splicing of the intron . Nucleic Acids Res . 18 : 6767 –6770 . 2263444 35 Beaujean A , Sangwan RS , Lecardonnel A , Sangwan-Norreel BS (1988 ) Agrobacterium-mediated transformation of three economically important potato cultivars using sliced internodal explants: an efficient protocol of transformation . J. Exp. Bot . 49 : 1589 –1595 . 36 Kasai A , Bai S , Li T , Harada T (2011 ) Graft-transmitted siRNA signal from the root induces visual manifestation of endogenous post-transcriptional gene silencing in the scion . PLoS ONE 6 : e16895 10.1371/journal.pone.0016895 21347381 37 Hovenkamp-Hermelink JHM . et al (1987 ) Isolation of an amylose-free starch mutant of the potato (Solanum tuberosum L.) . Theor. Appl. Genet . 75 : 217 –221 .

PMC005xxxxxx/PMC5001711.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756485210.1371/journal.pone.0161908PONE-D-16-25182Research ArticleBiology and Life SciencesPhysiologyImmune PhysiologyAntibodiesAutoantibodiesMedicine and Health SciencesPhysiologyImmune PhysiologyAntibodiesAutoantibodiesBiology and Life SciencesImmunologyImmune System ProteinsAntibodiesAutoantibodiesMedicine and Health SciencesImmunologyImmune System ProteinsAntibodiesAutoantibodiesBiology and Life SciencesBiochemistryProteinsImmune System ProteinsAntibodiesAutoantibodiesMedicine and Health SciencesPulmonologyPneumoniaMedicine and Health SciencesPulmonologyInterstitial Lung DiseasesMedicine and Health SciencesDiagnostic MedicineMedicine and Health SciencesDiagnostic MedicinePrognosisBiology and Life SciencesAnatomyHistologyMedicine and Health SciencesAnatomyHistologyBiology and Life SciencesDevelopmental BiologyFibrosisPeople and PlacesDemographyDeath RatesBiology and Life SciencesPopulation BiologyPopulation MetricsDeath RatesClinical Features of Idiopathic Interstitial Pneumonia with Systemic Sclerosis-Related Autoantibody in Comparison with Interstitial Pneumonia with Systemic Sclerosis Interstitial Pneumonia with Systemic Sclerosis-Related AutoantibodyYamakawa Hideaki 16*Hagiwara Eri 1Kitamura Hideya 16Yamanaka Yumie 16Ikeda Satoshi 1Sekine Akimasa 1Baba Tomohisa 1Iso Shinichiro 2Okudela Koji 3Iwasawa Tae 4Takemura Tamiko 5Kuwano Kazuyoshi 6Ogura Takashi 11 Department of Respiratory Medicine, Kanagawa Cardiovascular and Respiratory Center, Yokohama, Japan2 Department of Radiology, Yokohama Rousai Hospital for Labour Welfare Corporation, Yokohama, Japan3 Department of Pathobiology, Yokohama City University Graduate School of Medicine, Yokohama, Japan4 Department of Radiology, Kanagawa Cardiovascular and Respiratory Center, Yokohama, Japan5 Department of Pathology, Japanese Red Cross Medical Center, Tokyo, Japan6 Department of Respiratory Medicine, Tokyo Jikei University Hospital, Tokyo, JapanKuwana Masataka EditorJAPANCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: HY TO. Investigation: HY SI KO TI TT TO. Methodology: HY EH HK YY SI AS TB KK. Project administration: HY TO. Supervision: TO. Validation: HY KK TO. Writing – original draft: HY EH. Writing – review & editing: EH TT KK TO. * E-mail: hide1144@jikei.ac.jp26 8 2016 2016 11 8 e016190823 6 2016 12 8 2016 © 2016 Yamakawa et al2016Yamakawa et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Background Patients with idiopathic interstitial pneumonias sometimes have a few features of connective tissue disease (CTD) and yet do not fulfil the diagnostic criteria for any specific CTD. Objective This study was conducted to elucidate the characteristics, prognosis, and disease behavior in patients with interstitial lung disease (ILD) associated with systemic sclerosis (SSc)-related autoantibodies. Methods We retrospectively analyzed medical records of 72 ILD patients: 40 patients with SSc (SSc-ILD) and 32 patients with SSc-related autoantibody-positive ILD but not with CTD (ScAb-ILD), indicating lung-dominant CTD with SSc-related autoantibody. Results Patients with SSc-ILD were predominantly females and non-smokers, and most had nonspecific interstitial pneumonia confirmed by high-resolution computed tomography (HRCT) and pathological analysis. However, about half of the patients with ScAb-ILD were male and current or ex-smokers. On HRCT analysis, honeycombing was more predominant in patients with ScAb-ILD than with SSc-ILD. Pathological analysis showed the severity of vascular intimal or medial thickening in the SSc-ILD patients to be significantly higher than that in the ScAb-ILD patients. Survival curves showed that the patients with ScAb-ILD had a significantly poorer outcome than those with SSc-ILD. Conclusion Data from this study suggest that lung-dominant CTD with SSc-related autoantibody is a different disease entity from SSc-ILD. The authors received no specific funding for this work. Data AvailabilityIndividual patient data cannot be publicly deposited or included in the supporting information due to patient privacy. If researchers would like to access these data, data are available from Ms. Yukiko Ogasawara at ogasawara@kanagawa-junko.jp.Data Availability Individual patient data cannot be publicly deposited or included in the supporting information due to patient privacy. If researchers would like to access these data, data are available from Ms. Yukiko Ogasawara at ogasawara@kanagawa-junko.jp. ==== Body Introduction Connective tissue disease (CTD) is often associated with interstitial pneumonia. An evaluation for underlying CTD is recommended in the diagnosis and management of idiopathic interstitial pneumonias [1]. Some patients with idiopathic interstitial pneumonia have a few features of CTD and yet do not fulfil the diagnostic criteria for any specific CTD [2, 3]. These patients have been described previously as having undifferentiated CTD-associated interstitial lung disease (UCTD-ILD), lung-dominant CTD, and autoimmune-featured ILD, and the recently proposed term “interstitial pneumonia with autoimmune features” (IPAF) [1–8]. Among the various CTDs, rheumatoid arthritis (RA), systemic sclerosis (SSc), and polymyositis/dermatomyositis are more likely to be associated with ILD [9, 10]. Previous reports have indicated that clinical characteristics are similar between RA-ILD and ILD with anti-CCP antibody but not with RA, and between polymyositis/dermatomyositis-ILD and anti-ARS antibody-associated ILD but not with polymyositis/dermatomyositis [11, 12]. Specific study of each autoantibody found in ILD may be necessary to assess the appropriate strategy to diagnose and treat these disorders. We hypothesized that ILD with SSc-related autoantibodies would not resemble SSc-ILD, and this point may contribute to the reasons for the difference prognoses in previous reports of UCTD-ILD, lung-dominant CTD, autoimmune-featured ILD, and IPAF [2–7, 13]. The aims of our study were thus to review and assess clinical characteristics, prognosis, and disease behavior of SSc-related autoantibody-associated ILD. Material and Methods Study sample We retrospectively surveyed all patients who were diagnosed as having SSc-ILD or SSc-related autoantibody (anti-centromere, anti-scleroderma-70, and anti-U1 RNP antibody)-positive ILD but not CTD (ScAb-ILD) at Kanagawa Cardiovascular and Respiratory Center, Kanagawa, Japan, between March 1997 and July 2015. Among the patients with SSc-ILD, 3 patients with SSc-RA overlap and 1 patient with SSc- dermatomyositis overlap were excluded from this study. Moreover, among the patients with ScAb-ILD, 4 patients with RA, 4 patients with Sjögren’s syndrome, 2 patients with chronic hypersensitive pneumonitis, and 1 patient with IgG4-associated lung disease were also excluded. Diagnosis of SSc was made by rheumatologists at other institutions. Patients with SSc fulfilled the revised criteria for SSc of the American College of Rheumatology/European League Against Rheumatism (ACR/EULAR) classification [14]. In 4 patients with SSC-ILD, SSc preceded the onset of ILD (follow-up: 2.17–17.5 years). On the other hand, 2 patients developed manifestations of SSc during their follow-up period (5–7 months), and these patients were also included as SSc-ILD subjects. Although mixed connective tissue disease (MCTD) was originally described as a syndrome, the presence of anti-U1 RNP antibody is not restricted to MCTD but is also occasionally observed in patients with systemic lupus erythematosus and SSc [15]. Patients with SSc present with many disease features that are also found in MCTD, and if a single CTD becomes dominant, most patients who have MCTD develop SSc, and some MCTD patients often evolve toward severe SSc [16–18]. Therefore, in our study, 11 patients with MCTD were included as having SSc-ILD, and 15 anti-U1 RNP-positive patients with ILD were also included as having ScAb-ILD. Diagnosis of MCTD required fulfillment of at least one set of the three widely accepted diagnostic criteria (Sharp, Alarcon-Segovia, and Kasukawa criteria) [15, 19]. This study received approval from the institutional review board of Kanagawa Cardiovascular and Respiratory Center (no. 27–39). Informed consent was not required because of the retrospective nature of the study. Data collection Baseline clinical measures, other than bronchoalveolar lavage and histological findings by surgical lung biopsy, were obtained within one month of the initial diagnosis of SSc-ILD or ScAb-ILD at our hospital. Bronchoalveolar lavage and histological findings by surgical lung biopsy were obtained within 3 months (27 patients), 3–6 months (4 patients), and 3.25 years (1 patient) of diagnosis. A broad panel of autoantibodies were screened in a clinical setting even if there were no symptoms suspicious of CTD. Anti-centromere, anti-scleroderma-70, and anti-U1 RNP antibodies were tested using fluoroenzymeimmunoassay testing of sera (Thermo Fischer Scientific Inc., Tokyo, Japan). A positive result for each autoantibody was defined as a measurement greater than 10 U/mL. Radiological analysis Two radiologists (T. Iwasawa and S. Iso) reviewed high-resolution computed tomography (HRCT) scans for consensus of diagnosis of ILD in our hospital without information of each patient’s clinical data. Patients were classified as presenting a HRCT pattern either “suggestive or consistent with nonspecific interstitial pneumonia (NSIP)” or “suggestive of usual interstitial pneumonia (UIP)” [20, 21]. Pleuroparenchymal fibroelastosis was defined Reddy’s criteria [22]. The HRCT scans were analyzed for the following characteristics: honeycombing, ground-glass opacity, consolidation, reticulation, localization of attenuation in conjunction with honeycombing and cysts, traction bronchiectasis, bronchial wall thickening, pulmonary artery dilation, enlarged mediastinal lymph nodes, and pleural thickening. These features were selected on the basis of previous studies or from our experience [23–25]. Disagreements between the two radiologists after the first assessment were resolved by discussion. We also used computer-aided 3D quantitative analysis of chest HRCT to automatically categorize the lungs in the 3D CT images pixel-by-pixel with Gaussian histogram-normalized relations, and relative volume of disease extent (%) to HRCT lung volume was calculated [26, 27]. A previous study reported that an easily applicable, limited/extensive disease staging system for SSc-ILD based on combined evaluation with HRCT and pulmonary function testing (PFT) provides discriminatory prognostic information [28]. We thus used this staging system to categorize our study patients as having limited disease (HRCT extent ≤ 10% or, when HRCT extent was 10–30%, forced vital capacity [FVC] ≥ 70%) or extensive disease (HRCT > 30% or, when HRCT extent was 10–30%, FVC < 70%) [28]. Pathological analysis The surgical lung biopsy slides were reviewed by two pulmonary pathologists (K. Okudela and T. Takemura) who were blinded to the clinical and radiologic information. Histologic patterns were classified according to the current classification of idiopathic interstitial pneumonia [21]. Moreover, the following pathological features were semi-quantitatively graded as 0 (absent), 1 (mild), 2 (moderate), or 3 (severe): lung parenchyma, airway, and pleural lesions [29–31]. Any disagreements between the two pathologists were discussed until consensus was reached. Statistical analysis Categorical baseline characteristics are summarized by frequency and percentage, and continuous characteristic are reported as mean ± SD. To detect differences between groups, the Wilcoxon test or Fisher’s exact test was used as appropriate. We investigated potential risk factors of mortality with each variable chosen for entry into univariate Cox regression analysis and then performed multivariate Cox regression analysis with backward variable selection. The Kaplan-Meier method was used to display and the log-rank test to compare survival curves for the cohort stratified for each group (SSc-ILD and ScAb-ILD; limited and extensive staging of disease; Krebs von den Lungen-6 (KL-6) ≥ 1000 U/mL and < 1000 U/mL; and anti-centromere, anti-scleroderma-70, and anti-U1RNP antibody). Analysis of disease behavior based on FVC over time was performed with linear mixed-effects models in which separate fits for subjects with SSc-ILD or ScAb-ILD were allowed. Each model included random terms for intercept and slope (for time from diagnosis) to account for the data structure (repeated measures over time within subject). We considered p < 0.05 to represent statistical significance in all analyses. Missing data were categorized as “unknown” and were entered into each statistical analysis model. All data were analyzed with SAS version 9.4 (SAS Institute Inc.). Results Patient characteristics We identified 72 subjects of whom 40 were patients with SSc-ILD and 32 were patients with ScAb-ILD. Significantly more of the patients with SSc-ILD were women and never-smokers than were patients with ScAb-ILD (p = 0.013) (Table 1). The CD4/CD8 ratio of bronchoalveolar lavage fluid was significantly lower in the patients with SSc-ILD (p = 0.042). A quarter of the patients with SSc-ILD used cyclophosphamide. The median follow-up period was 2.51 years (range: 0.20–17.25 years) in SSc-ILD patients and 3.25 years (range: 0.15–9.75 years) in ScAb-ILD patients. 10.1371/journal.pone.0161908.t001Table 1 Baseline characteristics at the time of diagnosis as interstitial lung disease. Characteristics All subjects SSc-ILD ScAb-ILD P value No. of patients 72 40 32 Female N (%) 51 (70.8) 34 (85.0) 17 (53.1) 0.004* Age, mean ± SD 65.3 ± 14.6 61.7 ± 16.4 69.7 ± 10.8 0.059 Current or ex-smoker N (%) 26 (36.1) 9 (22.5) 17 (53.1) 0.013* Body mass index, kg/m2 22.18 ± 3.49 22.41 ± 3.18 21.89 ± 3.88 0.537 Positive autoantibody N (%)  Anti-centromere antibody 26 (36.1) 13 (32.5) 13 (40.6) 0.161  Anti-scleroderma-70 antibody 15 (20.8) 11 (27.5) 4 (12.5)  Anti-U1 RNP antibody 26 (36.1) 11 (27.5) 15 (46.9) Lymphocyte, /μL 1712.5 ± 644.2 1703.3 ± 601.8 1724.1 ± 703.3 0.747 Hb, g/dL 13.25 ± 1.44 13.14 ± 1.42 13.39 ± 1.47 0.289 Albumin, g/dL 3.96 ± 0.39 3.94 ± 0.38 3.98 ± 0.40 0.601 LDH, IU/L 248.0 ± 61.3 241.7 ± 64.4 255.9 ± 57.3 0.153 CRP, mg/dL 0.5 ± 0.892 0.462 ± 0.978 0.548 ±0.784 0.31 KL-6 (available N) 70 39 31  U/mL 1307.4 ± 951.2 1266.9 ± 938.0 1358.4 ± 980.6 0.727  ≥ 1000 U/mL N (%) 36 (50.0) 19 (47.5) 17 (53.1) 0.638 SP-D (available N) 64 36 28  ng/mL 190.52 ± 130.65 196.61 ± 141.94 182.7 ± 116.61 0.855  ≥ 110 ng/mL N (%) 43 (59.7) 24 (60.0) 19 (59.4) 1 Pulmonary function tests  Subjects (available N) 65 39 26  FEV1/ FVC ratio, % 79.8 ± 10.7 80.1 ± 8.6 79.4 ± 13.4 0.952  FVC, % predicted 85.74 ± 21.04 83.97 ± 20.61 88.39 ± 21.81 0.547   ≥ 70% N 50 29 21 0.765  Subjects (available N) 55 34 21  DLCO, % predicted 70.0 ± 20.94 69.41 ± 19.44 70.94 ± 23.65 1   ≥ 50% N 46 29 17 0.719 Staging  Limited disease N (%) 33 (45.8) 19 (47.5) 14 (43.8) 1  Extensive disease N (%) 33 (45.8) 18 (45.0) 15 (46.9)  Unknown N (%) 6 (8.3) 3 (7.5) 3 (9.4) Bronchoalveolar lavage  Subjects (available N) 25 18 7  Total cells (×104 mL) 27.4 ± 19.08 27.45 ± 19.26 27.28 ± 20.39 0.685  CD4/ CD8 ratio 1.393 ± 1.422 1.141 ± 1.230 2.065 ± 1.790 0.042*  Lymphocytes, % 19.91 ± 13.74 17.69 ± 14.76 25.60 ± 9.22 0.115  Neutrophils, % 6.666 ± 6.470 7.319 ± 7.265 4.986 ± 3.654 0.785  Eosinophils, % 3.770 ± 4.710 4.340 ± 5.300 2.290 ± 2.360 0.543 Medication (during follow-up), N (%)  PAH-specific drug therapy use† 4 (5.6) 4 (10.0) 0 (0.0) 0.124  Steroid use 30 (41.7) 19 (47.5) 11 (34.4) 0.338  Cyclophosphamid use 6 (8.3) 6 (15.0) 0 (0.0) 0.03*  Cyclosporine or tacrolimus or azathioprine use 15 (20.8) 9 (22.5) 6 (18.8) 0.776  Pirfenidone use 5 (6.9) 1 (2.5) 4 (12.5) 0.164 Deaths N (%) 18 (25.0) 5 (12.5) 13 (40.6)  Cause of death N   Interstitial pneumonia 10 1 9   Bacterial pneumonia 2 2 0   Cardiac or cerebral infarction 3 0 3   Breast cancer 1 1 0   Unknown 2 1 1 Median follow-up years (range) 2.84 (0.15–17.25) 2.51 (0.20–17.25) 3.25 (0.15–9.75) Data are presented as mean ± SD, unless otherwise stated. SSc: systemic sclerosis; ILD: interstitial lung disease; Ab: autoantibody; Hb: hemoglobin; LDH: lactate dehydrogenase; CRP: C-reactive protein; SP-D: surfactant protein-D; FEV1: forced expiratory volume in 1 s; FVC: forced vital capacity; DLCO: diffusing capacity of the lung for carbon monoxide; PAH: pulmonary arterial hypertension. †PAH-specific drugs include beraprost sodium (N = 2) and sildenafil (N = 2). *P value less than 0.05. Radiographic features Radiographic features in SSc-ILD included UIP in 3 (7.5%) patients and NSIP in 32 (80.0%), and in ScAb-ILD included UIP in 7 (21.9%) patients and NSIP in 16 (50.0%), indicating that NSIP was a more frequent pattern in SSc-ILD, and the frequency of UIP was slightly more common in ScAb-ILD than in SSc-ILD (p = 0.024) (Table 2). Honeycombing was observed significantly more frequently in ScAb-ILD (p = 0.003). In contrast, cystic changes (non-honeycombing, emphysema) were significantly more frequent in SSc-ILD (p = 0.013) (Fig 1). 10.1371/journal.pone.0161908.t002Table 2 Comparison of HRCT and pathological findings between SSc-ILD and ScAb-ILD. Characteristics All subjects SSc-ILD ScAb-ILD P value No. of patients 72 40 32 HRCT pattern N (%)  Suggestive of UIP 10 (13.9) 3 (7.5) 7 (21.9) 0.024†*  Suggestive or consistent with NSIP 48 (66.7) 32 (80.0) 16 (50.0)  Others†† 14 (19.4) 5 (12.5) 9 (28.1) HRCT findings N (%)  GGO 65 (90.3) 38 (95.0) 27 (84.4) 0.23   Distribution (unilateral/ bilateral) 2 (2.8)/ 63 (87.5) 1 (2.5)/ 37 (92.5) 1 (3.1)/ 26 (81.3) 1    (upper/ lower/ diffuse or random) 1 (1.4)/ 49 (68.1)/ 15 (20.8) 0 (0.0)/ 32 (80.0)/ 6 (15.0) 1 (3.1)/ 17 (53.1)/ 9 (28.1) 0.097    (peribronchovascular/ subpleural/ diffuse) 14 (19.4)/ 21 (29.2)/ 30 (41.7) 7 (17.5)/ 15 (37.5)/ 16 (40.0) 7 (21.9)/ 6 (18.8)/ 14 (43.8) 0.333  Consolidation 17 (23.6) 8 (20.0) 9 (28.1) 0.577   Distribution (unilateral/ bilateral) 3 (4.2)/ 14 (19.4) 2 (5.0)/ 6 (15.0) 1 (3.1)/ 8 (25.0) 0.576    (upper/ lower/ diffuse or random) 4 (5.6)/ 12 (16.7)/ 1 (1.4) 1 (2.5)/ 6 (15.0)/ 1 (2.5) 3 (9.4)/ 6 (18.8)/ 0 (0.0) 0.576    (peribronchovascular/ subpleural/ diffuse) 2 (2.8)/ 8 (11.1)/ 7 (9.7) 2 (5.0)/ 1 (2.5)/ 5 (12.5) 0 (0.0)/ 7 (21.9)/ 2 (6.3) 0.016*  Reticulation 68 (94.4) 39 (97.5) 29 (90.6) 0.317   Distribution (unilateral/ bilateral) 3 (4.2)/ 65 (90.3) 1 (2.5)/ 39 (97.5) 2 (6.3)/ 27 (84.4) 0.571    (upper/ lower/ diffuse or random) 3 (4.2)/ 62 (86.1)/ 3 (4.2) 1 (2.5)/ 35 (87.5)/ 3 (7.5) 2 (6.3)/ 27 (84.4)/ 0 (0.0) 0.293    (peribronchovascular/ subpleural/ diffuse) 16 (22.2)/ 32 (44.4)/ 20 (27.8) 11 (27.5)/ 16 (40.0)/ 12 (30.0) 5 (15.6)/ 16 (50.0)/ 8 (25.0) 0.51  Honeycombing 15 (20.8) 3 (7.5) 12 (37.5) 0.003*  Traction bronchiectasis 61 (84.7) 32 (80.0) 29 (90.6) 0.325  Bronchial wall thickening 57 (79.2) 29 (72.5) 28 (87.5) 0.151  Micro-nodules 10 (13.9) 5 (12.5) 5 (15.6) 0.743  Emphysema 18 (25.0) 8 (20.0) 10 (31.3) 0.29  Cyst (non-honeycombing, emphysema) 17 (23.6) 14 (35.0) 3 (9.4) 0.013*  Mosaic attenuation (air trapping) 32 (44.4) 17 (42.5) 15 (46.9) 0.813  Enlarged mediastinal lymph node 14 (19.4) 8 (20.0) 6 (18.8) 1  Pleural thickening or effusion 7 (9.7) 4 (10.0) 3 (9.4) 1  Pulmonary artery dilatation 27 (37.5) 19 (47.5) 8 (25.0) 0.056  Volume loss 56 (77.8) 32 (80.0) 24 (75.0) 0.776 HRCT  Disease extent, (available N) 69 38 31  % 30.79 ± 15.327 28.638 ± 12.676 33.429 ± 17.923 0.379  ≥ 30% N (%) 32 (44.4) 17 (42.5) 15 (46.9) 0.812 Pathological pattern (available N) 32 25 7  UIP 1 (1.4) 1 (2.5) 0 (0.0) 0.085†  Fibrotic NSIP 20 (27.8) 18 (45.0) 2 (6.3)  Unclassifiable 11 (15.3) 6 (15.0) 5 (15.6) Pathological features, grade 0/ 1/ 2/ 3  Lung parenchyma lesion   Cellular infiltration 0/ 10/ 16/ 6 0/ 8/ 11/ 6 0/ 2/ 5/ 0 0.085   Plasma cell infiltration 0/ 10/ 13/ 9 0/ 8/ 9/ 8 0/ 2/ 4/1 0.698   Lymphoid follicle with germinal center 15/ 14/ 2/ 1 10/ 12/ 2/ 1 5/ 2/ 0/ 0 0.13   Fibrosis 0/ 4/ 17/ 11 0/ 3/ 13/ 9 0/ 1/ 4/ 2 0.742   Honeycombing 24/ 7/ 1/ 0 19/ 5/ 1/ 0 5/ 2/ 0/ 0 0.88   Fibroblastic foci 10/ 19/ 1/ 2 7/ 16/ 1/ 1 3/ 3/ 0/ 1 0.714   Organizing pneumonia (intra-alveolar polyp) 16/ 12/ 3/ 1 15/ 8/ 1/ 1 1/ 4/ 2/ 0 0.033*   Atelectasis (collapse) 3/ 12/ 9/ 8 3/ 9/ 6/ 7 0/ 3/ 3/ 1 0.962   Cyst formation 25/ 4/ 3/ 0 20/ 3/ 2/ 0 5/ 1/ 1/ 0 0.636  Airway lesion   Cellular infiltration 2/ 20/ 5/ 5 2/ 14/ 4/ 5 0/ 6/ 1/ 0 0.429   Lymphoid follicle 26/ 5/ 1/ 0 21/ 3/ 1/ 0 5/ 2/ 0/ 0 0.523   Fibrosis 29/ 3/ 0/ 0 22/ 3/ 0/ 0 7/ 0/ 0/ 0 1   Traction bronchiectasis 9/ 16/ 6/ 1 7/ 11/ 6/ 1 2/ 5/ 0/ 0 0.386  Vascular intimal or medial thickening 19/ 8/ 4/ 1 12/ 8/ 4/ 1 7/ 0/ 0/ 0 0.020*  Pleural fibrosis 8/ 21/ 3/ 0 6/ 18/ 1/ 0 2/ 3/ 2/ 0 0.514  Pleural inflammation 17/ 10/ 4/ 1 12/ 8/ 4/ 1 5/ 2/ 0/ 0 0.208  Smoking-related lesion   Emphysema 13/ 13/ 6/ 0 11/ 10/ 4/ 0 2/ 3/ 2/ 0 0.403   Respiratory bronchiolitis 27/ 4/ 1/ 0 23/ 2/ 0/ 0 4/ 2/ 1/ 0 0.025*   Bronchial metaplasia 9/ 14/ 7/ 2 8/ 9/ 6/ 2 1/ 5/ 1/ 0 0.981   DIP reaction 17/ 10/ 5/ 0 14/ 8/ 3/ 0 3/ 2/ 2/ 0 0.42 Data are presented as mean ± SD, unless otherwise stated. SSc: systemic sclerosis; ILD: interstitial lung disease; Ab: autoantibody; HRCT: high-resolution computed tomography; UIP: usual interstitial pneumonia; NSIP: nonspecific interstitial pneumonia; GGO: ground glass opacity; DIP: desquamative interstitial pneumonia. †In relation to HRCT pattern except for others. ††Others includes cases with pleuroparenchymal fibroelastosis (N = 4) and unclassifiable (N = 10). †P values were calculated except others. *P value less than 0.05 10.1371/journal.pone.0161908.g001Fig 1 HRCT scans of cyst formation. (A-B) HRCT scan demonstrates cyst formation (arrowheads) with pulmonary fibrosis, traction bronchiectasis, and architectural distortion in two patients with SSc-ILD. There is no continuity between the cysts and traction bronchiectasis. (C) HRCT scan shows cysts (arrowheads) with opacity separated from the pleura in a patient with MCTD-ILD. Pathologic features Of the 72 patients, 32 underwent surgical lung biopsy for diagnosis of ILD. Of the 25 patients with SSc-ILD, the major histologic pattern was fibrotic NSIP in 18 patients, UIP in 1 patient, and “unclassifiable” in 6 patients (Table 2). The histologic pattern in the 7 patients with ScAb-ILD included “unclassifiable” in 5 patients, fibrotic NSIP in 2 patients, and UIP in none. Organizing pneumonia (intra-alveolar polyps) and respiratory bronchiolitis were present in 40% and 8% of the patients with SSc-ILD and in 85.7% and 42.9%, respectively, of the patients with ScAb-ILD, indicating significantly more frequent findings of organizing pneumonia in ScAb-ILD (p = 0.033). The severity of vascular intimal or medial thickening in the patients with SSc-ILD was significantly higher than that in those with ScAb-ILD (p = 0.020). Typical examples of pathological features in each grade are shown in Fig 2. 10.1371/journal.pone.0161908.g002Fig 2 Examples of pathological scoring. (A-B) Typical imaging in each grade of organizing pneumonia (intra-alveolar polyps) as found in patients with ScAb-ILD (hematoxylin-eosin stain) ([A]: grade 1 and [B]: grade 3). (C-D) Vascular intimal or medial thickening as found in patients SSc-ILD (Elastica van Gieson stain) ([C]: grade 1 and [D] grade 3]. Mortality Death from any cause occurred in 18 patients (20.0%) over a median 2.84-year follow-up period (range: 0.15–17.25 years). The overall cumulative 5-year mortality rate was 24.4%, whereas that of the patients with SSc-ILD and with ScAb-ILD was 10.9% and 35.9%, respectively (Table 1). Prognostic factors of all-cause mortality Log-rank testing showed that subjects with ScAb-ILD had significantly worse survival than those with SSc-ILD (p = 0.011) (Fig 3A). Analysis of staging (limited versus extensive disease) of all subjects and those with SSc-ILD showed no significant differences (p = 0.120 and 0.338, respectively), whereas just for ScAb-ILD, the patients with extensive disease had worse survival than those with limited disease (p = 0.015) (Fig 3B). For serum KL-6 in just the SSc-ILD patients, those with KL-6 ≥ 1000 U/mL had worse survival than those with KL-6 < 1000 U/mL (p = 0.049) (Fig 3C). In the analysis of each autoantibody, no significant difference in survival was found between patients with anti-centromere, anti-scleroderma-70, and anti-U1 RNP antibody (Fig 3D). ScAb-ILD, age ≥ 65 years, CRP ≥ 1 mg/dL, and honeycombing on HRCT were significant predictors of mortality in univariate analysis. A multivariate Cox proportional hazard model showed only age ≥ 65 years and CRP ≥ 1 mg/dL to be negative prognostic factors (Table 3). 10.1371/journal.pone.0161908.g003Fig 3 Kaplan-Meier survival curves of all-cause mortality. Overall cumulative 5-year mortality was 24.4%. (A) Patients with ScAb-ILD (dotted line) had worse survival than those with SSc-ILD (dashed line) (p = 0.011). (B) In patients with ScAb-ILD, those with extensive disease (dashed line) had worse survival than those with limited disease (solid line) (p = 0.015). (C) In patients with SSc-ILD, those with KL-6 ≥ 1000 U/mL (dashed line) had worse survival than those with KL-6 < 1000 U/mL (solid line) (p = 0.049). (D) The survival curves for patients with each type of autoantibody were not significantly different (p = 0.905 for comparison between anti-scleroderma-70 and anti-U1 RNP antibody, p = 0.089 for comparison between anti-scleroderma-70 and anti-centromere antibody, and p = 0.137 for comparison between anti-U1 RNP and anti-centromere antibody). 10.1371/journal.pone.0161908.t003Table 3 Analysis of factors associated with mortality. Univariate Cox Regression Multivariate Cox Regression Characteristics Crude HR (95% CI) P value Adjusted HR (95% CI) P value Disease type  ScAb-ILD 3.973 (1.267–12.464) 0.018* 1.189 (0.337–4.199) 0.788 Autoantibody  Anti-scleroderma-70 antibody Reference  Anti-centromere antibody 2.649 (0.682–10.284) 0.159  Anti-U1 RNP antibody 1.142 (0.283–4.607) 0.852 Age, ≥ 65 years 8.067 (1.824–35.687) 0.006* 8.794 (1.878–41.186) 0.006* Ever-smoker 1.312 (0.512–3.357) 0.572 FVC, < 70% 1.115 (0.338–3.682) 0.858 DLCO, < 50% 1.277 (0.268–6.088) 0.759 CRP, ≥ 1 mg/dL 5.070 (1.874–13.717) 0.001* 7.917 (2.303–27.211) 0.001* KL-6, ≥ 1000 U/mL 1.031 (0.383–2.776) 0.952 Disease extent on HRCT, ≥ 30% 1.599 (0.608–4.205) 0.342 Extensive disease 2.410 (0.755–7.692) 0.137 UIP pattern on HRCT 1.017 (0.214–4.519) 0.983 Honeycombing on HRCT 4.301 (1.582–11.691) 0.004* 0.482 (0.087–2.651) 0.401 Ab: autoantibody; ILD: interstitial lung disease; CRP: C-reactive protein; FVC: forced vital capacity; DLCO: diffusing capacity of the lung for carbon monoxide; HRCT: high-resolution computed tomography; UIP: usual interstitial pneumonia. *P value less than 0.05. Disease behavior during follow-up Baseline FVC (intercept) in ScAb-ILD (mean, 2.574 L [95% confidence interval (CI): 2.313–2.835]) was significantly higher than that in SSc-ILD (mean, 2.191 L [95% CI: 1.974–2.407]) (p = 0.027). The declining slopes of FVC were not significantly different between the two groups (SSc-ILD: mean, -0.03979 L year-1 [95% CI: -0.05449 to -0.02509]; ScAb-ILD: mean, -0.03740 L year-1 [95% CI: -0.06446 to -0.01034] (p = 0.878) (Fig 4). 10.1371/journal.pone.0161908.g004Fig 4 Changes in forced vital capacity (FVC) during follow-up. (A) SSc-ILD (B) ScAb-ILD (C) Regression lines were calculated by solving the linear mixed-effects model. Baseline FVC (intercept) with ScAb-ILD (mean, 2.574 L [95% CI: 2.313–2.835]) was significantly higher than that with SSc-ILD (mean, 2.191 L [95% CI: 1.974–2.407]) (p = 0.027). The declining slopes of FVC between both groups were not significantly different (SSc-ILD: mean, -0.03979 L year-1 [95% CI: -0.05449 to -0.02509]; ScAb-ILD: mean, -0.03740 L year-1 [95% CI: -0.06446 to -0.01034] (p = 0.878). Discussion The present study clarified the differences in patient characteristics, disease behavior, and prognosis between SSc-ILD and ScAb-ILD and showed that lung-dominant CTD with ScAb-ILD is a different disease entity from that of SSc-ILD. Although analyses of SSc-ILD were often reported previously, to the best of our knowledge, the association between SSc-ILD and SSc-related autoantibody-positive ILD but not SSc indicates that lung-dominant CTD has not been reported before. SSc is a CTD characterized by tissue fibrosis in the skin and internal organs. ILD develops in more than half of the patients with SSc and is an important risk factor of mortality as is pulmonary hypertension [32, 33]. As in previous studies, our patients with SSc-ILD were more likely to be females and non-smokers, and most cases were diagnosed as NSIP through both HRCT and pathological analysis [34]. In contrast, about half of the patients with ScAb-ILD were male and current or ex-smokers. In HRCT analysis, although one-half of the patients with ScAb-ILD were diagnosed as having NSIP, over one-fifth of them had UIP, and honeycombing was more predominant in those with ScAb-ILD rather than SSc-ILD. Interestingly, HRCT findings of cysts were more likely to be present in those with SSc-ILD rather than ScAb-ILD. Certain segments of the population have secondary Sjögren's syndrome associated with SSc, and thin-walled cysts are often seen in Sjögren's syndrome with ILD [35]. In our study, two patients were diagnosed as having secondary Sjögren's syndrome with SSc. The precise frequency of such patients may actually be higher in our study and thus might lead to significantly different results. Pathological analysis showed the severity of vascular intimal or medial thickening to be significantly higher in the patients with SSc-ILD than ScAb-ILD. Although only a few patients underwent specific drug therapy for pulmonary arterial hypertension in our study, the primary findings in SSc patients are intimal fibrosis (affecting the small vessels adjacent to the alveoli), medial hyperplasia, and adventitial fibrosis affecting the pulmonary arterioles; thus, our results were compatible with these findings [36]. In contrast, none of the patients with ScAb-ILD had such lesions. Moreover, the histologic pattern of most patients (71.4%) with ScAb-ILD who underwent surgical lung biopsy was “unclassifiable.” The high frequency of smoking history in the ScAb-ILD patients and the existence of the autoantibodies themselves may contribute to this result, but the exact cause remains unclear. The survival curves showed that patients with ScAb-ILD had a significantly poorer outcome than those with SSc-ILD. Although the ScAb-ILD patients with extensive disease had a significantly poorer survival curve than those with limited disease, there was no significant difference in prognosis between the SSc-ILD patients with limited versus extensive disease. A previous study reported that the limited/extensive staging system strongly predicted mortality [28]. The SSc-ILD patients in our study did not show the same result, and the small sample size might be one reason for this finding. The SSc-ILD patients, and not the ScAb-ILD patients, with the biomarker of serum KL-6 ≥ 1000 U/mL had a worse survival curve than those with KL-6 < 1000 U/mL. A serum KL-6 of ≥ 1000 U/mL was previously reported to be a predictor of poor prognosis in patients with idiopathic pulmonary fibrosis [37], and in SSc-ILD, the presence of elevated KL-6 values is also a poor prognostic factor [38]. When disease behavior was assessed on the basis of FVC, the declining slope of FVC was not significantly different between the two groups. SSc-ILD was reported to progress much more frequently in the first 4 years of systemic disease (especially in the first 2 years) [39], but similar results were not seen in our study. Interestingly, despite the severity of vascular intimal and medial thickening in the pathological findings of the patients with SSc-ILD and the baseline FVC calculated by solving the linear mixed-effects model being significantly higher in those with ScAb-ILD, the patients with ScAb-ILD appeared to have a worse prognosis. Disease behavior related to the declining slope of FVC was not significantly different between the two groups, which indicates that there was missing data during follow-up period because of the poor prognosis of the patients with ScAb-ILD. To summarize, there were many differences between the baseline characteristics and HRCT and pathological findings of the two disease entities, and primarily, the prognosis of the patients with ScAb-ILD was poorer than that of the patients with SSc-ILD. All of our subjects fulfilled the criteria for lung-dominant CTD proposed by Fischer et al. [1]. Most of the subjects with ScAb-ILD matched the diagnostic criteria for IPAF because these patients have a serologic domain (anti-centromere, anti-scleroderma-70, and anti-U1 RNP antibodies) and morphologic domain as indicated predominantly by radiological findings of NSIP, intrinsic airway disease with bronchial wall thickening and air trapping, and pathological findings, and by cellular infiltration of the airway and organizing pneumonia. Therefore, in our investigation of these many aspects, we suggest that patients with ILD who have not met the diagnosis of SSc but are positive for SSc-related autoantibodies have a disease entity distinct from that of SSc-ILD, and this is an important reason for the difference in some reports between the prognosis of CTD-ILD versus that of UCTD-ILD, lung-dominant CTD, autoimmune-featured ILD, and IPAF. The limitations of this study are as follows. First, the study involved a relatively small number of patients from a single center. However, in our institution, chest clinicians carry out screening while always keeping routine examination of each CTD-related autoantibody in mind, even if the patients with ILD have no symptoms suspicious of CTD. Moreover, if the patients have autoantibodies related to CTD, we usually consult the rheumatologists to determine whether the diagnosis of CTD can be fulfilled. Therefore, we could examine the patients in greater detail. Second, the patients with MCTD were included as patients with SSc in this study. As mentioned in the Study sample section, anti-U1 RNP antibody is occasionally observed in patients with SSc and systemic lupus erythematosus, and chronic interstitial pneumonitis as a complication of systemic lupus erythematosus is rare [14, 40]. Moreover, most patients with MCTD have characteristics of SSc and ultimately develop SSc [16–18]. In fact, 7 (63.6%) of the 11 patients with MCTD fulfilled the criteria for SSc during the follow-up period. Therefore, ILD with MCTD is included in SSc-ILD, and ILD with positive U1 RNP antibody is included in ScAb-ILD. We conclude that despite these limitations, our study suggests that lung-dominant CTD and IPAF with SSc-related autoantibody is a different disease entity from that of SSc-ILD. We believe that the results of our study will be helpful in determining whether the management of IPAF should be similar to that of CTD-ILD. Further studies with each of the specific autoantibodies of IPAF may be necessary to assess the appropriate strategy to diagnose and treat IPAF. We offer our sincerest thanks to Drs. Shigeru Komatsu, Takeshi Shinohara, and Shinko Sadoyama of the Department of Respiratory Medicine, Kanagawa Cardiovascular and Respiratory Center, for their handling of the diagnosis and treatment of the patients with SSc-ILD and ScAb-ILD, and to Hideyo Oda of Medical Toukei Co., Ltd. for his advice on statistical analysis. All work was performed at the Kanagawa Cardiovascular and Respiratory Center. ==== Refs References 1 Fischer A , West SG , Swigris JJ , Brown KK , du Bois RM . Connective tissue disease-associated interstitial lung disease: a call for clarification . CHEST 2010 ; 138 :251 –256 . 10.1378/chest.10-0194 20682528 2 Kinder BW , Collard HR , Koth L , Daikh DI , Wolters PJ , Elicker B , et al Idiopathic nonspecific interstitial pneumonia: lung manifestation of undifferentiated connective tissue disease? Am J Respir Crit Care Med 2007 ; 176 :691 –697 . 17556720 3 Corte TJ , Copley SJ , Desai SR , Zappala CJ , Hansell DM , Nicholson AG , et al Significance of connective tissue disease features in idiopathic interstitial pneumonia . Eur Respir J 2012 ; 39 :661 –668 . 10.1183/09031936.00174910 21920896 4 Vij R , Noth I , Strek ME . Autoimmune-featured interstitial lung disease: a distinct entity . Chest 2011 ; 140 :1292 –1299 . 10.1378/chest.10-2662 21565966 5 Kim HC , Ji W , Kim MY , Colby TV , Jang SJ , Lee CK , et al Interstitial pneumonia related to undifferentiated connective tissue disease: pathologic pattern and prognosis . Chest 2015 ; 147 :165 –172 . 10.1378/chest.14-0272 25211168 6 Nunes H , Schubel K , Piver D , Magois E , Feuillet S , Uzunhan Y , et al Nonspecific interstitial pneumonia: survival is influenced by the underlying cause . Eur Respir J 2015 ; 45 :746 –755 . 10.1183/09031936.00148613 25537566 7 Omote N , Taniguchi H , Kondoh Y , Watanabe N , Sakamoto K , Kimura T , et al Lung-Dominant Connective Tissue Disease: Clinical, Radiologic, and Histologic Features . Chest 2015 ; 148 :1438 –1446 . 10.1378/chest.14-3174 25950648 8 Fischer A , Antoniou KM , Brown KK , Cadranel J , Corte TJ , du Bois RM , et al An official European Respiratory Society/American Thoracic Society research statement: interstitial pneumonia with autoimmune features . Eur Respir J 2015 ; 46 :976 –987 . 10.1183/13993003.00150-2015 26160873 9 Fischer A , du Bois R . Interstitial lung disease in connective tissue disorders . Lancet 2012 ; 380 :689 –698 . 10.1016/S0140-6736(12)61079-4 22901890 10 Gutsche M , Rosen GD , Swigris JJ . Connective Tissue Disease-associated Interstitial Lung Disease: A review . Curr Respir Care Rep 2012 ; 1 :224 –232 . 23125954 11 Fischer A , Solomon JJ , du Bois RM , Deane KD , Olson AL , Fernandez-Perez ER , et al Lung disease with anti-CCP antibodies but not rheumatoid arthritis or connective tissue disease . Respir Med 2012 ; 106 :1040 –1047 . 10.1016/j.rmed.2012.03.006 22503074 12 Takato H , Waseda Y , Watanabe S , Inuzuka K , Katayama N , Ichikawa Y , et al Pulmonary manifestations of anti-ARS antibody positive interstitial pneumonia—with or without PM/DM . Respir Med 2013 ; 107 :128 –133 . 10.1016/j.rmed.2012.09.005 23137883 13 Oldham JM , Adegunsoye A , Valenzi E , Lee C , Witt L , Chen L , et al Characterisation of patients with interstitial pneumonia with autoimmune features . Eur Respir J 2016 ; 47 :1767 –1775 . 10.1183/13993003.01565-2015 27103387 14 van den Hoogen F , Khanna D , Fransen J , Johnson SR , Baron M , Tyndall A , et al 2013 classification criteria for systemic sclerosis: an American College of Rheumatology/European League against Rheumatism collaborative initiative . Arthritis Rheum 2013 ; 65 :2737 –2747 . 10.1002/art.38098 24122180 15 Sharp GC , Irvin WS , May CM , Holman HR , McDuffie FC , Hess EV , et al Association of antibodies to ribonucleoprotein and Sm antigens with mixed connective-tissue disease, systematic lupus erythematosus and other rheumatic diseases . N Engl J Med 1976 ; 295 :1149 –1154 . 1086429 16 Nimelstein SH , Brody S , McShane D , Holman HR . Mixed connective tissue disease: a subsequent evaluation of the original 25 patients . Medicine (Baltimore) 1980 ; 59 :239 –248 .6967141 17 Smolen JS , Steiner G . Mixed connective tissue disease: to be or not to be? Arthritis Rheum 1998 ; 41 :768 –777 . 9588728 18 Aringer M , Steiner G , Smolen JS . Does mixed connective tissue disease exist? Yes . Rheum Dis Clin North Am 2005 ; 31 :411 –420 , v 16084315 19 Cappelli S , Bellando Randone S , Martinović D , Tamas MM , Pasalić K , et al "To be or not to be," ten years after: evidence for mixed connective tissue disease as a distinct entity . Semin Arthritis Rheum 2012 ; 41 :589 –598 . 10.1016/j.semarthrit.2011.07.010 21959290 20 American Thoracic Society ; European Respiratory Society . American Thoracic Society/European Respiratory Society International Multidisciplinary Consensus Classification of the Idiopathic Interstitial Pneumonias. This joint statement of the American Thoracic Society (ATS), and the European Respiratory Society (ERS) was adopted by the ATS board of directors, June 2001 and by the ERS Executive Committee, June 2001 . Am J Respir Crit Care Med 2002 ; 165 :277 –304 . 11790668 21 Travis WD , Costabel U , Hansell DM , King TE Jr, Lynch DA , Nicholson AG , et al An official American Thoracic Society/European Respiratory Society statement: Update of the international multidisciplinary classification of the idiopathic interstitial pneumonias . Am J Respir Crit Care Med 2013 ; 188 :733 –748 . 10.1164/rccm.201308-1483ST 24032382 22 Reddy TL , Tominaga M , Hansell DM , von der Thusen J , Rassl D , Parfrey H , et al Pleuroparenchymal fibroelastosis: a spectrum of histopathological and imaging phenotypes . Eur Respir J 2012 ; 40 :377 –385 . 10.1183/09031936.00165111 22441748 23 Hansell DM , Bankier AA , MacMahon H , McLoud TC , Müller NL , Remy J . Fleischner Society: glossary of terms for thoracic imaging . Radiology 2008 ; 246 :697 –722 . 10.1148/radiol.2462070712 18195376 24 Lynch DA , Godwin JD , Safrin S , Starko KM , Hormel P , Brown KK , et al High-resolution computed tomography in idiopathic pulmonary fibrosis: diagnosis and prognosis . Am J Respir Crit Care Med 2005 ; 172 :488 –493 . 15894598 25 Best AC , Meng J , Lynch AM , Bozic CM , Miller D , Grunwald GK , et al Idiopathic pulmonary fibrosis: physiologic tests, quantitative CT indexes, and CT visual scores as predictors of mortality . Radiology 2008 ; 246 :935 –940 . 10.1148/radiol.2463062200 18235106 26 Iwasawa T , Kato S , Ogura T , Kusakawa Y , Iso S , Baba T , et al Low-normal lung volume correlates with pulmonary hypertension in fibrotic idiopathic interstitial pneumonia: computer-aided 3D quantitative analysis of chest CT . AJR Am J Roentgenol 2014 ; 203 :W166 –W173 . 10.2214/AJR.13.11409 25055292 27 Iwasawa T , Kanauchi T , Hoshi T , Ogura T , Baba T , Gotoh T , et al Multicenter study of quantitative computed tomography analysis using a computer-aided three-dimensional system in patients with idiopathic pulmonary fibrosis . Jpn J Radiol 2016 ; 34 :16 –27 . 10.1007/s11604-015-0496-0 26546034 28 Goh NS , Desai SR , Veeraraghavan S , Hansell DM , Copley SJ , Maher TM , et al Interstitial lung disease in systemic sclerosis: a simple staging system . Am J Respir Crit Care Med 2008 ; 177 :1248 –1254 . 10.1164/rccm.200706-877OC 18369202 29 Flaherty KR , Colby TV , Travis WD , Toews GB , Mumford J , Murray S , et al Fibroblastic foci in usual interstitial pneumonia: idiopathic versus collagen vascular disease . Am J Respir Crit Care Med 2003 ; 167 :1410 –1415 . 12615630 30 Song JW , Do KH , Kim MY , Jang SJ , Colby TV , Kim DS . Pathologic and radiologic differences between idiopathic and collagen vascular disease-related usual interstitial pneumonia . Chest 2009 ; 136 :23 –30 . 10.1378/chest.08-2572 19255290 31 Enomoto Y , Takemura T , Hagiwara E , Iwasawa T , Okudela K , Yanagawa N , et al Features of usual interstitial pneumonia in patients with primary Sjögren’s syndrome compared with idiopathic pulmonary fibrosis . Respir Investig 2014 ; 52 :227 –235 . 10.1016/j.resinv.2014.02.003 24998369 32 Silver RM . Scleroderma. Clinical problems. The lungs . Rheum Dis Clin North Am 1996 ; 22 :825 –840 . 8923598 33 Hasegawa M , Asano Y , Endo H , Fujimoto M , Goto D , Ihn H , et al Investigation of prognostic factors for skin sclerosis and lung function in Japanese patients with early systemic sclerosis: a multicentre prospective observational study . Rheumatology (Oxford) 2012 ; 51 :129 –133 .22072085 34 Hoffmann-Vold AM , Aaløkken TM , Lund MB , Garen T , Midtvedt Ø , Brunborg C , et al Predictive value of serial high-resolution computed tomography analyses and concurrent lung function tests in systemic sclerosis . Arthritis Rheumatol 2015 ; 67 :2205 –2212 . 10.1002/art.39166 25916462 35 Fox RI . Sjögren's syndrome . Lancet 2005 ; 366 :321 –331 . 16039337 36 Solomon JJ , Olson AL , Fischer A , Bull T , Brown KK , Raghu G . Scleroderma lung disease . Eur Respir Rev 2013 ; 22 :6 –19 . 10.1183/09059180.00005512 23457159 37 Yokoyama A , Kondo K , Nakajima M , Matsushima T , Takahashi T , Nishimura M , et al Prognostic value of circulating KL-6 in idiopathic pulmonary fibrosis . Respirology 2006 ; 11 :164 –168 . 16548901 38 Winstone TA , Assayag D , Wilcox PG , Dunne JV , Hague CJ , Leipsic J , et al Predictors of mortality and progression in scleroderma-associated interstitial lung disease: a systematic review . Chest 2014 ; 146 :422 –436 . 10.1378/chest.13-2626 24576924 39 Wells AU . Interstitial lung disease in systemic sclerosis . Presse Med 2014 ; 43 :e329 –e343 . 10.1016/j.lpm.2014.08.002 25217474 40 Keane MP , Lynch JP 3rd. Pleuropulmonary manifestations of systemic lupus erythematosus . Thorax 2000 ; 55 :159 –166 . 10639536

PMC005xxxxxx/PMC5001712.txt

==== Front PLoS GenetPLoS GenetplosplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, CA USA 2756439410.1371/journal.pgen.1006258PGENETICS-D-16-00654Research ArticleBiology and Life SciencesGeneticsGene ExpressionMedicine and Health SciencesInfectious DiseasesBacterial DiseasesSalmonellaSalmonella TyphimuriumBiology and Life SciencesMicrobiologyMedical MicrobiologyMicrobial PathogensBacterial PathogensSalmonellaSalmonella TyphimuriumMedicine and Health SciencesPathology and Laboratory MedicinePathogensMicrobial PathogensBacterial PathogensSalmonellaSalmonella TyphimuriumBiology and Life SciencesOrganismsBacteriaEnterobacteriaceaeSalmonellaSalmonella TyphimuriumBiology and Life SciencesGeneticsGene ExpressionGene RegulationBiology and Life SciencesGeneticsGene TypesRegulator GenesBiology and Life SciencesGeneticsGene ExpressionGene RegulationPost-Transcriptional Gene RegulationBiology and Life SciencesGeneticsGene ExpressionGene RegulationRegulonsBiology and life sciencesGeneticsGene expressionDNA transcriptionBiology and life sciencesBiochemistryProteinsDNA-binding proteinsTranscription FactorsBiology and Life SciencesGeneticsGene ExpressionGene RegulationTranscription FactorsBiology and Life SciencesBiochemistryProteinsRegulatory ProteinsTranscription FactorsThe Impact of 18 Ancestral and Horizontally-Acquired Regulatory Proteins upon the Transcriptome and sRNA Landscape of Salmonella enterica serovar Typhimurium Regulatory Landscape of Salmonella TyphimuriumColgan Aoife M. 1http://orcid.org/0000-0003-0461-1530Kröger Carsten 1Diard Médéric 2Hardt Wolf-Dietrich 2Puente José L. 3Sivasankaran Sathesh K. 1Hokamp Karsten 4Hinton Jay C. D. 15*1 Department of Microbiology, School of Genetics and Microbiology, Moyne Institute of Preventive Medicine, Trinity College, Dublin, Ireland2 Institute of Microbiology, ETH Zürich, Zürich, Switzerland3 Departamento de Microbiología Molecular, Instituto de Biotecnología, Universidad Nacional Autónoma de Mexico, Cuernavaca, Morelos, Mexico4 Department of Genetics, School of Genetics and Microbiology, Smurfit Institute of Genetics, Trinity College, Dublin, Ireland5 Institute of Integrative Biology, University of Liverpool, Liverpool, United KingdomCasadesús Josep EditorUniversidad de Sevilla, SPAINThe authors have declared that no competing interests exist. Conceived and designed the experiments: AMC CK MD WDH JLP JCDH. Performed the experiments: AMC CK. Analyzed the data: AMC CK JLP JCDH. Contributed reagents/materials/analysis tools: MD WDH SKS KH. Wrote the paper: AMC CK MD WDH JLP SKS KH JCDH. * E-mail: jay.hinton@liverpool.ac.uk26 8 2016 8 2016 12 8 e100625821 3 2016 25 7 2016 © 2016 Colgan et al2016Colgan et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.We know a great deal about the genes used by the model pathogen Salmonella enterica serovar Typhimurium to cause disease, but less about global gene regulation. New tools for studying transcripts at the single nucleotide level now offer an unparalleled opportunity to understand the bacterial transcriptome, and expression of the small RNAs (sRNA) and coding genes responsible for the establishment of infection. Here, we define the transcriptomes of 18 mutants lacking virulence-related global regulatory systems that modulate the expression of the SPI1 and SPI2 Type 3 secretion systems of S. Typhimurium strain 4/74. Using infection-relevant growth conditions, we identified a total of 1257 coding genes that are controlled by one or more regulatory system, including a sub-class of genes that reflect a new level of cross-talk between SPI1 and SPI2. We directly compared the roles played by the major transcriptional regulators in the expression of sRNAs, and discovered that the RpoS (σ38) sigma factor modulates the expression of 23% of sRNAs, many more than other regulatory systems. The impact of the RNA chaperone Hfq upon the steady state levels of 280 sRNA transcripts is described, and we found 13 sRNAs that are co-regulated with SPI1 and SPI2 virulence genes. We report the first example of an sRNA, STnc1480, that is subject to silencing by H-NS and subsequent counter-silencing by PhoP and SlyA. The data for these 18 regulatory systems is now available to the bacterial research community in a user-friendly online resource, SalComRegulon. Author Summary The transcriptional networks and the functions of small regulatory RNAs of Salmonella enterica serovar Typhimurium are being studied intensively. S. Typhimurium is becoming the ideal model pathogen for linking transcriptional and post-transcriptional gene regulation to bacterial virulence. Here, we systematically defined the regulatory factors responsible for controlling the expression of S. Typhimurium coding genes and sRNAs under infection-relevant growth conditions. As well as confirming published regulatory inputs for Salmonella pathogenicity islands, such as the positive role played by Fur in the expression of SPI1, we report, for the first time, the global impact of the FliZ, HilE and PhoB/R transcription factors and identify 124 sRNAs that belong to virulence-associated regulons. We found a subset of genes of known and unknown function that are regulated by both HilD and SsrB, highlighting the cross-talk mechanisms that control Salmonella virulence. An integrative analysis of the regulatory datasets revealed 5 coding genes of unknown function that may play novel roles in virulence. We hope that the SalComRegulon resource will be a dynamic database that will be constantly updated to inspire new hypothesis-driven experimentation, and will contribute to the construction of a comprehensive transcriptional network for S. Typhimurium. http://dx.doi.org/10.13039/501100001602Science Foundation Ireland08/IN.1/B2104Hinton Jay C. D. http://dx.doi.org/10.13039/501100001602Science Foundation Ireland07/IN.1/B918Hinton Jay C. D. This work was supported by the Stokes Professorial Fellowship and Principal Investigator grant awarded to JCDH by Science Foundation Ireland (Ref 08/IN.1/B2104 and 07/IN.1/B918; www.sfi.ie). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityThe RNA-seq data generated from this study are deposited at NCBI GEO under the accession numbers GSM2091439 to GSM2091466, and can be accessed at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE79314.Data Availability The RNA-seq data generated from this study are deposited at NCBI GEO under the accession numbers GSM2091439 to GSM2091466, and can be accessed at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE79314. ==== Body Introduction Salmonella enterica serovar Typhimurium (S. Typhimurium) is an important foodborne pathogen that causes self-limiting gastroenteritis, or more serious systemic infections in susceptible hosts. In the developed world, there are an estimated 93.8 million incidences of salmonellosis caused by non-typhoidal Salmonella (NTS) strains, resulting in 155,000 deaths each year [1]. In developing countries, NTS strains cause bloodstream infections that kill about 20% of patients. This high mortality rate reflects the combination of pre-disposing conditions such as HIV, malaria and malnutrition, and the emergence of invasive NTS strains [2, 3]. S. Typhimurium colonises a wide range of mammals and birds, and encounters a series of stressful conditions within various host environments. The bacteria express a Type III Secretion System (T3SS) encoded on a pathogenicity island (SPI1) that mediates invasion of the host intestinal epithelium. Once internalised, S. Typhimurium expresses a second T3SS, encoded on a second pathogenicity island (SPI2), which is responsible for its survival and replication in the intracellular environment within the Salmonella containing vacuole (SCV) and for the establishment of systemic infection [4, 5]. Salmonella has the ability to rapidly remodel its gene expression profile when exposed to different spatial and temporal cues. This co-ordinated transcriptional programme allows S. Typhimurium to interact with the microbiota and the mammalian host, to multiply and survive in the host intestine and to cause systemic disease [6]. For Salmonella bacteria, infection involves the generation of genetically-identical cells with different but cooperating phenotypes to ensure fitness and optimal spatio-temporal gene expression [7–12]. RNA-seq analysis of individual host cells has revealed that the gene expression profile of infecting bacterial cells can have profound effects on the resulting host response [13]. Such a flexible bacterial gene expression program must be tightly controlled by the dynamic interactions of hundreds of transcriptional regulators. S. Typhimurium encodes 156 protein factors which act at the level of the initiation of transcription [14]. In addition, gene expression is regulated at the post-transcriptional level, largely by trans-acting small regulatory RNAs (sRNA) [15]. Trans-acting sRNAs usually affect expression of target genes through short imperfect base-pairing interactions, and often require the RNA chaperone, Hfq [16]. We require an integrated understanding of the regulatory inputs that coordinate the expression of all RNA transcripts of Salmonella. Here, we explore the interconnections between transcriptional and post-transcriptional regulation of S. Typhimurium protein coding genes and sRNAs using bacterial mutants that lack key components of the global transcriptional networks that are associated with SPI1 and SPI2 expression. We used an RNA-seq-based transcriptomic approach to explore the regulons of 18 virulence-associated regulatory systems including sigma factors, transcription factors, two-component systems and an RNA chaperone, under in vitro conditions. Our particular focus was the regulation of sRNA gene expression and we identified 124 sRNA genes that are controlled by at least one of the regulatory systems tested in this work. Molecular and computational approaches were used to validate regulatory interactions. We showed that the putative virulence-associated sRNA, STnc1480, is silenced by the nucleoid-associated protein H-NS, and that the transcription factors PhoP and SlyA counteract the repressive effects of H-NS at the STnc1480 promoter. Finally, we describe 13 sRNAs which we predict to play important roles in S. Typhimurium virulence, based on their patterns of regulation. We assembled an online compendium of our RNA-seq-based transcriptomic analysis of the regulons of 18 systems that control S. Typhimurium virulence, as a community resource. One important caveat is that these regulons contain genes that are both directly and indirectly regulated. Further experiments will be required to identify the binding sites for each transcription factor across the chromosome. This investigation of the regulatory inputs to the expression of S. Typhimurium coding genes and sRNAs significantly extends current knowledge about the interconnections between transcriptional and post-transcriptional regulatory elements, and is a step towards the elucidation of the topology of regulatory networks that control S. Typhimurium pathogenesis. Results and Discussion RNA-seq analysis of regulatory mutants of S. Typhimurium under infection-relevant growth conditions Previously, we profiled the transcriptome of wild-type S. Typhimurium 4/74 under 22 environmental conditions [9]. We discovered that S. Typhimurium possesses 280 sRNAs and that the expression of each sRNA was highly dynamic and environmentally-responsive [9]. However, little was known about the protein factors responsible for modulating sRNA expression. Here we investigated global transcriptional and post-transcriptional regulators of Salmonella gene expression to identify regulatory inputs that control the dynamic expression of S. Typhimurium sRNAs. We selected a panel of 18 virulence-associated S. Typhimurium regulatory proteins, and disrupted their expression by generating a set of isogenic deletion mutants (Tables 1 and 2; Materials and Methods). Our published transcriptomic and transcriptional start site (TSS) data from wild-type strain 4/74 [9] were used to investigate whether the expression of downstream genes was affected by the genetic constructions used to delete the individual regulatory genes. Typically, the entire coding regions of 18 regulatory genes were removed. To avoid polar effects, resistance cassettes were removed from mutant strains if the deleted gene had its own TSS and was in the 5’ region of an operon. For the sirA mutation, a 432 bp region at the 5’ end of the sirA coding sequence (CDS) was deleted to maintain an intact uvrC TSS. For the slyA mutation, the 5’ and 3’ ends of the slyA CDS were left intact to avoid disruption of anmK expression and deletion of the overlapping 3’ end of the slyB CDS on the opposite strand. 10.1371/journal.pgen.1006258.t001Table 1 Bacterial strains and plasmids. Name Genotype Reference Bacterial strains Wild-type prototroph S. Typhimurium 4/74. StrR [88] JH3630 S. Typhimurium 4/74 ΔrpoE::frt P22 transduction from JVS-1028 JH3632 S. Typhimurium 4/74 Δfur::frt P22 transduction from JH3305 JH3635 S. Typhimurium 4/74 ΔhilD::frt P22 transduction from JVS-5924 JH3636 S. Typhimurium 4/74 ΔslyA::kan This Study JH3637 S. Typhimurium 4/74 ΔphoB/R::kan P22 transduction from JH3312 JH3643 S. Typhimurium 4/74 ΔfliZ::frt This Study JH3652 S. Typhimurium 4/74 ΔompR/envZ::frt P22 transduction from [99] JH3653 S. Typhimurium 4/74 Δdam::frt P22 transduction from [37] JH3657 S. Typhimurium 4/74 ΔssrA/B::frt This Study JH3675 S. Typhimurium 4/74 ΔssrA::frt This Study JH3733 S. Typhimurium 4/74 ΔssrB::frt This Study JH3660 S. Typhimurium 4/74 ΔphoP/Q::frt This Study JH3763 S. Typhimurium 4/74 ΔphoP::frt P22 transduction from JVS-8781 JH3673 S. Typhimurium 4/74 ΔbarA/sirA-432::kan This Study JH3674 S. Typhimurium 4/74 ΔrpoS::kan P22 transduction from JH3575 JH3584 S. Typhimurium 4/74 Δhfq::kan P22 transduction from JVS-00225 JH3765 S. Typhimurium 4/74 ΔhilA::frt P22 transduction from JVS-1195 JH3766 S. Typhimurium 4/74 ΔhilC::cat JH3767 S. Typhimurium 4/74 ΔhilE::cat P22 transduction from M2008 [11] JH3774 S. Typhimurium 4/74 hns-1::kan [100] JH3775 S. Typhimurium 4/74 phoP-3xFLAG (C-terminal) This Study JH3777 S. Typhimurium 4/74 hns-3xFLAG::kan (C-terminal) [71] JH3782 S. Typhimurium 4/74 ΔslyA hns-1::kan This Study JH3783 S. Typhimurium 4/74 ΔphoP hns-1::kan This Study TOP10 E. coli TOP10 F- mcrA Δ(mrr-hsdRMS-mcrBC) φ80lacZΔM15 ΔlacX74 nupG recA1 araD139 Δ(ara-leu)7697 galE15 galK16 rpsL(StrR) endA1 λ- Invitrogen Plasmids pKD3 Template plasmid for gene deletion; AmpR, FRT-flanked cat gene [89] pKD4 Template plasmid for gene deletion; AmpR, FRT-flanked kan gene [89] pKD46 Red helper plasmid; repA101 (ts), pBAD-γ-β-exo, araC+, AmpR [89] pCP20 Plasmid for temperature sensitive FLP synthesis; ts-rep, FLP, CmR, AmpR [91] pSUB11 Template plasmid for epitope tagging; AmpR, FRT-flanked 3xFLAG tag and kan gene [90] pBAD-myc-hisA pBR322 origin, araC+, araBAD, C-terminal myc and polyhistidine tags, AmpR Invitrogen pBAD-slyA slyA gene cloned in MCS of pBAD-myc-hisA This Study pBAD-phoP phoP gene cloned in MCS of pBAD-myc-hisA This Study Antibiotics were used at the following final concentrations: Ampicillin (Amp) 100 μg/mL, Kanamycin (Kan) 50 μg/mL, Chloramphenicol (Cm) 35 μg/mL. 10.1371/journal.pgen.1006258.t002Table 2 Panel of regulatory mutants and environmental conditions used for RNA-seq experiments. Strain Function of Regulator Growth Condition Description of Growth Rationale for Chosen Growth Condition wild-type Mid exponential phase (MEP) growth to OD6000.3 in Lennox broth Exponentially growing cultures contain hemi-methylated DNA sites for targeting by the Dam methylase. Δdam DNA adenine methyltransferase protein. Targets GATC sites on hemi-methylated DNA, important for initiation of chromosome replication and mismatch DNA repair [101]. Dam activates SPI1 genes through post-transcriptional control of HilD [38]. Mid exponential phase (MEP) growth to OD6000.3 in Lennox broth wild-type Intermediate exponential phase (IEP) growth to OD6000.8 in Lennox broth S. Typhimurium grown to OD6000.8 in Lennox broth expresses SPI1 in a bistable fashion and the competitive growth disadvantage between cells expressing SPI1 and their non-SPI1-expressing siblings is greatest at this time point [11]. ΔhilD SPI1-encoded HilD is the primary regulator of genes necessary for the expression of the SPI1 T3SS, through formation of a feed-forward regulatory loop with HilC and RtsA [102]. Intermediate exponential phase (IEP) growth to OD6000.8 in Lennox broth ΔhilC SPI1-encoded HilC forms part of a feed-forward regulatory loop necessary for the expression of the SPI1 T3SS [102]. Intermediate exponential phase (IEP) growth to OD6000.8 in Lennox broth ΔhilA SPI1-encoded HilA is the central regulator of SPI1 T3SS genes, either directly or indirectly through control of another SPI1-encoded TF, InvF [103]. Intermediate exponential phase (IEP) growth to OD6000.8 in Lennox broth ΔhilE The hilE gene is transcriptionally activated by the regulator of fimbrial gene expression, FimZ [29], and HilE interacts with the HilD protein to repress hilA transcription [52]. Intermediate exponential phase (IEP) growth to OD6000.8 in Lennox broth wild-type Early stationary phase (ESP) growth to OD6002.0 in Lennox broth We have previously shown that growth to ESP induces expression of genes encoding the components of the SPI1 T3SS [9]. Δfur Fur is the ferric uptake regulator, responsible for maintaining cellular iron homeostasis. In iron-replete conditions Fe2+-bound Fur binds to promoters of target genes, involved in acquisition and transport of iron, and blocks their transcription. When iron levels are low, Fur dissociates from the ferrous ion and the target DNA, allowing target gene transcription [104]. Fur activates hilA expression through transcriptional repression of the hns gene. H-NS normally silences hilA expression and is counter-silenced by HilD at the hilA promoter [20]. Early stationary phase (ESP) growth to 0D6002.0 in Lennox broth ΔhilD SPI1-encoded HilD is the primary regulator of all genes necessary for the expression of the SPI1 T3SS, through formation of a feed-forward regulatory loop with HilC and RtsA [102]. Early stationary phase (ESP) growth to OD6002.0 in Lennox broth ΔbarA/sirA BarA/SirA is a TCS involved in the indirect activation of SPI1 via the sRNAs CsrB and CsrC. CsrB and CsrC bind and titrate the RNA-binding protein CsrA from its target mRNAs, including hilD mRNA [18, 105, 106]. Early stationary phase (ESP) growth to OD6002.0 in Lennox broth ΔfliZ FliZ is involved in regulation of flagellar gene expression. FliZ also activates hilA expression through post-translational interaction with the HilD protein [22]. Early stationary phase (ESP) growth to OD6002.0 in Lennox broth Δhfq Hfq is an RNA binding protein that is a core component of post-transcriptional regulatory networks, through facilitation and stabilisation of interactions between trans-acting sRNAs and their target mRNAs [77]. Hfq affects gene function in a number of ways and a Δhfq mutant is attenuated for invasion, intracellular replication and motility [107]. Early stationary phase (ESP) growth to OD6002.0 in Lennox broth wild-type Late stationary phase (LSP) growth to OD6002.0 in Lennox broth + 6 hours further growth Growth to late stationary phase exposes S. Typhimurium to nutrient deprivation, oxygen depletion and membrane damage. This environment activates the general stress response sigma factor, RpoS, and the alternative sigma factor RpoE, associated with the extracytoplasmic stress response. ΔrpoE RpoE is an alternative sigma factor that regulates genes encoding components of the extracytoplasmic stress response (ESR). RpoE contributes to S. Typhimurium pathogenesis by regulating expression of genes required for survival of stressful host environments and a ΔrpoE mutant is attenuated for virulence in macrophages and mice [34]. Late stationary phase (LSP) growth to OD6002.0 in Lennox broth + 6 hours further growth ΔrpoS RpoS is the general stress response alternative sigma factor, responsible for transcription of genes necessary for bacterial survival in a diverse range of stressful conditions. RpoS contributes to S. Typhimurium pathogenesis through activation of the virulence plasmid-encoded spv genes and curli fibre-encoding csg genes [32, 108, 109]. A ΔrpoS mutant is attenuated for virulence following oral infection of mice [32]. Late stationary phase (LSP) growth to OD6002.0 in Lennox broth + 6 hours further growth wild-type SPI2-inducing (InSPI2) growth to OD6000.3 in PCN (pH 5.8, 1 mM MgCl2+, 0.4 mM Pi) It has previously been shown that a slightly acidic pH (5.8) and a limitation of inorganic phosphate (0.4 mM) induces expression of the SPI2 T3SS, as these conditions mimic aspects of the conditions encountered by intra-macrophage Salmonella [31]. ΔphoB/R PhoBR comprise a TCS that responds to environmental phosphate levels and regulates target genes to maintain cellular phosphate homeostasis. PhoB/R is thought to play a regulatory role in the phosphate-limiting intracellular environment. PhoB/R is involved in repression of hilA expression when phosphate levels are low and invasion gene expression is no longer required [103]. SPI2-inducing (InSPI2) growth to OD6000.3 in PCN (pH 5.8, 1 mM MgCl2+, 0.4 mM Pi) ΔslyA SlyA is required for resistance to oxidative stress and antimicrobial peptides, macrophage survival and virulence in mice but is not required for invasion of epithelial cells. SlyA binds to and activates transcription from the ssrA promoter of the SPI2-encoded TCS SsrA/B [44, 110]. SlyA plays a key role in regulation of virulence gene expression through counter-silencing of H-NS target genes [111]. SPI2-inducing (InSPI2) growth to OD6000.3 in PCN (pH 5.8, 1 mM MgCl2+, 0.4 mM Pi) ΔompR/envZ OmpR/EnvZ is a TCS involved in the acid tolerance response in S. Typhimurium. OmpR binds to and directly regulates expression from the ssrA promoter [99], while, in conjunction with the nucleoid-associated protein Fis, OmpR primes SPI2 genes for expression while S. Typhimurium is still in the host intestinal lumen [112]. SPI2-inducing (InSPI2) growth to OD6000.3 in PCN (pH 5.8, 1 mM MgCl2+, 0.4 mM Pi) ΔphoP/Q PhoP/Q is a TCS that senses the concentration of divalent cations, such as Mg2+ and Ca2+ and activates target genes in response to starvation of these cations, for example during intracellular growth. PhoP/Q contributes to S. Typhimurium pathogenesis through differential control of SPI1 and SPI2 gene expression. PhoP negatively regulates hilA, via activation of hilE in a FimZ-dependent manner [29], and positively regulates SPI2 expression through transcriptional activation of ssrB and post-transcriptional activation of ssrA [113]. SPI2-inducing (InSPI2) growth to OD6000.3 in PCN (pH 5.8, 1 mM MgCl2+, 0.4 mM Pi) ΔssrA SsrA is the SPI2-encoded sensor kinase of the SsrA/B TCS that is the central regulator of all components of the SPI2 T3SS [56]. SPI2-inducing (InSPI2) growth to OD6000.3 in PCN (pH 5.8, 1 mM MgCl2+, 0.4 mM Pi) ΔssrB SsrB is the SPI2-encoded response regulator of the SsrA/B TCS that is the central regulator of all components of the SPI2 T3SS [56]. SPI2-inducing (InSPI2) growth to OD6000.3 in PCN (pH 5.8, 1 mM MgCl2+, 0.4 mM Pi) ΔssrA/B The SsrA/B TCS is the central regulator of all components of the SPI2 T3SS. Multiple environmental and regulatory inputs are integrated at the ssrA and ssrB promoters to ensure Salmonella survival within the intracellular environment [56]. SPI2-inducing (InSPI2) growth to OD6000.3 in PCN (pH 5.8, 1 mM MgCl2+, 0.4 mM Pi) ΔhilD HilD is the primary regulator of all components of the SPI1 T3SS and also mediates regulatory cross-talk between SPI1 and SPI2, resulting in the activation of the SPI2 T3SS through counter-silencing of H-NS at the ssrB promoter [30]. SPI2-inducing (InSPI2) growth to OD6000.3 in PCN (pH 5.8, 1 mM MgCl2+, 0.4 mM Pi) It is evident from Fig 1A and 1B that each of the 18 selected regulatory genes was expressed in the wild-type strain under the growth conditions used in this study, and that the gene deletions did not usually cause significant polar downstream effects. Exceptions are discussed in S1 Text. Analysis of the transcriptome and sRNA expression landscape in this panel of mutants, using RNA-seq, revealed the complexity of gene regulation, and offers clues to the function of uncharacterised sRNAs. 10.1371/journal.pgen.1006258.g001Fig 1 Confirmation of 18 chromosomal deletions by RNA-seq. Visualisation of sequence reads in the surrounding region of each deleted gene in the relevant mutant strain and wild-type comparator, grown under identical conditions in Lennox medium (A) or SPI2-inducing PCN medium (B), in the Integrated Genome Browser. The colours of each track represent the sequencing reads which map to that locus and the height of the normalised reads is directly proportional to the level of expression at that locus (Materials and Methods). The individual panels demonstrate that the gene encoding each regulator was expressed in the wild-type strain under the growth condition chosen for analysis of that regulator, and that no sequencing reads mapped to the deleted region of the mutant strain. Neighbouring genes were generally not affected by polar mutations (see S1 Text for exceptions). White arrows with a black outline denote protein-coding genes. White arrows with a blue outline denote sRNA genes. Black bent arrows indicate TSS. All arrows indicate the direction of transcription. Predicted Rho (ρ)-independent terminators [93] are denoted by stem-loop structures. The panel of regulatory proteins (Table 2) included SPI1-encoded transcription factors (HilA, HilC and HilD) and transcription factors previously reported to control SPI1 gene expression (HilE, BarA/SirA, Fur and FliZ) [17–24]. Mutants lacking these transcription factors were studied under SPI1-inducing conditions. The SPI2-encoded regulators (SsrA/B) and regulatory systems that play an important role in intracellular survival and replication of S. Typhimurium (SlyA, HilD, OmpR/EnvZ, PhoP/Q, PhoB/R) [25–30] were investigated during growth in SPI2-inducing conditions [31]. Additionally, the regulons of the DNA adenine methyltransferase protein, Dam, and two alternative sigma factors, RpoS (σ38) and RpoE (σ24), were investigated in Lennox medium; these three transcription factors (TF) also modulate S. Typhimurium virulence gene expression [32–38]. Some of the regulatory proteins are encoded by the core ancestral S. Typhimurium genome, while others are associated with horizontally-acquired pathogenicity islands (S5 Table). The selected regulators represent a diverse range of systems that control the extracellular (SPI1-dependent) and intracellular (SPI2-dependent) infection modes of S. Typhimurium (S1 Fig), as recently reviewed [39]. The regulatory role of each protein was determined by comparing the transcriptome of isogenic mutant and wild-type strains grown in identical environmental conditions. The rationale and abbreviation used for each growth condition is detailed in Table 2. The wild-type strain and the regulatory mutants shared a similar growth rate in the relevant media (S1 Table). The mapping statistics from three independent RNA-seq runs are detailed in S2 Table. The genes controlled by each of the 18 regulatory systems were defined, as described in Materials and Methods. S3 Table contains lists of all differentially-expressed protein-coding (CDS) genes and sRNA genes (>3-fold change in expression in mutant strain compared to wild-type) in the panel of regulatory mutants. As previously mentioned, these regulons contain genes that are both directly and indirectly regulated, and further experiments will be required to identify the binding sites for each TF across the chromosome. Control of SPI1 and SPI2 expression by transcriptional regulatory systems under infection-relevant in vitro conditions As SPI1 and SPI2 are the two main Salmonella pathogenicity islands, we began by investigating the transcriptional landscape of these regions in our panel of regulatory mutants (Fig 2A and 2B). To simplify analysis of the SsrA/B regulon, differentially-expressed genes in the ΔssrA, ΔssrB and ΔssrAB mutant strains were combined as ΔssrAB (see S2 Fig and S3 Table for individual ΔssrA and ΔssrB datasets). Expression of the SPI1 CDS genes was reduced by an average of 40-fold in the absence of the SPI1-encoded transcription factors. The SPI2 CDS gene expression was reduced by an average of 109-fold in the absence of the SPI2-encoded and SPI2-associated transcription factors. 10.1371/journal.pgen.1006258.g002Fig 2 Visualisation of mapped sequence reads in the SPI1 (A) and SPI2 (B) pathogenicity islands in wild-type S. Typhimurium and the panel of regulatory mutants. Strains were grown under 5 in vitro environmental conditions, as indicated. Abbreviations for the 5 growth conditions (MEP, IEP, ESP, LSP and InSPI2) are defined in Materials and Methods and Table 2. CDS names are labelled in black, small RNA gene names in blue. All arrows indicate the direction of transcription, and transcriptional start sites are indicated as bent arrows. Strain names and conditions are labelled at the right of the image and the wild-type comparators for the relevant mutants are labelled in bold. The figure was prepared with the IGB browser, and the scale is 0–100 normalized reads for every sample. Analysis of gene expression across a panel of regulatory mutants, rather than focusing on individual mutants in isolation, allowed us to observe patterns of co-regulation between Salmonella regulons. Our transcriptomic approach highlights the regulatory connections that exist between the two main pathogenicity islands, and shows the HilD-mediated cross-talk between SPI1 and SPI2 that has previously been reported [30]. We also found CDS genes that were differentially-expressed (>3-fold) in the absence of both HilD and SsrB, and the majority of these 52 HilD/ SsrB-controlled genes were virulence-associated (Fig 3A and 3B). During growth at ESP, the absence of HilD caused the down-regulation of many SPI2-encoded and SPI2-associated genes which confirmed the previously reported HilD-mediated cross-talk between SPI1 and SPI2. Mechanistically, HilD indirectly regulates SsrB-regulated genes by antagonizing the H-NS-mediated repression of ssrAB at stationary phase under SPI-1 inducing conditions [30] (see also Fig 2). However, the up-regulation of SPI1-associated genes in the absence of SsrB suggests an additional layer of transcriptional control for SPI1 genes, via SsrB-mediated repression under SPI2-inducing conditions. 10.1371/journal.pgen.1006258.g003Fig 3 Overlap between HilD and SsrB-controlled gene expression reveals 8 putative virulence factors. Absolute expression (A) and relative expression (B) of the 52 CDS genes which are differentially-expressed (>3-fold) in the absence of HilD and SsrB, compared to wild-type grown under ESP and InSPI2 conditions, respectively (putative co-regulated genes). (C) Presence of a transposon insertion in the 52 co-regulated genes leads to attenuation or amplification of fitness in chicken, pig or calf models [41]. The scoring methods used by the authors of each publication were applied independently to each dataset (Materials and Methods). A negative score indicates attenuation of fitness as a result of the transposon insertion (blue), while an increase in fitness is denoted by a positive fitness score (orange). If no output reads were identified for a particular insertion an arbitrary negative fitness score of -15 was assigned [41]. Replacement of co-regulated genes with either a sense-oriented Kanamycin resistance cassette or an antisense-oriented chloramphenicol resistance cassette resulted in fitness attenuation in colonisation of either the spleen or liver of BALB/c mice following infection via the i.p. route [42]. A negative fitness score indicates attenuation of fitness as a result of the gene deletion (blue), while an amplification of fitness is reflected by a positive fitness score (orange). Grey indicates that no virulence data were available for that gene. Eight genes that were co-regulated by HilD and SsrB have putative roles in virulence, and are indicated in blue. Eight of the 52 HilD/ SsrB-controlled genes have not previously been directly linked to the SPI1 or SPI2 regulons and in most cases have no known function (Function UNknown, FUN). Two of the 8 genes, mcpA (STM3138) and mcpC (STM3216), are predicted to encode Salmonella-specific methyl-accepting chemotaxis proteins [40] and display similar expression patterns to the SPI1-encoded, SPI1-associated and SPI4-encoded genes which are directly or indirectly activated by HilD, and repressed by SsrB, but are not regulated by HilA suggesting they are controlled by the HilD feed-forward loop. The other six genes STM1329, STM1330, STM1600, STM1854, STM2585 and yneC (STM4079S) show a pattern of expression that resembles the SPI2-encoded and SPI2-associated genes, that are positively regulated by both HilD and SsrB under SPI1 and SPI2-inducing conditions, respectively, and are upregulated within macrophages [14]. We speculate that the HilD/ SsrB-controlled FUN genes could play a role in virulence and to further support this hypothesis, we interrogated published data from transposon-directed insertion site sequencing (TraDIS) involving oral infections of chicken, pigs and calves [41] and data from intraperitoneal infection of BALB/c mice using single gene deletion mutant libraries [42]. The importance of the 52 HilD/ SsrB-controlled genes during infection of these 4 animal models is summarised in Fig 3C. Seven of the 8 HilD/ SsrB-controlled FUN genes are required for infection of at least one of the animal models. The Gifsy-1-encoded putative transposase gene, STM2585, is required for virulence in all 4 animal models. Moreover, a deletion mutant of STM2585 (steE) is attenuated for colonization of mouse spleen and the SteE protein is translocated into the host cell cytoplasm [43]. Furthermore, STM2585 and STM1330 are regulated by SlyA and PhoP [44], the master regulators of the SsrB regulon. In contrast, mcpC mutants show increased fitness in all animal models, consistent with a role in down-regulating the expression or function of specific traits that may have a fitness cost for the bacterium during the invasion process. The identification of these putative novel virulence factors through investigation of their regulatory patterns highlights the power of our approach to both recapitulate current knowledge on SPI1 and SPI2 regulatory cross-talk, and to predict the role of novel FUN genes in virulence. The influence of transcriptional regulatory systems upon the expression landscape of sRNAs The CDS targets of many Salmonella TFs have previously been identified by microarray and single gene analysis studies [37, 44–50], and data from these published studies were used to validate our method of investigating bacterial regulons. Table 3 highlights the effects of the panel of regulatory mutants on flagellar, SPI1 and SPI2 gene expression in the context of the literature. Our data confirm published S. Typhimurium virulence gene regulatory interactions, and identify new regulatory links described below. Any differences between the regulons defined in this study and the literature are discussed in S1 Text, and typically arise from differences in the growth conditions that were used, strain background or the increased dynamic range and sensitivity of the RNA-seq-based method of transcriptional profiling. Overall, these comparisons confirm that RNA-seq-based transcriptomics accurately characterises CDS gene expression, and so allows individual sRNAs to be assigned to specific cellular regulons. 10.1371/journal.pgen.1006258.t003Table 3 Comparison of virulence gene expression in published regulons with panel of regulatory mutantsa. Flagella SPI1 SPI2 Supported by the literature Disagrees with the literature Novel findings Strain HilA HilA does not affect flhD expression [114]. HilA activates SPI1 [115]. 14028 [114] SL1344 [115] HilC HilC activates SPI1 [116]. SL1344 [116] HilD HilD activates flhD expression [114]. HilD activates SPI1 [102]. HilD activates SPI2 [30]. 14028 [114] 14028[102] SL1344 [30] BarA/SirA BarA/SirA represses flagellar gene expression [17]. BarA/SirA activates SPI1 [18]. fljB expression decreases in the absence of BarA/SirA. Expression of SPI2 genes decreases in the absence of BarA/SirA. 14028 [17] 14028 [18] FliZ FliZ activates most flagellar genes [117]. FliZ activates SPI1 [118]. Expression of fljAB increases in the absence of FliZ. Expression of SPI2 genes decreases in the absence of FliZ. 14028 [117] SL1344 [118] Fur Fur activates flagellar genes [119]. Fur activates SPI1 genes [19]. Fur represses SPI2 genes [120]. 14028 [119] 14028 [19] 14028 [120] Hfq Hfq activates flagellar genes, SPI1 genes and SPI2 genes [78]. SL1344 [78] HilE Dam Dam methylation represses fliD [37]. Expression of fliC decreases in the absence of Dam. Expression of SPI1 increases in the absence of Dam. Expression of fljAB increases in the absence of Dam. SL1344 [37] RpoE RpoS fliC expression decreases in the absence of RpoS [63]. Expression of invF decreases in the absence of RpoS. Expression of ssa genes increase in the absence of RpoS. 14028 [63] PhoB/R PhoP/Q PhoP/Q represses flagellar genes [121]. PhoP/Q represses SPI1 [29]. PhoP/Q activates SPI2 [113]. SL1344 [121] SL1344 [29] 14028 [113] OmpR/EnvZ OmpR/EnvZ represses flagellar genes [50]. OmpR/EnvZ represses SPI1 [57]. OmpR/EnvZ activates SPI2 [26]. SL1344 [50] SL1344 [57] 14028 [26] SlyA SlyA activates SPI2 [44]. Expression of flagellar genes increases in the absence of SlyA. Expression of SPI1 genes increases in the absence of SlyA. 14028 [44] SsrA/B SsrA/B activates SPI2 [25]. Expression of flagellar genes increases in the absence of SsrA/B. Expression of SPI1 genes increases in the absence of SsrA/B. SL1344 [25] a Coloured squares indicate the putative regulatory effect of the indicated regulatory system on the indicated group of genes based on differential gene expression in the panel of regulatory mutants in this study. Blue indicates putative repression (higher gene expression in mutant than wild-type); Red indicates putative activation (lower gene expression in mutant than wild-type); Magenta indicates a mixture of up- and down-regulated genes; Grey indicates no change in expression. The Transcripts Per Million (TPM) approach was used to generate expression values from the RNA-seq data, and to define whether individual genes were expressed (threshold = TPM 10). Approximately 75% of the 280 sRNAs of wild-type S. Typhimurium strain 4/74 were expressed in all 5 conditions used in this study, while a further 20% of sRNAs were expressed in at least one environmental condition (Fig 4A). In the panel of transcription factor mutants, 44% of S. Typhimurium sRNA genes (124) were differentially-expressed by at least 3-fold. Almost 50% of the differentially-expressed sRNAs received only one regulatory input (direct or indirect) from the panel of TFs and approximately 11% of differentially-expressed sRNAs received 5 or more regulatory inputs from the panel of TFs (Fig 4B). The sRNA genes in each of the categories shown in Fig 4B are detailed in S6 Table. Fig 5 shows the shared sRNA targets between each regulatory system. We identified many sRNAs that are controlled by both the SPI1 and SPI2 regulatory systems, reflecting the hierarchical regulatory structures that control expression of the two main pathogenicity islands. 10.1371/journal.pgen.1006258.g004Fig 4 Expression and regulation of S. Typhimurium sRNAs. (A) sRNAs expressed in wild-type S. Typhimurium 4/74 under 5 in vitro environmental conditions. (B) Bar chart showing the number of differentially-expressed sRNAs in the context of the number of putative regulatory inputs to each gene from the panel of 18 TFs. 10.1371/journal.pgen.1006258.g005Fig 5 Pattern of differential expression of S. Typhimurium sRNAs in the panel of regulatory mutants. Concentric circle diagram representing all 280 S. Typhimurium sRNAs, which are labelled outside the circle according to their relative chromosomal position. The putative regulatory inputs reflect the pattern of sRNA gene expression in the panel of regulatory mutants. Each circular track represents a single regulatory system. Red boxes indicate putative activation of sRNA expression (>3-fold decrease in expression in mutant compared to wild-type); Blue boxes indicate putative repression of sRNA expression (>3-fold increase in expression in mutant compared to wild-type); yellow indicates no change in expression in mutant compared to wild-type. The outermost ring indicates the conservation status of each sRNA, with Salmonella-specific sRNAs shown in green. The sRNAs that are conserved in other enteric species are orange, based on our previous analysis [14]. Diagram generated with Circos (http://www.circos.ca). The majority (>60%) of differentially-expressed sRNAs were transcribed from intergenic regions, while over 20% of differentially-expressed sRNAs were located in the 3’UTR of a coding gene, a chromosomal location that constitutes a large reservoir for small non-coding RNAs [51]. Forty-eight percent of the 3’UTR-derived sRNAs are transcribed from their own promoter, while the remaining 52% are likely to be processed from the mRNA of the upstream ORF [9]. We speculate that processed sRNAs are often functionally-related to the upstream co-transcribed ORFs, while expression of the sRNAs that are transcribed from their own promoter may be controlled by alternative transcription factors and these sRNAs have a distinct function. Expression of the majority (57%) of ORFs upstream of putative processed sRNAs follow the same regulatory pattern as their 3’-derived sRNAs in the panel of regulatory mutants (>3-fold change; S6 Table), while only one upstream ORF displays the same regulatory pattern as its non-processed downstream 3’UTR-derived sRNA (S6 Table). Over 60% of differentially-expressed 3’-encoded sRNAs were Hfq-enriched [51], suggesting that many of the RNAs in this category function as canonical trans-acting sRNAs and require the RNA chaperone Hfq. Conversely, only 20% of antisense-encoded sRNAs were enriched for Hfq binding, highlighting the fact that many of this class of sRNAs are Hfq-independent and may act in cis to their target genes (S3 Fig and S6 Table). The size of the sRNA-based regulons ranges from 63 differentially-expressed sRNAs (RpoS) to one differentially-expressed sRNA (HilA and PhoB/R) (Fig 6A and S4A Fig). The SPI1-encoded regulators HilC and HilD function as part of a feed-forward regulatory loop [4], and considerable overlap was observed between the sRNA genes that were controlled by these TFs (Fig 5). The HilE CDS and sRNA regulons were distinct from the HilA, HilC and HilD regulons at IEP, and show a novel positive regulatory role for HilE in the control of metabolic gene expression (S3 Table), in addition to the reported role as a negative regulator of SPI1 genes [52]. The sRNA genes which were differentially-expressed in the absence of the BarA/SirA regulatory system were a subset of the HilC and HilD-regulated genes, reflecting the indirect activation of SPI1 by BarA/SirA, via the sRNAs CsrB and CsrC (Fig 5) [18]. More sRNA genes were up-regulated than down-regulated in the mutant strain that lacks Fur, consistent with the primary function of Fur as a transcriptional repressor [53]. RyhB-1 and RyhB-2, the S. Typhimurium homologues of the E. coli Fur-repressed sRNA RyhB, were up-regulated in the absence of Fur. 10.1371/journal.pgen.1006258.g006Fig 6 sRNA and CDS expression in the panel of regulatory mutants. Bubble diagram showing the numbers of differentially-expressed sRNA genes (A) and the numbers of differentially-expressed CDS (B) in the panel of regulatory mutants. The bubble size is proportional to the number of differentially-expressed genes in the isogenic mutant that lacks each regulatory system. The x-axis indicates the number of genes which show increased expression (repressed) in the mutant strain, the y-axis indicates the number of genes which show decreased expression (activated) in the mutant strain. The numbers of differentially-expressed genes are indicated in the key to the right of each graph. We investigated whether promoters of sRNAs which were up-regulated in the absence of Fur contained “Fur-boxes” or recognition motifs for Fur binding, consistent with direct repression of expression by Fur. The promoters of five up-regulated sRNA genes contained high-scoring putative Fur boxes (S5 Fig). The highest scoring Fur box was present in the promoter of the Fur-dependent RyhB-1 gene [54]. Other high-scoring Fur boxes were located in the promoters of sRNA genes located adjacent to iron-associated or Fur-regulated operons. STnc4000 contains a putative Fur binding site overlapping the -10 site of the promoter region. This sRNA is encoded in an intergenic region and is transcribed divergently from the bfd gene, which encodes a Fur-regulated bacterioferritin-associated ferredoxin protein. The second highest scoring putative Fur binding site was obtained for a sequence overlapping the -10 site of the STnc3250 promoter. STnc3250 is intergenic and transcribed divergently from fhuA, which is the first gene in a Fur-dependent operon that encodes ferrichrome-iron associated proteins. The proximity of STnc4000 and STnc3250 to genes involved in iron homeostasis raises the possibility that Fur regulation of STnc4000 and STnc3250 may be involved in the control of iron metabolism. The Fur-repressed divergently-oriented promoters of STnc3250 and STnc4000 may represent new examples of the bidirectional promoters that belong to the Fur regulon, such as, fepA-fes, fepD-ybdA and fepB-entCEBA [55]. We found that SPI2-associated regulators OmpR/EnvZ, PhoPQ, SsrAB and SlyA control more sRNAs than the SPI1-associated regulators (Figs 5 and 6A and S4A Fig). Many of the sRNA genes which are differentially-expressed in the SPI2-associated regulatory mutants are likely to be indirectly regulated, reflecting the complexities of the SPI2 regulatory hierarchy [56]. The sRNAs which are differentially-expressed in the absence of SsrA/B are a subset of those that are differentially-expressed in the absence of the PhoP/Q and OmpR/EnvZ two-component systems (TCS) (Fig 5), reflecting the roles that PhoP and OmpR play in the control of virulence gene expression as well as the regulation of components of the ancestral genome [57, 58]. STnc1860 was the only differentially-expressed sRNA in the ΔphoB/R mutant (>4-fold decrease in expression). STnc1860 is located downstream of the phoU gene which encodes a transcriptional regulator of the PstSCAB-PhoU high affinity phosphate transport system operon and is a target of the PhoB/R TCS [59]. STnc1860 is co-transcribed with this operon from the pstS start site [9] and could be involved in the control of phosphate assimilation. Overall, each regulatory system modulated the expression of a similar proportion of sRNAs and CDS (Fig 6B and S4B Fig). The general effect of the panel of regulatory mutations is negative on expression of CDS and sRNA genes. More differentially-expressed CDS genes were controlled by the SPI2-associated regulatory systems than the SPI1-associated regulators. The key difference between the sRNA-based and CDS-based regulatory networks was seen in the context of RpoS-dependent transcriptional control. RpoS modulated the expression of more sRNA genes than other regulatory proteins (S4B Fig), whereas the CDS-based network of RpoS was among the smaller regulons, containing only 190 putative RpoS-dependent coding genes. We propose that RpoS is a hub for sRNA-mediated gene regulation in S. Typhimurium. RpoS-regulated genes play a role in the general stress response at LSP [60]. The prominent role played by RpoS in regulation of sRNA expression at LSP may reflect the pleiotropic functions that sRNAs play in mediating cellular responses to stress [16], or the lack of expression of many CDS under LSP conditions. The large number of sRNAs which were differentially-expressed in the absence of RpoS highlights the importance of RpoS as a hub for post-transcriptional regulation [61–63]. Of the 63 sRNAs that are differentially-expressed in the absence of RpoS, 36 are Salmonella-specific and 27 are conserved in other species [14] (Fig 5). ChIP analysis has shown that three sRNAs, OmrA, SibC (RygC) and RyeB (SdsR) belong to the core RpoS regulon of E. coli [64]. These three conserved sRNAs were also differentially-expressed in the absence of RpoS in our study (S3 Table). Furthermore, Peano et al. identified an RpoS binding site in the intergenic region between the divergently transcribed tisB gene and the conserved sRNA IstR-1_2. The expression of IstR-1_2 is reduced in the absence of RpoS in our study, suggesting that RpoS activates IstR-1_2 expression in both species. We anticipate that, in future, the joint interrogation of our transcriptomic data with other global ChIP-based studies will elucidate entire regulons. We aligned the promoter sequences of the 49 sRNAs that were down-regulated (>3-fold) in the absence of RpoS to determine if the promoters of the Salmonella-specific sRNAs contain hallmarks of RpoS-dependent promoters and are, therefore, likely to be genuine members of the RpoS regulon. A C nucleotide at position -13 relative to the transcriptional start site is highly conserved among RpoS-dependent promoters, and does not favour binding by RpoD [65]. Twenty-one of the 49 RpoS-dependent promoters contained a C at position -13, including the promoters responsible for the expression of OmrA, SdsR and IstR-1_2, (S6 Fig). Placing of a G or T at position -14 and degeneracy of the sequence and position of the -35 hexamer are also typical of RpoS-dependent promoters [66] and are common features of the RpoS-dependent sRNA promoters (S6 Fig). sRNA promoters are not qualitatively different from the promoters of CDS [67] and, with the exception of the RpoS regulon, the striking similarities of regulatory input on sRNA promoters with the regulatory input on CDS promoters demonstrate that the transcriptional regulation of sRNAs is mediated by established cellular regulatory networks. Our panel of S. Typhimurium regulatory systems did not reveal a single dedicated TF for transcriptional control of all sRNAs, suggesting that sRNAs are likely to have been integrated into existing networks as required. Confirmation of key regulatory inputs for five sRNAs To confirm the regulatory findings for sRNAs of interest, we used northern blots to confirm differential sRNA gene expression between wild-type and regulatory mutants (Fig 7). Expression of the sRNA STnc520 decreased 17-fold in the absence of the primary SPI1 regulator, HilD, at ESP (Fig 7A). In S. Typhimurium strain 14028, expression of STnc520 was directly activated by the SPI1-encoded transcription factor, SprB (Joseph T. Wade, pers. comm.). Direct regulation of STnc520 by SprB was also confirmed in S. Typhimurium 4/74 in this study (S7A and S7B Fig). 10.1371/journal.pgen.1006258.g007Fig 7 Confirmation of novel regulatory interactions that control five sRNAs. (A) STnc520, (B) FnrS, (C) STnc1330, (D) RyeF, (E) STnc1480. Each panel shows the visualisation of sequencing reads obtained from RNA-seq experiments using the IGB browser beside northern blots that validate the differential expression data using RNA extracted from biological replicates of wild-type and mutant strains grown under the same in vitro environmental condition, as indicated. 5S RNA was probed as a loading control. Expression of the FNR-dependent sRNA FnrS decreased approximately 32-fold in the absence of Fur at ESP (Fig 7B), consistent with the reported overlap between the Fur and FNR regulons [45]. The Fur-mediated up-regulation of FnrS is unlikely to be a direct effect, however, as Fur does not directly activate transcription [53]. The wild-type strain accumulates the sRNA STnc1330 at LSP and, in keeping with previous findings [63], STnc1330 expression is reduced approximately 61-fold in the absence of RpoS (Fig 7C). RyeF expression was reduced 25-fold in the absence of S. Typhimurium RpoE (Fig 7D) and we note that MicL, the E. coli orthologue of the RyeF sRNA, belongs to the RpoE regulon in E. coli [68]. The Salmonella-specific sRNA, STnc1480, was most highly expressed under environmental conditions that mimic the host intracellular environment and within murine macrophages [9, 14]. STnc1480 has multiple regulatory inputs from SPI2-associated regulatory systems, with the largest reduction in STnc1480 expression (approximately 58-fold) being seen in the absence of both the PhoP/Q and SlyA regulatory systems (Fig 7E). Transcriptional regulation of the STnc1480 sRNA The expression and regulatory profile of STnc1480 suggests that this sRNA may be important during the intracellular lifestyle of Salmonella, leading us to investigate the expression pattern of this sRNA in greater detail. The PhoP and SlyA-dependent expression of STnc1480 was confirmed by ectopic expression of PhoP or SlyA from the arabinose-inducible PBAD promoter, which specifically restored STnc1480 expression in ΔphoP and ΔslyA mutant backgrounds respectively; however ectopic expression of either protein was unable to restore STnc1480 expression in a mutant strain which did not express the other regulator (Fig 8A and 8B). These data argue that both SlyA and PhoP are required for optimal expression of STnc1480. There is significant overlap between the SlyA and PhoP regulons and a transcriptional requirement for both SlyA and PhoP has previously been reported for horizontally-acquired genes that are subject to silencing by the nucleoid-associated protein H-NS [44, 69, 70]. We, therefore, determined whether STnc1480 transcription was subject to H-NS-mediated silencing and subsequent counter-silencing by SlyA and PhoP. 10.1371/journal.pgen.1006258.g008Fig 8 Regulation of STnc1480 expression. (A) Northern blot showing that STnc1480 is expressed in wild-type cells and is SlyA-dependent, even when PhoP expression is induced from the PBAD promoter. STnc1480 expression is restored when PhoP is ectopically expressed in a ΔphoP background and addition of L-arabinose to a ΔphoP mutant carrying the empty pBAD vector does not affect STnc1480 expression. (B) Northern blot showing that STnc1480 is expressed in wild-type cells and accumulates to a higher level when SlyA is ectopically expressed from the PBAD promoter. STnc1480 expression was rescued when SlyA was exogenously supplied in a ΔslyA mutant and addition of L-arabinose to a ΔslyA mutant carrying the empty pBAD vector does not affect STnc1480 expression. STnc1480 expression was not rescued by the ectopic expression of SlyA in the absence of PhoP. -/+ indicates the respective absence or presence of the inducer L-arabinose. 5S RNA was probed as a loading control. Each northern blot is representative of at least 3 independent experiments. Chromatin immunoprecipitation followed by qPCR was used to investigate H-NS occupancy of the STnc1480 promoter under non-inducing conditions (MEP) and inducing conditions (Low Mg2+). The proV promoter and the hemX gene were used as positive and negative control regions, respectively, as H-NS binds to the proV promoter and does not bind to the hemX gene [71]. Fig 9A and 9B are representative of two independent biological replicates and show that H-NS associated with the STnc1480 promoter, with greater enrichment observed for the experimental IP sample (FLAG), compared to the mock IP sample under both inducing and non-inducing conditions. These data indicate that H-NS bound directly to the STnc1480 promoter under non-inducing conditions and that H-NS was not displaced from the STnc1480 promoter under conditions when the sRNA was highly expressed. 10.1371/journal.pgen.1006258.g009Fig 9 H-NS occupancy of the STnc1480 promoter. qPCR data of representative ChIP assay from two independent experiments demonstrating H-NS binding to the STnc1480 promoter under (A) non-inducing (MEP) and (B) inducing (LowMg2+) environmental conditions. Experimental (FLAG) and mock DNA from the STnc1480 promoter and the positive (proV) and negative (hemX) control regions was normalised to the starting amount of DNA, which was extracted prior to immunoprecipitation (IP/Input). (C) Northern blot showing expression of STnc1480 under non-inducing and inducing conditions in wild-type and in the absence of a functional H-NS protein in combination with ΔslyA and ΔphoP mutations. 5S RNA was probed as a loading control. To further understand the relationship between STnc1480 and the TFs PhoP and SlyA, and to characterise the roles of PhoP and SlyA in counter-silencing, we investigated the expression of STnc1480 in strains lacking a functional H-NS protein in either ΔphoP or ΔslyA mutant backgrounds. Fig 9C confirms that STnc1480 was not highly expressed during logarithmic growth in rich medium in a wild-type background, and that STnc1480 expression was de-repressed in the absence of a functional H-NS protein under this growth condition. SlyA was no longer fully required for STnc1480 transcription under non-inducing or inducing conditions, arguing that the key role of SlyA in STnc1480 expression is to counteract the repressive effects of H-NS, rather than to activate transcription; SlyA plays a similar function at the pagC promoter [70]. Counter-silencing of H-NS is likely to be achieved through SlyA-mediated restructuring of the STnc1480 promoter architecture, rather than displacement of H-NS. We speculate that PhoP plays the role of a classical transcriptional activator at the STnc1480 promoter, rather than counter-silencer, as the presence of PhoP is required for transcription even in the absence of a functional H-NS protein. Identification of sRNAs with SPI1-like and SPI2-like expression profiles The “guilt by association” hypothesis posits that groups of genes which perform similar functions are co-expressed and/or co-regulated, allowing transcriptomic data to be used to identify genes which may share related functions [72]. Several approaches are available to reveal novel interactions between TF and co-expressed genes [73], and correlative patterns are becoming widely used for the inference of causal influence and to define transcriptional networks [74]. We previously identified transcriptional signatures for SPI1- and SPI2-related genes, based on the expression profiles of the archetypical SPI1 gene, prgH, and the archetypical SPI2 gene, ssaG [9]. To identify sRNA genes which may play important roles in S. Typhimurium virulence, we searched for sRNAs with expression profiles that closely correlate to the expression of SPI1 and SPI2 genes. Our global transcriptomic analyses identified 2 sRNA genes that showed a SPI1-like pattern of expression across 22 environmental conditions [9], within murine macrophages [14] and in our panel of regulatory mutants (Pearson correlation coefficient > 0.7 [prgH]) (Table 4 and Fig 10A). InvR (located within SPI1) is transcriptionally activated by the primary SPI1 transcription factor, HilD [75], confirming the value of correlative analysis for identifying genuine regulatory interactions. As previously discussed, the second SPI1-like sRNA, STnc520, is directly regulated by SPI1-encoded SprB. 10.1371/journal.pgen.1006258.g010Fig 10 Putative regulatory inputs for thirteen SPI1-like and SPI2-like sRNAs. (A) Heatmap showing relative expression of two SPI1-like sRNAs in comparison to the archetypical SPI1 gene prgH and 11 SPI2-like sRNAs in comparison to the archetypical SPI2 gene ssaG in a panel of 15 regulatory mutants, compared to the wild-type strain grown under the same environmental condition. SPI1-like sRNAs were defined as sRNAs which show a similar pattern of expression (Pearson correlation coefficient >0.7) under 22 in vitro environmental conditions, within murine macrophages and in a panel of 18 regulatory mutants to the archetypical SPI1 prgH gene. SPI2-like sRNAs were defined under exactly the same conditions using the archetypical SPI2 gene ssaG. Yellow indicates no change in expression; red indicates an increase in expression; and blue indicates a decrease in expression in the mutant strain, compared to wild-type. (B) Regulatory network representing a hypothetical model of the transcriptional regulation of the SPI1-like and SPI2-like sRNAs, based on differential expression in the absence of the indicated regulatory protein. Salmonella-specific sRNAs are green circles; Enterobacteriaceae-conserved sRNAs are orange circles; transcription factors are multi-coloured octagons; up-regulation (>3-fold decrease in expression in mutant/wild-type) shown with a red arrow; down-regulation (>3-fold increase in expression in mutant/wild-type) shown with a blue T-bar. Regulatory network was generated using www.cytoscape.org. 10.1371/journal.pgen.1006258.t004Table 4 sRNAs which show co-regulation with SPI1 or SPI2. SPI1-like sRNA Upstream Gene Downstream Gene Genomic context Pearson correlation coefficient with prgHa Significant TraDIS attenuation STnc520 pliC pagC > < < 0.84 Yes InvR invH STM2901 > > > 0.75 Yes SPI2-like sRNA Upstream Gene Downstream Gene Genomic context Pearson correlation coefficient with ssaGb Significant TraDIS attenuation IsrH_1_2 sseL glpC > < > 0.98 Yes STnc3730 yfcC pta > < > 0.89 - c STnc3090 oadG STM0057 < > < 0.87 - STnc470 (InvS) STM0082 STM0081 > < < 0.85 - STnc3020 prgJ prgH as to prgI 0.84 Yes STnc3180 ybdO dsbG < > < 0.83 - PinT STM4310 STM4312 > > < 0.82 - STnc1480 yeaJ yeaH < > < 0.77 Yes STnc3050 oadG SL3326 < < < 0.75 Yes sRNA10 Rnk Rna < < < 0.75 - STnc3170 intA thdF > < > 0.72 - a Pearson correlation coefficient with prgH across 22 environmental conditions, within murine macrophages and in panel of 18 regulatory mutants b Pearson correlation coefficient with ssaG across 22 environmental conditions, within murine macrophages and in panel of 18 regulatory mutants c-indicates either that there was either no transposon insertion within this gene, or that there was a transposon insertion that was not associated with significant attenuation. Eleven sRNA transcripts showed a SPI2-like expression pattern (Pearson correlation coefficient > 0.7 [ssaG]) (Table 4 and Fig 10A). None of the SPI2-like sRNA genes are located in the SPI2 pathogenicity island. In fact, the SPI2-like sRNA STnc3020 is encoded within the SPI1 island, antisense to prgI, and shows a modest negative correlation with expression of the archetypical SPI1 gene, prgH (Pearson correlation coefficient: -0.21). A high scoring SsrB binding motif was identified adjacent to the putative -35 site of the STnc3020 promoter (S8 Fig). This observation raises the intriguing possibility that STnc3020 has been co-opted by SsrB to integrate regulatory and environmental cues and mediate cross-talk between SPI1 and SPI2. The STnc1480 sRNA shows a SPI2-like expression profile [9]. The previously-discussed PhoP- and SlyA-dependence of this sRNA suggests that STnc1480 could play a role in the expression of SPI2-associated regulatory systems. Furthermore, a second SPI2-like sRNA, PinT (STnc440) is PhoP-dependent, and was one of the most highly up-regulated sRNA transcripts upon internalisation of Salmonella within macrophages [14] and various other host cell types [76]. PinT represses both SPI1- and SPI2-associated virulence genes, and is a post-transcriptional timer of Salmonella gene expression during infection [76]. We speculated that some uncharacterised sRNAs that show SPI1 or SPI2-like expression patterns could have important functions in S. Typhimurium virulence and we investigated potential virulence phenotypes in the context of TraDIS datasets [41]. The two SPI1-like and four SPI2-like sRNAs were required for optimal fitness during infection of the chicken, pig or calf models (summarised in Table 4). We present a hypothetical model of a regulatory network which highlights the key regulatory inputs (>3-fold change in expression) of the 13 SPI1-like and SPI2-like sRNAs (Fig 10B), showing direct or indirect regulation of the SPI1-like and SPI2-like sRNAs mainly by SPI-associated regulatory proteins. The two SPI1-like sRNAs, InvR and STnc520, receive negative regulatory inputs from the SPI2-associated regulatory proteins in addition to the positive regulatory inputs from the SPI1-associated regulators, consistent with cross-talk between the expression of the SPI1 and SPI2 systems. The 11 SPI2-like sRNAs are highly interconnected by their regulatory inputs reflecting the complexity of the SPI2 regulatory hierarchy, although a number of the regulatory interactions are likely to be indirect. This hypothetical regulatory network provides the foundation for future experimentation to identify the direct mechanism of transcriptional regulation of the SPI1-like and SPI2-like sRNAs, which present interesting candidates for further analysis. Conservation of SPI1-like and SPI2-like sRNAs Our laboratory recently published a phylogenetic analysis of the 280 Salmonella sRNAs across 29 enterobacterial genomes [14]. One hundred and seventy-six “Salmonella-specific” sRNAs were identified, that showed >90% sequence identity across the Salmonella genus and <70% sequence identity with other members of the Enterobacteriaceae. Eleven of the 13 SPI1-like and SPI2-like sRNAs were Salmonella-specific. This high proportion of Salmonella-specific sRNAs is consistent with a role for the SPI1-like and SPI2-like sRNAs in the evolution of S. Typhimurium as a pathogen (S5 Table). Despite the fact that none of the 11 SPI2-like sRNAs are encoded within the SPI2 island, six SPI2-like sRNAs are S. enterica-specific and are not conserved in Salmonella bongori. The limited conservation of these sRNAs outwith the S. enterica species suggests that these elements were either acquired with or after their SPI2-encoded regulators, and were co-opted to perform regulatory functions within the intracellular environment after the evolutionary divergence of S. enterica and S. bongori. We performed a similar phylogenetic analysis of the 15 TFs, response regulators and sigma factors used in this study and confirmed that HilD, HilA, HilC, HilE and SsrB are Salmonella-specific regulators, while the remaining 10 regulators are conserved outwith the Salmonella genus (S5 Table). We found that the SPI1-like and SPI2-like sRNAs are controlled by a mixture of regulatory inputs from both Salmonella-specific and Enterobacteriaceae-conserved regulatory systems. The Hfq regulon The RNA chaperone Hfq facilitates binding between sRNA and target mRNA molecules and contributes to sRNA-mediated post-transcriptional regulation by various mechanisms. Hfq also controls sRNA stability prior to target recognition, either through protection from ribonucleases or through promotion of sRNA decay [77]. The role of Hfq in modulating the global expression of Salmonella sRNAs has not yet been reported as our previous analysis of the Salmonella Hfq regulon relied upon DNA microarrays [78]. Previous co-immunoprecipitation analysis, combined with RNA-seq, demonstrated that approximately half of S. Typhimurium sRNAs were associated with Hfq in a number of environmental conditions [51]. This led us to investigate how the absence of the Hfq protein affects the steady state levels of the Hfq-bound and non-bound sRNAs. Clearly, interactions between Hfq and other proteins or with many mRNAs have pleiotropic effects on gene expression, and these effects cannot always be directly attributed to Hfq [79]. Therefore, we combined the available Hfq co-immunoprecipitation data [51] with our transcriptomic data to identify the Hfq-dependent sRNAs that were physically associated with Hfq, and represent candidate canonical trans-acting sRNAs (S6 Table). Sixty-three of the 280 S. Typhimurium sRNAs were differentially-expressed in the absence of Hfq (3-fold or greater change in transcript level), and were designated as Hfq-regulated (S3 Table). The majority (87%; 55 sRNAs) of Hfq-regulated sRNA transcripts were enriched by co-immunoprecipitation with Hfq in at least one of the environmental conditions used by Chao et al, and 67% (42 sRNAs) were Hfq-enriched at ESP (Fig 11A). 10.1371/journal.pgen.1006258.g011Fig 11 Analysis of the Hfq-dependent sRNA regulon. (A) Venn diagram comparing sRNA genes that were differentially-expressed in Δhfq compared to wild-type, and sRNAs which were associated with Hfq under any condition, or specifically at ESP. Fifty-five sRNAs indicated within the dashed black box were included for analysis of differential expression in the panel of regulatory mutants for panel (C). (B) Bar chart demonstrating the number of sRNA genes that are up- or down-regulated in Δhfq compared to wild-type in the context of association with Hfq. (C) Regulatory network representing a hypothetical model of the transcriptional regulation of 30 out of 55 Hfq-associated sRNAs, based on differential expression in the absence of the indicated regulatory protein. Salmonella-specific sRNAs are green circles; Enterobacteriaceae-conserved sRNAs are orange circles; transcription factors are multi-coloured octagons; up-regulation (>3-fold decrease in expression in mutant/wild-type) shown with a red arrow; down-regulation (>3-fold increase in expression in mutant/wild-type) shown with a blue T-bar. Association with Hfq was determined by Chao et al (2012) using Hfq co-immunoprecipitation [51]. Nineteen sRNAs were enriched for Hfq at ESP but did not show Hfq-dependent expression at ESP. Two of these sRNAs, AmgR and STnc2100, were not expressed at ESP or in the Δhfq mutant and so were excluded from our analysis. The remaining 17 Hfq-enriched, non-differentially-expressed sRNAs may require Hfq binding for their activity and to aid binding to their target mRNAs but do not require Hfq for stability. A further 8 sRNAs were differentially-expressed in the Δhfq mutant but not enriched for Hfq in any of the conditions used by Chao et al. The differential expression of a number of these sRNAs may have an indirect cause as the absence of Hfq has wide-ranging effects on sigma factors and TFs [78]. Therefore, the differential expression of some sRNAs in the Δhfq mutant may reflect altered transcription rather than an altered rate of turnover. For example, approximately 16% of Hfq-regulated sRNAs were up-regulated in the absence of Hfq (Fig 11B), including the RpoE-dependent sRNAs RybB, MicA and RyeF [68, 80]. RpoE, and genes in the RpoE regulon, were more highly expressed in a Δhfq mutant due to activation of the extra-cytoplasmic stress response [81], suggesting that increased levels of RpoE-dependent sRNAs in the Δhfq mutant simply reflect an induction of RpoE. Approximately 30% of up-regulated sRNAs were not enriched for Hfq, compared to only 9% of down-regulated sRNAs which were not enriched for Hfq, reflecting an indirect regulatory effect on the up-regulated sRNA genes (Fig 11B). We investigated how the 55 Hfq-regulated and Hfq-enriched sRNAs (highlighted using a dashed black box on Fig 11A) were expressed in the panel of regulatory mutants. The expression of thirty of the 55 Hfq-associated sRNAs was modulated by at least one of the regulatory systems under investigation. We generated a model of the transcriptional regulatory network for these thirty canonical trans-acting sRNAs, based on the sRNA expression patterns in the panel of regulatory mutants (Fig 11C). Approximately half of the Hfq-associated differentially regulated sRNAs were Salmonella-specific, while the other half are conserved in other members of the Enterobacteriaceae family (S5 Table). The potential for diversity of function of Hfq-dependent trans-acting sRNAs is reflected in the diverse range of transcriptional regulators of these sRNAs. The sRNAs with multiple regulatory inputs connect multiple hubs, and we speculate that these sRNAs play a physiological role in connecting regulons to integrate multiple regulatory signals and generate a co-ordinated genetic output in vivo. The sRNAs with fewer regulatory inputs may play more specific roles within their respective regulons. This transcriptomic analysis of the Δhfq mutant, coupled with the Hfq co-immunoprecipitation data [51], identified a core group of sRNAs that required Hfq for both their activity and stability. Many uncharacterised sRNAs belong to this core group, and represent interesting candidates for further investigation, which are likely to function as canonical trans-acting sRNAs under virulence-associated conditions. Community data resource We previously developed an online tool, named SalComMac, which allows users to interrogate the expression profiles of S. Typhimurium genes in the wild-type strain grown under a suite of 22 infection-relevant environmental conditions [9] and within murine macrophages [14]. The RNA-seq sequence reads may be visualised in the context of the chromosome using the interactive online browser Jbrowse [9, 14]. We now provide the gene expression profiles of S. Typhimurium genes in the panel of regulatory mutants as a compendium database (http://tinyurl.com/SalComRegulon), and in the context of the chromosome (http://tinyurl.com/SalComRegulon-Jbrowse). The SalComRegulon tool provides absolute gene expression values in wild-type and mutant strains, and the relative expression of genes in mutant strains compared to the wild-type strain grown in the same environment. This transcriptomic dataset is intended to provide a valuable resource for the investigation of the expression profiles of genes of interest across a panel of regulatory mutants by the bacterial research community. Perspective In the past, regulons have been defined in individual Salmonella strains, grown in various conditions and following different experimental criteria. Here, we have used a single S. Typhimurium strain, and just five growth conditions to perform a systematic and high-resolution analysis that offers the first opportunity for direct comparison between these infection-relevant regulatory systems. This analysis has allowed us to examine previously unseen regulatory interactions and uncover novel putative virulence genes. This database will provide the foundation for many hypothesis-driven future studies. We have expanded the regulons of key S. Typhimurium virulence-associated regulatory proteins under infection-relevant growth conditions and we present for the first time a detailed global view of the genes controlled by HilE, FliZ, RpoE and PhoB/R. The RNA-seq approach has given an unprecedented view of S. Typhimurium gene expression at single nucleotide resolution. By analysing the transcriptome of 18 mutants lacking important sigma factors, TFs, TCSs and an RNA chaperone we have identified new interactions between well-characterised regulatory systems, and the regulatory inputs that control the expression of the majority of uncharacterised S. Typhimurium sRNAs. As shown in S1 Fig, the regulatory systems under investigation in this study are key components of the complex transcriptional network responsible for the rapid adaptation required for successful infection of mammalian hosts. The transcriptional control of sRNAs can generate regulatory loops which function via feed-forward and positive or negative feedback [82–84]. We predict that many of the 124 sRNAs that respond to the 18 regulatory systems will belong to new regulatory loops that link TFs to their target genes. The rapid kinetics of sRNA-mediated gene regulation [82, 83, 85–87] is likely to extend the flexibility and dynamic range of the regulatory systems of Salmonella. Future analysis of our transcriptomic data in conjunction with other datasets such as the environmentally-responsive and macrophage-regulated sRNA gene expression profiles [9, 14], global sRNA target identification and chromatin immunoprecipitation-derived TF binding sites will complete the picture of mixed regulatory interactions within the Salmonella cell. Here, the focus on the SPI1- and SPI2-associated regulons has identified 13 sRNAs that share the transcriptional signatures of S. Typhimurium virulence genes, including six sRNAs that are required for successful infection. We propose that these sRNAs control aspects of the pathogenesis of S. Typhimurium. The study takes us a step closer to the goal of elucidating the high-precision regulatory map of S. Typhimurium that allows this dangerous pathogen to cause hundreds of thousands of human deaths each year. Materials and Methods Bacterial strains and routine growth conditions Salmonella enterica serovar Typhimurium strain 4/74 was used throughout this study [88]. All mutant strains and plasmids used in this study are listed in Table 1. Cells were routinely cultured in Lennox broth (10 g/L tryptone, 5 g/L sodium chloride, 5 g/L yeast extract). Unless otherwise stated, overnight cultures were sub-inoculated 1:1,000 in 25 mL of Lennox broth in a 250 mL Erlenmeyer flask with appropriate antibiotics. For growth in SPI2-inducing phosphate carbon nitrogen (PCN) medium [31], the inoculum was taken from overnight Lennox broth cultures, as previously described [9]. One mL of the overnight culture was harvested by centrifugation at 13,000 rpm at room temperature. The cells were washed 3 times in pre-warmed PCN medium and sub-inoculated 1:500 in 25 mL of minimal medium in a 250 mL flask with appropriate antibiotics. All cultures were incubated at 37°C and 220 rpm in an Innova 3100 water-bath shaker (New Brunswick Scientific). Growth curves Relative growth rates of wild-type and mutant strains were determined in Lennox and PCN media using a Synergy H1 plate reader (Biotek). Overnight cultures were grown as described above. After overnight growth, strains grown in Lennox medium were diluted in fresh Lennox medium to a normalised OD600 of 0.003. Strains grown under InSPI2 conditions were washed 3 times in InSPI2 PCN medium and diluted in fresh PCN to a normalised OD600 of 0.03. Growth of each strain was assayed in a 96-well plate in a final volume of 200 μL at 37°C, with orbital shaking. Optical density was measured every 10 minutes throughout growth. Cell doubling times were calculated from the optical density measurements taken during exponential growth using www.doubling-time.com. Each strain was assayed in triplicate on 3 independent occasions. Ectopic expression analysis For ectopic expression of proteins from the arabinose-inducible PBAD promoter, overnight cultures of strains carrying the pBAD plasmid containing the cloned gene and an empty pBAD vector were set up as previously described. Overnight cultures were diluted in 25 mL of the appropriate medium in a 250 mL flask and grown to the desired OD600. Cultures were split into two 250 mL flasks and L-arabinose was added at a final concentration of 0.2% to one flask, to induce expression from the PBAD promoter, while no arabinose was added to the second culture. Induction proceeded for the indicated length of time and cells were harvested for analysis. Genetic manipulation of S. Typhimurium strain 4/74 Gene deletion mutants were generated as previously described [89]. Genes were tagged in their chromosomal location using an adapted λ Red recombineering protocol as previously described [90]. Mutations or FLAG tags were moved into clean genetic backgrounds by P22 transduction and confirmed by PCR and DNA sequencing (Source Biosciences, Dublin). To ensure antibiotic resistance cassettes would not affect transcription of downstream genes, the resistance genes were removed from mutant strains using the pCP20 plasmid [89, 91]. Antibiotic resistant mutants or tagged strains were grown to mid-log phase and cells were transformed with the FLP recombinase-harbouring pCP20 plasmid. Cells were recovered for 1 hour at 30°C with aeration and plated on ampicillin or chloramphenicol plates. Transformants were passaged at 37°C without antibiotics to cure the strain of the pCP20 plasmid. Loss of the antibiotic resistance cassette and the pCP20 plasmid were screened for using appropriate antibiotic selection plates and incubation at permissive temperatures. Sequence and ligation independent cloning (SLIC) was carried out with modifications [92]. Wild-type chromosomal DNA and plasmid DNA were used as templates to amplify the insert DNA and vector backbone respectively. One μg of DpnI-digested vector backbone and 1μg of insert DNA were treated with 5 units of T4 DNA polymerase in the absence of deoxynucleotide triphosphates to generate single strand overhangs. T4 DNA polymerase treatment proceeded at 23°C for 30 minutes in the presence of 5 mM DTT, 200 mM Urea, 1× BSA, and 1× reaction buffer (67 mM Tris-HCl pH8.8, 6.6 mM MgCl2, 1 mM DTT, 16.8 mM (NH4)2SO4). The reaction was stopped by addition of 25 mM EDTA and incubation at 75°C for 20 minutes. 100 ng of T4 DNA polymerase treated vector was mixed with an equal amount of T4 DNA polymerase treated “insert DNA” in a final volume of 10 μL. Samples were incubated at 65°C for 10 minutes followed by “Touch-down” annealing, during which the incubation temperature was reduced from 65°C to 25°C in 1°C decrements and samples were held for 1 minute at each temperature. Annealed vector and insert mixtures were then transformed into chemically competent E. coli TOP10 cells. Positive clones were selected for using appropriate antibiotic selection plates and overnight incubation at 37°C. Clones were screened by colony PCR using plasmid-specific and gene-specific primers. All constructs, gene deletions, and gene tags were confirmed by PCR and sequenced by DNA sequencing. Isolation of total RNA from S. Typhimurium 4/74 RNA was extracted from S. Typhimurium cells grown to a defined OD600 (Table 2). Total RNA was isolated from wild-type and mutant strains using TRIzol, as previously described [93]. Briefly, 5 OD600 units were removed from the culture and cellular transcription was stopped using 0.4× culture volume of a 5% phenol 95% ethanol “stop” solution. RNA was stabilised on ice, in stop solution, for at least 30 minutes before cells were harvested at 4,000 rpm for 10 minutes at 4°C. Pellets were re-suspended in 1 mL of TRIzol Reagent. 400 μL of chloroform was added and the samples were immediately and thoroughly mixed by inversion. Samples were moved to a Phase-lock tube (5 Prime) and the aqueous and organic phases were separated by centrifugation at 13,000 rpm for 15 minutes at room temperature. RNA was precipitated from the aqueous phase using isopropanol for 30 minutes at room temperature followed by centrifugation at 13,000 rpm for 30 minutes at room temperature. The RNA pellet was rinsed with 70% ethanol followed by centrifugation at 13,000 rpm for 10 minutes at room temperature. The RNA pellet was air-dried for 15 minutes and re-suspended in DEPC-treated water at 65°C with shaking at 900 rpm on a Thriller thermoshaker (Peqlab) for 5 minutes with occasional vortexing. RNA was kept on ice whenever possible and RNA was stored at -80°C. RNA quality was assessed using a 2100 Bioanalyser (Agilent). RNA to be used for cDNA library preparations was treated with 10 units of DNase I to remove any DNA present in the sample and samples were purified by phenol-chloroform extraction. Northern blotting In vitro transcription by T7 RNA polymerase was used to generate Dig-labelled riboprobes (Dig Northern Starter kit, Roche). Total RNA was electrophoresed through an 8.3 M Urea, 1× TBE, 7% polyacrylamide gel. RNA was transferred to a positively charged nylon membrane using the Biometra Fastblot B43 semi-dry blotting apparatus at a constant amplitude of 125 mA for 30 minutes at 4°C. RNA was UV-crosslinked to the membrane at 120 mJ for 2 minutes. The membrane was equilibrated in hybridisation buffer for 1 hour at 62°C in pre-warmed DIG Easy Hyb solution in a rotating hybridisation oven. Boiled riboprobe was added to the pre-hybridising solution and hybridisation proceeded at 62°C overnight. The membrane was washed twice for a total of 10 minutes in pre-heated (62°C) stringency wash buffer 1 (2× saline-sodium citrate (SSC) buffer, 0.1% SDS) at room temperature with rocking on a see-saw rocker (Stuart), followed by 2 washes for a total of 30 minutes with room temperature stringency wash buffer 2 (0.5× SSC buffer, 0.1% SDS) with rocking at room temperature. Unspecific sites on the membrane were blocked using 1× casein-based blocking solution, diluted in maleic acid buffer (0.1 M maleic acid, 0.15 M NaCl adjusted to pH 7.5 using NaOH pellets) for 30 minutes at room temperature with rocking. Alkaline phosphatase conjugated polyclonal anti-digoxigenin Fab-fragment was diluted 1:10,000 in 1× blocking buffer and immunological detection of the membrane proceeded for 30 minutes at room temperature with rocking. The membrane was then washed twice for a total of 30 minutes in wash buffer (maleic acid buffer, 0.3% Tween-20). The membrane was incubated for 5 minutes in detection buffer (0.1 M Tris-HCl, 0.1 M NaCl, pH 9.5) and CDP-star was used as the chemiluminescent substrate. Enzymatic de-phosphorylation of CDP-star by alkaline phosphatase was then visualised using an ImageQuant LAS4000 Imager. To determine if RNA samples were equally loaded on the gel, membranes were stripped and re-probed for the 5S ribosomal RNA. RNA adapter-based cDNA library construction and RNA-seq Strand-specific cDNA library preparation and high throughput cDNA sequencing (RNA-seq) of wild-type 4/74 and isogenic mutants was performed on DNase I-digested total RNA by Vertis Biotechnologie AG (Freising, Germany). No depletion or enrichment methods were used. RNA was fragmented by ultrasound. The 5’ end of each fragment was de-phosphorylated using tobacco acid pyrophosphatase (TAP) and a re-phosphorylated using polynucleotide kinase (PNK) for 5’ mono-phosphorylation. Poly(A)-tails were added to each fragment by poly(A) polymerase and an RNA adaptor, containing a 6–10 nucleotide bar-code, was ligated to the 5’-phosphate of each RNA fragment. First strand cDNA synthesis was performed using oligo (dT) priming and Moloney murine leukaemia virus reverse transcriptase (M-MLV RT). The resulting cDNA was amplified by PCR to approximately 20–30 ng/μL using a high fidelity DNA polymerase. cDNA was purified using the Agencourt AMPure XP kit (Beckman Coulter Genomics) and analysed by capillary electrophoresis. cDNA was sequenced on an Illumina HiSeq 2000 platform. Mutants were always sequenced in the same lane as their wild-type comparator. Mapping of RNA-seq data and differential expression analysis Sequence reads obtained from RNA-seq experiments were mapped to the 4/74 reference genome using the Segemehl mapping software [94]. To map reads which contained poor quality bases at the 3’ end, we used an iterative trimming approach in which nucleotides were truncated in a stepwise manner from the 3’ end until the reads mapped uniquely or until the read length fell below 20 bases. Reads that did not map uniquely to a single chromosomal location were discarded. Data were normalised using the transcripts per million (TPM) method [95, 96]. A TPM value of 10 was used as the threshold for gene expression based on TPM values of indicator genes which were previously shown not to be expressed under a particular condition [9]. Differential expression of genes between WT and isogenic mutants was calculated from TPM values. Two independent biological replicates of wild-type RNA from ESP, LSP and InSPI2 cultures were used in independent RNA-seq runs. Correlative analysis was used to demonstrate the reproducibility of the RNA extraction, cDNA library preparation and RNA-seq methods (S4 Table). During downstream analyses, independently extracted RNA was used to confirm RNA-seq-based findings by northern blot or quantitative PCR. For visualization of sequence reads in IGB and JBrowse, the read depth was adjusted in relation to the cDNA library with the lowest number of reads. This was achieved by dividing the read coverage at each genomic position by the library size and multiplying it by the size of the smallest library (ΔompR/envZ). Chromatin immunoprecipitation (ChIP) ChIP was carried out as previously described [71]. Briefly, protein-DNA complexes were cross-linked by adding formaldehyde to a final concentration of 1% in a drop-wise manner with gentle stirring at room temperature for 30 minutes. Cross-linking reactions were quenched by the addition of ice-cold glycine to a final concentration of 0.125 M for 5 minutes with gentle stirring at room temperature. The cross-linked cells were harvested by centrifugation at 4°C at 4,000 rpm for 8 minutes and were re-suspended in 600 μL of lysis buffer (50 mM Tris-HCl pH 8.1, 10 mM EDTA, 1% SDS, 1× protease inhibitor tablet stock) and incubated on ice for 10 minutes. 1.4 mL dilution buffer (20 mM Tris-HCl pH 8.1, 150 mM NaCl, 2 mM EDTA, 1% Trition X-100, 0.01% SDS, 1× protease inhibitor tablet stock) was added and the chromatin was sonicated on ice to reduce the average DNA fragment length to approximately 500 bp using an MSE Soniprep sonicator (Sanyo). The chromatin was pre-cleared by adding 50 μg of non-species specific IgG (normal rabbit IgG, Millipore). The chromatin was incubated for 1 hour at 4°C on a rotating wheel and 100 μL of protein G-agarose bead suspension was added. The chromatin was incubated for a further 3 hours at 4°C with rotation. Beads were pelleted at 4,000 rpm for 4 minutes at 4°C. 200 μL of pre-cleared chromatin was removed and stored at -20°C as “Input” DNA for downstream analysis. The remainder of the pre-cleared chromatin was used to set up Immunoprecipitation (IP) reactions. A “mock” IP reaction containing 1350 μL of chromatin and 10 μg of species specific IgG (normal mouse IgG, Millipore) was set up to measure background levels of DNA binding to antibodies and beads. Experimental IP reactions contained 1350 μL of chromatin and 10 μg of monoclonal mouse anti-FLAG M2 antibody (Sigma). IP reactions were incubated overnight at 4°C on a rotating wheel. 50 μL of protein G-agarose bead suspension was added to each IP sample and incubation continued for 3 hours at 4°C on a rotating wheel. The beads containing the bound antibody-protein-DNA complexes were carefully washed. Antibody-protein-DNA complexes were eluted from the beads at room temperature by adding 225 μL of elution buffer (100 mM NaHCO3, 1% SDS) followed by vortexing and pelleting of the beads twice. Both eluates were combined in the same tube. Input and IP samples were treated with 5 ng/μL RNase A and 0.3 M NaCl and incubated at 65°C for at least 6 hours or overnight. Protein-DNA cross-links were disrupted by treating Input and IP samples with 9 μg of proteinase K at 45°C for at least 3 hours or overnight. DNA was extracted from Input and IP samples by standard phenol-chloroform extraction followed by ethanol precipitation with yeast tRNA and glycogen as co-precipitants. DNA was re-suspended in nuclease free water at 37°C with shaking at 900 rpm for 1 hour. Input and IP DNA was analysed by quantitative PCR and quantified relative to a standard curve of chromosomal DNA. The quantity of immunoprecipitated DNA is relative to specific protein binding in that region and was calculated as a fraction of the starting amount of DNA (Input). The mock immunoprecipitate and experimental immunoprecipitate were compared to a control region which was negative for specific transcription factor binding. Assignment of fitness scores for HilD and SsrB-controlled genes Published data from two studies involving high-throughput analysis of mutant pools during infection showed the virulence-associated roles played by HilD and SsrB-controlled genes. The scoring methods used by the authors of each publication were applied independently to the relevant dataset. High-throughput sequencing of insertion sites of pools of S. Typhimurium transposon mutants was used to compare the ratio of input to output reads to determine relative fitness, following oral infections of chickens, pig or calves. A negative or positive fitness score indicates attenuation (blue) or amplification (orange) of fitness, respectively, as a result of the transposon insertion. If no output reads were identified for a particular insertion an arbitrary negative fitness score of -15 was assigned [41]. Two S. Typhimurium single gene deletion libraries were used to infect BALB/c mice via intraperitoneal infection, and were isolated from the spleen and liver; DNA microarrays were used to compare the ratio of input to output reads to determine fitness. A negative or positive fitness score indicates attenuation (blue) or amplification (orange) of fitness, respectively, as a result of the gene deletion [42]. In cases where different fitness scores were assigned to individual mutants isolated from different sites of the host, the most negative or most positive score was chosen. Hfq enrichment Enrichment for binding by Hfq was determined, as previously described [9], using published Hfq co-immunoprecipitation datasets [51, 78]. Briefly, a 5-fold enrichment factor of Hfq immunoprecipitation over a mock immunoprecipitation was used to determine if sRNAs were strongly associated with Hfq. Motif analysis of the Fur recognition site A position-specific scoring matrix (PSSM) was generated using alignment of the homologous Salmonella sequences of 15 published Fur binding sites from E. coli (available at http://arep.med.harvard.edu/ecoli_matrices/), by assigning a score to each possible base at each position within the binding site and normalising to the average G/C content in the S. Typhimurium chromosome. The PSSM was used to scan for direct Fur binding, in the 100 bp upstream of the 30 sRNA genes which were up-regulated in the Δfur mutant, using pattern searching software (available from rsat.ulb.ac.be), as 100 bp was the maximum distance reported for Fur binding upstream of the published Fur-regulated genes used for matrix assembly. The published Fur binding sequences were scanned with the same PSSM to establish a minimum threshold weighted score, below which predicted motifs were considered to be false-positives. Because the lowest weighted score of a published Fur binding site was 7, and a number of poorly conserved motifs had a score of approximately 7, we chose the more conservative threshold score of 10. Only motifs with a weighted score >10 were designated as putative Fur-binding sites. Motif analysis of the SsrB recognition site Previously published ChIP-chip analysis of the SsrB regulon identified an 18 bp palindromic sequence with an internal 7-4-7 organisation for SsrB recognition [47]. A position-specific scoring matrix (PSSM) was generated from an alignment of these previously identified SsrB-bound sites by assigning a score to each possible base at each position within the binding site and normalising to the average G/C content in the S. Typhimurium chromosome. The PSSM was used to scan approximately 500 bp upstream of the sRNAs which were down-regulated in the absence of SsrB, using pattern searching software available from rsat.ulb.ac.be. A score of 10 was used as the minimum threshold score for an SsrB recognition motif, based on the scores of defined SsrB bound sites scanned with the same PSSM. Phylogenetic analysis Investigation of the conservation of 15 transcription factors, TCS response regulators and Sigma factors was determined in 29 enterobacterial genomes using GLSEARCH. 1.00 indicates 100% sequence identity. Salmonella-specific regulators were defined as those with >90% sequence identity within the Salmonella genus and <70% sequence identity within other members of the Enterobacteriaceae family. Accession numbers The RNA-seq data generated from this study are deposited at NCBI GEO under the accession numbers GSM2091439 to GSM2091466, and can be accessed at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE79314. Supporting Information S1 Fig Schematic showing a simplified S. Typhimurium regulatory network. Schematic showing the key regulatory interactions required for S. Typhimurium pathogenicity and the control of sRNA expression. Transcriptional activation is represented by full red arrows, transcriptional repression is represented by full blue T-bars, post-transcriptional activation is represented by dotted red arrows, post-transcriptional repression is represented by dotted blue T-bars. Extracellular environmental signals are represented by wavy black arrows. Membrane-bound sensor kinases, outer membrane porins and transporter systems are shown within the cellular membrane. Cell surface appendages, such as Type III Secretion Systems, flagella, fimbriae and curli fibres are indicated outside the cell. In the case of indirect regulatory interactions, mediators of the interaction are indicated in grey boxes. The schematic is based on a number of publications, including [4, 20, 24, 29, 39, 56, 97] and others. Some regulatory interactions have been omitted for ease of viewing. (TIF) Click here for additional data file. S2 Fig Overlap between the SsrA, SsrB and SsrAB regulons. Venn diagram showing the high level of similarity between the differentially-expressed genes in the single ΔssrA, ΔssrB and the double ΔssrAB mutants compared to the wild-type strain grown under SPI2-inducing conditions. Individual gene lists are available in S3 Table. (TIF) Click here for additional data file. S3 Fig Chromosomal location and Hfq-enrichment of differentially-expressed sRNAs. Stacked bar chart showing the number of differentially-expressed sRNAs (>3-fold change in expression in panel of regulatory mutants, compared to wild-type) in terms of their chromosomal location relative to nearby coding genes and in the context of enrichment for Hfq, as determined by Chao et al (2012) using Hfq co-immunoprecipitation [51]. The height of each bar reflects the number of sRNAs in each category. (TIF) Click here for additional data file. S4 Fig Variation in size of regulons comprising sRNAs and CDS. (A) Stacked bar chart showing the numbers of differentially-expressed sRNAs and (B) the numbers of differentially-expressed CDS in the panel of regulatory mutants. Red indicates the number of genes which show decreased expression (activated by regulatory system) in the mutant strain. Blue indicates the number of genes which show increased expression (repressed by regulatory system) in the mutant strain. In each case the comparator is the wild-type strain grown under the same environmental condition. (TIF) Click here for additional data file. S5 Fig Predicted binding of Fur within sRNA promoter regions. (A) Motif consensus logo for Fur recognition motif was generated with software available from http://weblogo.berkeley.edu [98], using published Fur binding sites (http://arep.med.harvard.edu/ecoli_matrices/). (B) Consensus sequence for Fur recognition aligned with putative Fur recognition sites within Fur-regulated sRNA promoters. (C) Genetic organisation of the STnc4000 and STnc3250 promoters demonstrating locations of the putative Fur recognition sites. (TIF) Click here for additional data file. S6 Fig Predicted RpoS-dependent promoters. Multiple alignment of the promoters of RpoS-dependent sRNAs that contain conserved hallmarks of RpoS recognition [66]. Some promoter features that favour RpoS recognition are indicated in red above the alignment. Within the sequence alignment: Red >75% nucleotide sequence identity; Blue >35% <75% nucleotide sequence identity; Black <35% nucleotide sequence identity between aligned promoter sequences. (TIF) Click here for additional data file. S7 Fig The SPI1-like sRNA STnc520 is directly regulated by the SPI1-encoded transcription factor SprB. (A) Northern blot showing expression of STnc520 in the wild-type Salmonella Typhimurium 4/74 and isogenic mutants of SPI1-encoded or SPI1-associated transcription factors. 5S RNA was probed as a loading control. (B) Chromatin immunoprecipitation (ChIP) followed by qPCR demonstrates SprB binding to the STnc520 promoter. There is strong enrichment (approximately 14-fold) of the STnc520 promoter region in the experimental (FLAG) ChIP DNA, compared to the background “mock” ChIP DNA. The negative control gene, hemX, displayed little enrichment in the experimental ChIP DNA sample, compared to the mock DNA. Error bars are based on the standard deviation from 2 independent biological replicates. (TIF) Click here for additional data file. S8 Fig Predicted binding of SsrB within the STnc3020 promoter region. (A) Motif consensus logo for SsrB recognition motif was generated using Weblogo software (http://weblogo.berkeley.edu) [98] using SsrB bound sites determined by ChIP-chip analysis [47]. (B) Consensus sequence for SsrB recognition aligned with putative SsrB recognition site within the STnc3020 promoter. (C) Genetic organisation of the STnc3020 promoter demonstrating the location of the predicted SsrB recognition site. (TIF) Click here for additional data file. S1 Table Dataset 1: Doubling times of wild-type and mutant strains in relevant growth media. (XLSX) Click here for additional data file. S2 Table Dataset 2: RNA-seq statistics for 3 Illumina HiSeq experiments. (XLSX) Click here for additional data file. S3 Table Dataset 3–22: Differentially-expressed CDS (excluding deleted CDS for each mutant) and sRNA genes in all mutant strains. (XLSX) Click here for additional data file. S4 Table Dataset 23: Pearson correlation coefficients showing the high reproducibility of RNA-seq data between independent biological replicates. (XLSX) Click here for additional data file. S5 Table Dataset 24: Phylogenetic analysis of 15 transcription factors. Dataset 25: Phylogenetic analysis of 13 SPI1 and SPI2-like sRNAs. Dataset 26: Phylogenetic analysis of 30 Hfq-associated sRNAs. (XLSX) Click here for additional data file. S6 Table Dataset 27: Regulatory inputs to differentially-expressed sRNAs. Dataset 28: chromosomal location of differentially-expressed sRNAS. Dataset 29: 3’UTR sRNAs. Dataset 30: Hfq associated sRNAs. (XLSX) Click here for additional data file. S1 Text Discussion of the mutations used in this study that showed minor polar effects and the overlaps between virulence gene expression profiles from this study with the literature. (DOCX) Click here for additional data file. We are very grateful to Jane Twohig for excellent technical support and to Méabh Henry-Bezy and Michael Beckett for experimental assistance. We thank Fritz Thϋmmler (Vertis Biotechnologie AG) for the construction of high-quality cDNA libraries. We thank Pepe Casadesús and Joe Wade for sharing strains and unpublished data, Ian R. Monk for help with SLIC experiments, and Jörg Vogel’s lab for sharing strains. Shabarinath Srikumar, Kristian Händler, Charles Dorman, Shane Dillon, Andrew Cameron, Heather Quinn, Rocίo Canals, Disa L. Hammarlöf, Siân Owen, Nicolas Wenner and Kai Papenfort provided helpful comments during the experimental and writing phases of this project. ==== Refs References 1 Majowicz SE , Musto J , Scallan E , Angulo FJ , Kirk M , O'Brien SJ , et al The global burden of nontyphoidal Salmonella gastroenteritis . Clin Infect Dis . 2010 ;50 (6 ):882 –9 . Epub 2010/02/18. 10.1086/650733 20158401 2 Ao TT , Feasey NA , Gordon MA , Keddy KH , Angulo FJ , Crump JA . Global burden of invasive nontyphoidal Salmonella disease, 2010(1) . Emerg Infect Dis . 2015 ;21 (6 ). Epub 2015/04/11. 3 Feasey NA , Dougan G , Kingsley RA , Heyderman RS , Gordon MA . Invasive non-typhoidal Salmonella disease: an emerging and neglected tropical disease in Africa . Lancet . 2012 ;379 (9835 ):2489 –99 . Epub 2012/05/17. 10.1016/S0140-6736(11)61752-2 22587967 4 Ellermeier JR , Slauch JM . Adaptation to the host environment: regulation of the SPI1 type III secretion system in Salmonella enterica serovar Typhimurium . Curr Opin Microbiol . 2007 ;10 (1 ):24 –9 . Epub 2007/01/09. 17208038 5 Hensel M . Salmonella pathogenicity island 2 . Mol Microbiol . 2000 ;36 (5 ):1015 –23 . Epub 2000/06/09. 10844687 6 Kaiser P , Diard M , Stecher B , Hardt WD . The streptomycin mouse model for Salmonella diarrhea: functional analysis of the microbiota, the pathogen's virulence factors, and the host's mucosal immune response . Immunol Rev . 2012 ;245 (1 ):56 –83 . Epub 2011/12/16. 10.1111/j.1600-065X.2011.01070.x 22168414 7 Eriksson S , Lucchini S , Thompson A , Rhen M , Hinton JC . Unravelling the biology of macrophage infection by gene expression profiling of intracellular Salmonella enterica . Mol Microbiol . 2003 ;47 (1 ):103 –18 . Epub 2002/12/21. 12492857 8 Hautefort I , Thompson A , Eriksson-Ygberg S , Parker ML , Lucchini S , Danino V , et al During infection of epithelial cells Salmonella enterica serovar Typhimurium undergoes a time-dependent transcriptional adaptation that results in simultaneous expression of three type 3 secretion systems . Cell Microbiol . 2008 ;10 (4 ):958 –84 . Epub 2007/11/23. 18031307 9 Kröger C , Colgan A , Srikumar S , Handler K , Sivasankaran SK , Hammarlof DL , et al An infection-relevant transcriptomic compendium for Salmonella enterica serovar Typhimurium . Cell Host Microbe . 2013 ;14 (6 ):683 –95 . Epub 2013/12/18. 10.1016/j.chom.2013.11.010 24331466 10 Ackermann M , Stecher B , Freed NE , Songhet P , Hardt WD , Doebeli M . Self-destructive cooperation mediated by phenotypic noise . Nature . 2008 ;454 (7207 ):987 –90 . Epub 2008/08/23. 10.1038/nature07067 18719588 11 Sturm A , Heinemann M , Arnoldini M , Benecke A , Ackermann M , Benz M , et al The cost of virulence: retarded growth of Salmonella Typhimurium cells expressing type III secretion system 1 . PLoS Pathog . 2011 ;7 (7 ):e1002143 Epub 2011/08/11. 10.1371/journal.ppat.1002143 21829349 12 Diard M , Garcia V , Maier L , Remus-Emsermann MN , Regoes RR , Ackermann M , et al Stabilization of cooperative virulence by the expression of an avirulent phenotype . Nature . 2013 ;494 (7437 ):353 –6 . Epub 2013/02/22. 10.1038/nature11913 23426324 13 Avraham R , Haseley N , Brown D , Penaranda C , Jijon HB , Trombetta JJ , et al Pathogen Cell-to-Cell Variability Drives Heterogeneity in Host Immune Responses . Cell . 2015 ;162 (6 ):1309 –21 . Epub 2015/09/08. 10.1016/j.cell.2015.08.027 26343579 14 Srikumar S , Kroger C , Hebrard M , Colgan A , Owen SV , Sivasankaran SK , et al RNA-seq Brings New Insights to the Intra-Macrophage Transcriptome of Salmonella Typhimurium . PLoS Pathog . 2015 ;11 (11 ):e1005262 Epub 2015/11/13. 10.1371/journal.ppat.1005262 26561851 15 Altuvia S , Wagner EG . Switching on and off with RNA . Proc Natl Acad Sci U S A . 2000 ;97 (18 ):9824 –6 . Epub 2000/08/30. 10963653 16 Vogel J . A rough guide to the non-coding RNA world of Salmonella . Mol Microbiol . 2009 ;71 (1 ):1 –11 . Epub 2008/11/15. 10.1111/j.1365-2958.2008.06505.x 19007416 17 Teplitski M , Goodier RI , Ahmer BM . Pathways leading from BarA/SirA to motility and virulence gene expression in Salmonella . J Bacteriol . 2003 ;185 (24 ):7257 –65 . Epub 2003/12/04. 14645287 18 Fortune DR , Suyemoto M , Altier C . Identification of CsrC and characterization of its role in epithelial cell invasion in Salmonella enterica serovar Typhimurium . Infect Immun . 2006 ;74 (1 ):331 –9 . Epub 2005/12/22. 16368988 19 Ellermeier JR , Slauch JM . Fur regulates expression of the Salmonella pathogenicity island 1 type III secretion system through HilD . J Bacteriol . 2008 ;190 (2 ):476 –86 . Epub 2007/11/13. 17993530 20 Troxell B , Sikes ML , Fink RC , Vazquez-Torres A , Jones-Carson J , Hassan HM . Fur negatively regulates hns and is required for the expression of HilA and virulence in Salmonella enterica serovar Typhimurium . J Bacteriol . 2011 ;193 (2 ):497 –505 . Epub 2010/11/16. 10.1128/JB.00942-10 21075923 21 Teixido L , Carrasco B , Alonso JC , Barbe J , Campoy S . Fur activates the expression of Salmonella enterica pathogenicity island 1 by directly interacting with the hilD operator in vivo and in vitro . PloS One . 2011 ;6 (5 ):e19711 Epub 2011/05/17. 10.1371/journal.pone.0019711 21573071 22 Kage H , Takaya A , Ohya M , Yamamoto T . Coordinated regulation of expression of Salmonella pathogenicity island 1 and flagellar type III secretion systems by ATP-dependent ClpXP protease . J Bacteriol . 2008 ;190 (7 ):2470 –8 . Epub 2008/02/05. 10.1128/JB.01385-07 18245288 23 Chubiz JE , Golubeva YA , Lin D , Miller LD , Slauch JM . FliZ regulates expression of the Salmonella pathogenicity island 1 invasion locus by controlling HilD protein activity in Salmonella enterica serovar Typhimurium . J Bacteriol . 2010 ;192 (23 ):6261 –70 . Epub 2010/10/05. 10.1128/JB.00635-10 20889744 24 Golubeva YA , Sadik AY , Ellermeier JR , Slauch JM . Integrating global regulatory input into the Salmonella pathogenicity island 1 type III secretion system . Genetics . 2012 ;190 (1 ):79 –90 . Epub 2011/10/25. 10.1534/genetics.111.132779 22021388 25 Cirillo DM , Valdivia RH , Monack DM , Falkow S . Macrophage-dependent induction of the Salmonella pathogenicity island 2 type III secretion system and its role in intracellular survival . Mol Microbiol . 1998 ;30 (1 ):175 –88 . Epub 1998/10/24. 9786194 26 Feng X , Oropeza R , Kenney LJ . Dual regulation by phospho-OmpR of ssrA/B gene expression in Salmonella pathogenicity island 2 . Mol Microbiol . 2003 ;48 (4 ):1131 –43 . Epub 2003/05/20. 12753201 27 Kato A , Groisman EA . The PhoQ/PhoP regulatory network of Salmonella enterica . Adv Exp Med Biol . 2008 ;631 :7 –21 . Epub 2008/09/17. 18792679 28 Linehan SA , Rytkonen A , Yu XJ , Liu M , Holden DW . SlyA regulates function of Salmonella pathogenicity island 2 (SPI-2) and expression of SPI-2-associated genes . Infect Immun . 2005 ;73 (7 ):4354 –62 . Epub 2005/06/24. 15972530 29 Baxter MA , Jones BD . Two-Component Regulators Control hilA Expression by Controlling fimZ and hilE Expression within Salmonella enterica Serovar Typhimurium . Infect Immun . 2015 ;83 (3 ):978 –85 . Epub 2014/12/31. 10.1128/IAI.02506-14 25547794 30 Bustamante VH , Martinez LC , Santana FJ , Knodler LA , Steele-Mortimer O , Puente JL . HilD-mediated transcriptional cross-talk between SPI-1 and SPI-2 . Proc Natl Acad Sci U S A . 2008 ;105 (38 ):14591 –6 . Epub 2008/09/19. 10.1073/pnas.0801205105 18799744 31 Lober S , Jackel D , Kaiser N , Hensel M . Regulation of Salmonella pathogenicity island 2 genes by independent environmental signals . Int J Med Microbiol 2006 ;296 (7 ):435 –47 . Epub 2006/08/15. 16904940 32 Fang FC , Libby SJ , Buchmeier NA , Loewen PC , Switala J , Harwood J , et al The alternative sigma factor KatF (RpoS) regulates Salmonella virulence . Proc Natl Acad Sci U S A . 1992 ;89 (24 ):11978 –82 . Epub 1992/12/15. 1465428 33 Kowarz L , Coynault C , Robbe-Saule V , Norel F . The Salmonella Typhimurium katF (rpoS) gene: cloning, nucleotide sequence, and regulation of spvR and spvABCD virulence plasmid genes . J Bacteriol . 1994 ;176 (22 ):6852 –60 . Epub 1994/11/01. 7961444 34 Humphreys S , Stevenson A , Bacon A , Weinhardt AB , Roberts M . The alternative sigma factor, sigmaE, is critically important for the virulence of Salmonella Typhimurium . Infect Immun . 1999 ;67 (4 ):1560 –8 . Epub 1999/03/20. 10084987 35 Li J , Overall CC , Nakayasu ES , Kidwai AS , Jones MB , Johnson RC , et al Analysis of the Salmonella regulatory network suggests involvement of SsrB and H-NS in Eσ-regulated SPI-2 gene expression . Frontiers Microbiol . 2015 ;6 (27 ). 36 Garcia-Del Portillo F , Pucciarelli MG , Casadesus J . DNA adenine methylase mutants of Salmonella Typhimurium show defects in protein secretion, cell invasion, and M cell cytotoxicity . Proc Natl Acad Sci U S A . 1999 ;96 (20 ):11578 –83 . Epub 1999/09/29. 10500219 37 Balbontin R , Rowley G , Pucciarelli MG , Lopez-Garrido J , Wormstone Y , Lucchini S , et al DNA adenine methylation regulates virulence gene expression in Salmonella enterica serovar Typhimurium . J Bacteriol . 2006 ;188 (23 ):8160 –8 . Epub 2006/09/26. 16997949 38 Lopez-Garrido J , Casadesus J . Regulation of Salmonella enterica pathogenicity island 1 by DNA adenine methylation . Genetics . 2010 ;184 (3 ):637 –49 . Epub 2009/12/17. 10.1534/genetics.109.108985 20008574 39 Erhardt M , Dersch P . Regulatory principles governing Salmonella and Yersinia virulence . Frontiers Microbiol . 2015 ;6 :949 . Epub 2015/10/07. 40 Frye J , Karlinsey JE , Felise HR , Marzolf B , Dowidar N , McClelland M , et al Identification of new flagellar genes of Salmonella enterica serovar Typhimurium . J Bacteriol . 2006 ;188 (6 ):2233 –43 . Epub 2006/03/04. 16513753 41 Chaudhuri RR , Morgan E , Peters SE , Pleasance SJ , Hudson DL , Davies HM , et al Comprehensive assignment of roles for Salmonella Typhimurium genes in intestinal colonization of food-producing animals . PLoS Genet . 2013 ;9 (4 ):e1003456 Epub 2013/05/03. 10.1371/journal.pgen.1003456 23637626 42 Silva-Valenzuela CA , Molina-Quiroz RC , Desai P , Valenzuela C , Porwollik S , Zhao M , et al Analysis of Two Complementary Single-Gene Deletion Mutant Libraries of Salmonella Typhimurium in Intraperitoneal Infection of BALB/c Mice . Frontiers Microbiol . 2015 ;6 :1455 . Epub 2016/01/19. 43 Niemann GS , Brown RN , Gustin JK , Stufkens A , Shaikh-Kidwai AS , Li J , et al Discovery of novel secreted virulence factors from Salmonella enterica serovar Typhimurium by proteomic analysis of culture supernatants . Infect Immun . 2011 ;79 (1 ):33 –43 . Epub 2010/10/27. 10.1128/IAI.00771-10 20974834 44 Navarre WW , Halsey TA , Walthers D , Frye J , McClelland M , Potter JL , et al Co-regulation of Salmonella enterica genes required for virulence and resistance to antimicrobial peptides by SlyA and PhoP/PhoQ . Mol Microbiol . 2005 ;56 (2 ):492 –508 . Epub 2005/04/09. 15813739 45 Troxell B , Fink RC , Porwollik S , McClelland M , Hassan HM . The Fur regulon in anaerobically grown Salmonella enterica sv. Typhimurium: identification of new Fur targets . BMC Microbiol . 2011 ;11 :236 Epub 2011/10/25. 10.1186/1471-2180-11-236 22017966 46 Thijs IM , De Keersmaecker SC , Fadda A , Engelen K , Zhao H , McClelland M , et al Delineation of the Salmonella enterica serovar Typhimurium HilA regulon through genome-wide location and transcript analysis . J Bacteriol . 2007 ;189 (13 ):4587 –96 . Epub 2007/05/08. 17483226 47 Tomljenovic-Berube AM , Mulder DT , Whiteside MD , Brinkman FS , Coombes BK . Identification of the regulatory logic controlling Salmonella pathoadaptation by the SsrA-SsrB two-component system . PLoS Genet . 2010 ;6 (3 ):e1000875 Epub 2010/03/20. 10.1371/journal.pgen.1000875 20300643 48 Rytkonen A , Poh J , Garmendia J , Boyle C , Thompson A , Liu M , et al SseL, a Salmonella deubiquitinase required for macrophage killing and virulence . Proc Natl Acad Sci U S A . 2007 ;104 (9 ):3502 –7 . Epub 2007/03/16. 17360673 49 Yoon H , McDermott JE , Porwollik S , McClelland M , Heffron F . Coordinated regulation of virulence during systemic infection of Salmonella enterica serovar Typhimurium . PLoS Pathog . 2009 ;5 (2 ):e1000306 Epub 2009/02/21. 10.1371/journal.ppat.1000306 19229334 50 Perkins TT , Davies MR , Klemm EJ , Rowley G , Wileman T , James K , et al ChIP-seq and transcriptome analysis of the OmpR regulon of Salmonella enterica serovars Typhi and Typhimurium reveals accessory genes implicated in host colonization . Mol Microbiol . 2013 ;87 (3 ):526 –38 . Epub 2012/11/30. 10.1111/mmi.12111 23190111 51 Chao Y , Papenfort K , Reinhardt R , Sharma CM , Vogel J . An atlas of Hfq-bound transcripts reveals 3' UTRs as a genomic reservoir of regulatory small RNAs . EMBO J . 2012 ;31 (20 ):4005 –19 . Epub 2012/08/28. 10.1038/emboj.2012.229 22922465 52 Baxter MA , Fahlen TF , Wilson RL , Jones BD . HilE interacts with HilD and negatively regulates hilA transcription and expression of the Salmonella enterica serovar Typhimurium invasive phenotype . Infect Immun . 2003 ;71 (3 ):1295 –305 . Epub 2003/02/22. 12595445 53 Carpenter BM , Whitmire JM , Merrell DS . This is not your mother's repressor: the complex role of Fur in pathogenesis . Infect Immun . 2009 ;77 (7 ):2590 –601 . Epub 2009/04/15. 10.1128/IAI.00116-09 19364842 54 Padalon-Brauch G , Hershberg R , Elgrably-Weiss M , Baruch K , Rosenshine I , Margalit H , et al Small RNAs encoded within genetic islands of Salmonella Typhimurium show host-induced expression and role in virulence . Nucleic Acids Res . 2008 ;36 (6 ):1913 –27 . Epub 2008/02/13. 10.1093/nar/gkn050 18267966 55 Escolar L , Perez-Martin J , de Lorenzo V . Coordinated repression in vitro of the divergent fepA-fes promoters of Escherichia coli by the iron uptake regulation (Fur) protein . J Bacteriol . 1998 ;180 (9 ):2579 –82 . Epub 1998/05/09. 9573216 56 Fass E , Groisman EA . Control of Salmonella pathogenicity island-2 gene expression . Curr Opin Microbiol . 2009 ;12 (2 ):199 –204 . Epub 2009/03/07. 10.1016/j.mib.2009.01.004 19264535 57 Quinn HJ , Cameron AD , Dorman CJ . Bacterial regulon evolution: distinct responses and roles for the identical OmpR proteins of Salmonella Typhimurium and Escherichia coli in the acid stress response . PLoS Genet . 2014 ;10 (3 ):e1004215 Epub 2014/03/08. 10.1371/journal.pgen.1004215 24603618 58 Monsieurs P , De Keersmaecker S , Navarre WW , Bader MW , De Smet F , McClelland M , et al Comparison of the PhoPQ regulon in Escherichia coli and Salmonella Typhimurium . J Mol Evol . 2005 ;60 (4 ):462 –74 . Epub 2005/05/11. 15883881 59 Wanner BL . Phosphorous assimilation and control of the phosphate regulon in Escherichia coli and Salmonella Typhimurium Cell Mol Biol . 2 ed : ASM Press ; 1996 p. 1357 –81 . 60 Hengge-Aronis R . Signal transduction and regulatory mechanisms involved in control of the sigma(S) (RpoS) subunit of RNA polymerase . Microbiol Mol Biol R . 2002 ;66 (3 ):373 –95 . Epub 2002/09/05. 61 Frohlich KS , Papenfort K , Berger AA , Vogel J . A conserved RpoS-dependent small RNA controls the synthesis of major porin OmpD . Nucleic Acids Res . 2012 ;40 (8 ):3623 –40 . Epub 2011/12/20. 10.1093/nar/gkr1156 22180532 62 Silva IJ , Ortega AD , Viegas SC , Garcia-Del Portillo F , Arraiano CM . An RpoS-dependent sRNA regulates the expression of a chaperone involved in protein folding . RNA . 2013 ;19 (9 ):1253 –65 . Epub 2013/07/31. 10.1261/rna.039537.113 23893734 63 Levi-Meyrueis C , Monteil V , Sismeiro O , Dillies MA , Monot M , Jagla B , et al Expanding the RpoS/sigmaS-Network by RNA Sequencing and Identification of sigmaS-Controlled Small RNAs in Salmonella . PloS One . 2014 ;9 (5 ):e96918 Epub 2014/05/09. 10.1371/journal.pone.0096918 24810289 64 Peano C , Wolf J , Demol J , Rossi E , Petiti L , De Bellis G , et al Characterization of the Escherichia coli sigma(S) core regulon by Chromatin Immunoprecipitation-sequencing (ChIP-seq) analysis . Sci Rep . 2015 ;5 :10469 Epub 2015/05/29. 10.1038/srep10469 26020590 65 Becker G , Hengge-Aronis R . What makes an Escherichia coli promoter sigma(S) dependent? Role of the -13/-14 nucleotide promoter positions and region 2.5 of sigma(S) . Mol Microbiol . 2001 ;39 (5 ):1153 –65 . Epub 2001/03/17. 11251833 66 Typas A , Becker G , Hengge R . The molecular basis of selective promoter activation by the sigmaS subunit of RNA polymerase . Mol Microbiol . 2007 ;63 (5 ):1296 –306 . Epub 2007/02/17. 17302812 67 Gottesman S . Micros for microbes: non-coding regulatory RNAs in bacteria . Trends Genet . 2005 ;21 (7 ):399 –404 . Epub 2005/05/26. 15913835 68 Guo MS , Updegrove TB , Gogol EB , Shabalina SA , Gross CA , Storz G . MicL, a new sigmaE-dependent sRNA, combats envelope stress by repressing synthesis of Lpp, the major outer membrane lipoprotein . Genes Dev . 2014 ;28 (14 ):1620 –34 . Epub 2014/07/18. 10.1101/gad.243485.114 25030700 69 Shi Y , Latifi T , Cromie MJ , Groisman EA . Transcriptional control of the antimicrobial peptide resistance ugtL gene by the Salmonella PhoP and SlyA regulatory proteins . J Biol Chem . 2004 ;279 (37 ):38618 –25 . Epub 2004/06/23. 15208313 70 Perez JC , Latifi T , Groisman EA . Overcoming H-NS-mediated transcriptional silencing of horizontally acquired genes by the PhoP and SlyA proteins in Salmonella enterica . J Biol Chem . 2008 ;283 (16 ):10773 –83 . Epub 2008/02/14. 10.1074/jbc.M709843200 18270203 71 Dillon SC , Cameron AD , Hokamp K , Lucchini S , Hinton JC , Dorman CJ . Genome-wide analysis of the H-NS and Sfh regulatory networks in Salmonella Typhimurium identifies a plasmid-encoded transcription silencing mechanism . Mol Microbiol . 2010 ;76 (5 ):1250 –65 . Epub 2010/05/07. 10.1111/j.1365-2958.2010.07173.x 20444106 72 Quackenbush J . Genomics. Microarrays—guilt by association . Science . 2003 ;302 (5643 ):240 –1 . Epub 2003/10/11. 14551426 73 Cloots L , Marchal K . Network-based functional modeling of genomics, transcriptomics and metabolism in bacteria . Curr Opin Microbiol . 2011 ;14 (5 ):599 –607 . Epub 2011/09/29. 10.1016/j.mib.2011.09.003 21943683 74 De Smet R , Marchal K . Advantages and limitations of current network inference methods . Nat Rev Microbiol . 2010 ;8 (10 ):717 –29 . Epub 2010/09/02. 10.1038/nrmicro2419 20805835 75 Pfeiffer V , Sittka A , Tomer R , Tedin K , Brinkmann V , Vogel J . A small non-coding RNA of the invasion gene island (SPI-1) represses outer membrane protein synthesis from the Salmonella core genome . Mol Microbiol . 2007 ;66 (5 ):1174 –91 . Epub 2007/11/01. 17971080 76 Westermann AJ , Forstner KU , Amman F , Barquist L , Chao Y , Schulte LN , et al Dual RNA-seq unveils noncoding RNA functions in host-pathogen interactions . Nature . 2016 . Epub 2016/01/21. 77 Vogel J , Luisi BF . Hfq and its constellation of RNA . Nat Rev Microbiol . 2011 ;9 (8 ):578 –89 . Epub 2011/07/16. 10.1038/nrmicro2615 21760622 78 Sittka A , Lucchini S , Papenfort K , Sharma CM , Rolle K , Binnewies TT , et al Deep sequencing analysis of small noncoding RNA and mRNA targets of the global post-transcriptional regulator, Hfq . PLoS Genet . 2008 ;4 (8 ):e1000163 Epub 2008/08/30. 10.1371/journal.pgen.1000163 18725932 79 Wagner EG . Cycling of RNAs on Hfq . RNA Biol . 2013 ;10 (4 ):619 –26 . Epub 2013/03/08. 10.4161/rna.24044 23466677 80 Papenfort K , Pfeiffer V , Mika F , Lucchini S , Hinton JC , Vogel J . SigmaE-dependent small RNAs of Salmonella respond to membrane stress by accelerating global omp mRNA decay . Mol Microbiol . 2006 ;62 (6 ):1674 –88 . Epub 2007/04/12. 17427289 81 Figueroa-Bossi N , Lemire S , Maloriol D , Balbontin R , Casadesus J , Bossi L . Loss of Hfq activates the sigmaE-dependent envelope stress response in Salmonella enterica . Mol Microbiol . 2006 ;62 (3 ):838 –52 . Epub 2006/09/27. 16999834 82 Mandin P , Guillier M . Expanding control in bacteria: interplay between small RNAs and transcriptional regulators to control gene expression . Curr Opin Microbiol . 2013 ;16 (2 ):125 –32 . Epub 2013/02/19. 10.1016/j.mib.2012.12.005 23415757 83 Gopel Y , Gorke B . Rewiring two-component signal transduction with small RNAs . Curr Opin Microbiol . 2012 ;15 (2 ):132 –9 . Epub 2011/12/27. 10.1016/j.mib.2011.12.001 22197250 84 Papenfort K , Espinosa E , Casadesus J , Vogel J . Small RNA-based feedforward loop with AND-gate logic regulates extrachromosomal DNA transfer in Salmonella . Proc Natl Acad Sci U S A . 2015 ;112 (34 ):E4772 –81 . Epub 2015/08/27. 10.1073/pnas.1507825112 26307765 85 Beisel CL , Storz G . Base pairing small RNAs and their roles in global regulatory networks . FEMS Microbiol Rev . 2010 ;34 (5 ):866 –82 . Epub 2010/07/29. 10.1111/j.1574-6976.2010.00241.x 20662934 86 Adamson DN , Lim HN . Rapid and robust signaling in the CsrA cascade via RNA-protein interactions and feedback regulation . Proc Natl Acad Sci U S A . 2013 ;110 (32 ):13120 –5 . Epub 2013/07/24. 10.1073/pnas.1308476110 23878244 87 Shimoni Y , Friedlander G , Hetzroni G , Niv G , Altuvia S , Biham O , et al Regulation of gene expression by small non-coding RNAs: a quantitative view . Mol Syst Biol . 2007 ;3 :138 Epub 2007/09/26. 17893699 88 Wray C , Sojka WJ . Experimental Salmonella Typhimurium infection in calves . Res Vet Sci . 1978 ;25 (2 ):139 –43 . Epub 1978/09/01. 364573 89 Datsenko KA , Wanner BL . One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products . Proc Natl Acad Sci U S A . 2000 ;97 (12 ):6640 –5 . Epub 2000/06/01. 10829079 90 Uzzau S , Figueroa-Bossi N , Rubino S , Bossi L . Epitope tagging of chromosomal genes in Salmonella . Proc Natl Acad Sci U S A . 2001 ;98 (26 ):15264 –9 . Epub 2001/12/14. 11742086 91 Cherepanov PP , Wackernagel W . Gene disruption in Escherichia coli: TcR and KmR cassettes with the option of Flp-catalyzed excision of the antibiotic-resistance determinant . Gene . 1995 ;158 (1 ):9 –14 . Epub 1995/05/26. 7789817 92 Li MZ , Elledge SJ . Harnessing homologous recombination in vitro to generate recombinant DNA via SLIC . Nat Methods . 2007 ;4 (3 ):251 –6 . Epub 2007/02/13. 17293868 93 Kröger C , Dillon SC , Cameron AD , Papenfort K , Sivasankaran SK , Hokamp K , et al The transcriptional landscape and small RNAs of Salmonella enterica serovar Typhimurium . Proc Natl Acad Sci U S A . 2012 ;109 (20 ):E1277 –86 . Epub 2012/04/28. 10.1073/pnas.1201061109 22538806 94 Hoffmann S , Otto C , Kurtz S , Sharma CM , Khaitovich P , Vogel J , et al Fast mapping of short sequences with mismatches, insertions and deletions using index structures . PLoS Comput Biol . 2009 ;5 (9 ):e1000502 Epub 2009/09/15. 10.1371/journal.pcbi.1000502 19750212 95 Wagner GP , Kin K , Lynch VJ . Measurement of mRNA abundance using RNA-seq data: RPKM measure is inconsistent among samples . Theor Biosci . 2012 ;131 (4 ):281 –5 . Epub 2012/08/09. 96 Li B , Ruotti V , Stewart RM , Thomson JA , Dewey CN . RNA-Seq gene expression estimation with read mapping uncertainty . Bioinformatics . 2010 ;26 (4 ):493 –500 . Epub 2009/12/22. 10.1093/bioinformatics/btp692 20022975 97 Fabrega A , Vila J . Salmonella enterica serovar Typhimurium skills to succeed in the host: virulence and regulation . Clin Microbiol Rev . 2013 ;26 (2 ):308 –41 . Epub 2013/04/05. 10.1128/CMR.00066-12 23554419 98 Crooks GE , Hon G , Chandonia JM , Brenner SE . WebLogo: a sequence logo generator . Genome Res . 2004 ;14 (6 ):1188 –90 . Epub 2004/06/03. 15173120 99 Cameron AD , Dorman CJ . A fundamental regulatory mechanism operating through OmpR and DNA topology controls expression of Salmonella pathogenicity islands SPI-1 and SPI-2 . PLoS Genet . 2012 ;8 (3 ):e1002615 Epub 2012/03/30. 10.1371/journal.pgen.1002615 22457642 100 Falconi M , McGovern V , Gualerzi C , Hillyard D , Higgins NP . Mutations altering chromosomal protein H-NS induce mini-Mu transposition . The New biologist . 1991 ;3 (6 ):615 –25 . Epub 1991/06/01. 1655013 101 Low DA , Casadesus J . Clocks and switches: bacterial gene regulation by DNA adenine methylation . Curr Opin Microbiol . 2008 ;11 (2 ):106 –12 . Epub 2008/04/09. 10.1016/j.mib.2008.02.012 18396448 102 Ellermeier CD , Ellermeier JR , Slauch JM . HilD, HilC and RtsA constitute a feed forward loop that controls expression of the SPI1 type three secretion system regulator hilA in Salmonella enterica serovar Typhimurium . Mol Microbiol . 2005 ;57 (3 ):691 –705 . Epub 2005/07/28. 16045614 103 Lucas RL , Lostroh CP , DiRusso CC , Spector MP , Wanner BL , Lee CA . Multiple factors independently regulate hilA and invasion gene expression in Salmonella enterica serovar Typhimurium . J Bacteriol . 2000 ;182 (7 ):1872 –82 . Epub 2000/03/14. 10714991 104 Kadner RJ . Regulation by iron: RNA rules the rust . J Bacteriol . 2005 ;187 (20 ):6870 –3 . Epub 2005/10/04. 16199555 105 Martinez LC , Yakhnin H , Camacho MI , Georgellis D , Babitzke P , Puente JL , et al Integration of a complex regulatory cascade involving the SirA/BarA and Csr global regulatory systems that controls expression of the Salmonella SPI-1 and SPI-2 virulence regulons through HilD . Mol Microbiol . 2011 ;80 (6 ):1637 –56 . Epub 2011/04/27. 10.1111/j.1365-2958.2011.07674.x 21518393 106 Zere TR , Vakulskas CA , Leng Y , Pannuri A , Potts AH , Dias R , et al Genomic Targets and Features of BarA-UvrY (-SirA) Signal Transduction Systems . PloS One . 2015 ;10 (12 ):e0145035 Epub 2015/12/18. 10.1371/journal.pone.0145035 26673755 107 Sittka A , Pfeiffer V , Tedin K , Vogel J . The RNA chaperone Hfq is essential for the virulence of Salmonella Typhimurium . Mol Microbiol . 2007 ;63 (1 ):193 –217 . Epub 2006/12/14. 17163975 108 Chen CY , Buchmeier NA , Libby S , Fang FC , Krause M , Guiney DG . Central regulatory role for the RpoS sigma factor in expression of Salmonella Dublin plasmid virulence genes . J Bacteriol . 1995 ;177 (18 ):5303 –9 . Epub 1995/09/01. 7665519 109 Romling U , Bian Z , Hammar M , Sierralta WD , Normark S . Curli fibers are highly conserved between Salmonella Typhimurium and Escherichia coli with respect to operon structure and regulation . J Bacteriol . 1998 ;180 (3 ):722 –31 . Epub 1998/02/11. 9457880 110 Okada N , Oi Y , Takeda-Shitaka M , Kanou K , Umeyama H , Haneda T , et al Identification of amino acid residues of Salmonella SlyA that are critical for transcriptional regulation . Microbiol . 2007 ;153 (Pt 2 ):548 –60 . Epub 2007/01/30. 111 Stoebel DM , Free A , Dorman CJ . Anti-silencing: overcoming H-NS-mediated repression of transcription in Gram-negative enteric bacteria . Microbiol . 2008 ;154 (Pt 9 ):2533 –45 . Epub 2008/09/02. 112 Osborne SE , Coombes BK . Transcriptional priming of Salmonella Pathogenicity Island-2 precedes cellular invasion . PloS One . 2011 ;6 (6 ):e21648 Epub 2011/07/09. 10.1371/journal.pone.0021648 21738750 113 Bijlsma JJ , Groisman EA . The PhoP/PhoQ system controls the intramacrophage type three secretion system of Salmonella enterica . Mol Microbiol . 2005 ;57 (1 ):85 –96 . Epub 2005/06/14. 15948951 114 Singer HM , Kuhne C , Deditius JA , Hughes KT , Erhardt M . The Salmonella Spi1 virulence regulatory protein HilD directly activates transcription of the flagellar master operon flhDC . J Bacteriol . 2014 ;196 (7 ):1448 –57 . Epub 2014/02/04. 10.1128/JB.01438-13 24488311 115 Lostroh CP , Bajaj V , Lee CA . The cis requirements for transcriptional activation by HilA, a virulence determinant encoded on SPI-1 . Mol Microbiol . 2000 ;37 (2 ):300 –15 . Epub 2000/08/10. 10931326 116 Eichelberg K , Hardt WD , Galan JE . Characterization of SprA, an AraC-like transcriptional regulator encoded within the Salmonella Typhimurium pathogenicity island 1 . Mol Microbiol . 1999 ;33 (1 ):139 –52 . Epub 1999/07/20. 10411731 117 Saini S , Brown JD , Aldridge PD , Rao CV . FliZ Is a posttranslational activator of FlhD4C2-dependent flagellar gene expression . J Bacteriol . 2008 ;190 (14 ):4979 –88 . Epub 2008/05/13. 10.1128/JB.01996-07 18469103 118 Iyoda S , Kamidoi T , Hirose K , Kutsukake K , Watanabe H . A flagellar gene fliZ regulates the expression of invasion genes and virulence phenotype in Salmonella enterica serovar Typhimurium . Microb Pathogenesis . 2001 ;30 (2 ):81 –90 . Epub 2001/02/13. 119 Campoy S , Jara M , Busquets N , de Rozas AM , Badiola I , Barbe J . Intracellular cyclic AMP concentration is decreased in Salmonella typhimurium fur mutants . Microbiol . 2002 ;148 (Pt 4 ):1039 –48 . Epub 2002/04/05. 120 Choi E , Kim H , Lee H , Nam D , Choi J , Shin D . The iron-sensing Fur regulator controls expression timing and levels of Salmonella pathogenicity island 2 genes in the course of environmental acidification . Infect Immun . 2014 ;82 (6 ):2203 –10 . Epub 2014/03/20. 10.1128/IAI.01625-13 24643535 121 Adams P , Fowler R , Kinsella N , Howell G , Farris M , Coote P , et al Proteomic detection of PhoPQ- and acid-mediated repression of Salmonella motility . Proteomics . 2001 ;1 (4 ):597 –607 . Epub 2001/10/30. 11681212

PMC005xxxxxx/PMC5001713.txt

==== Front PLoS GenetPLoS GenetplosplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, CA USA 2756457610.1371/journal.pgen.1006276PGENETICS-D-16-00367Research ArticleBiology and Life SciencesAnatomyDigestive SystemGastrointestinal TractMedicine and Health SciencesAnatomyDigestive SystemGastrointestinal TractBiology and Life SciencesAnatomyReproductive SystemGenital AnatomyGonadsMedicine and Health SciencesAnatomyReproductive SystemGenital AnatomyGonadsResearch and Analysis MethodsModel OrganismsAnimal ModelsCaenorhabditis ElegansBiology and Life SciencesOrganismsAnimalsInvertebratesNematodaCaenorhabditisCaenorhabditis ElegansBiology and life sciencesGeneticsEpigeneticsRNA interferenceBiology and life sciencesGeneticsGene expressionRNA interferenceBiology and life sciencesGeneticsGenetic interferenceRNA interferenceBiology and life sciencesBiochemistryNucleic acidsRNARNA interferencePhysical SciencesChemistryChemical ElementsMagnesiumBiology and Life SciencesGeneticsPhenotypesBiology and Life SciencesCell BiologyCellular TypesAnimal CellsGerm CellsBiology and Life SciencesPhysiologyPhysiological ParametersMedicine and Health SciencesPhysiologyPhysiological ParametersMg2+ Extrusion from Intestinal Epithelia by CNNM Proteins Is Essential for Gonadogenesis via AMPK-TORC1 Signaling in Caenorhabditis elegans CNNM Mg2+ Transporters and GonadogenesisIshii Tasuku 1Funato Yosuke 1*Hashizume Osamu 1Yamazaki Daisuke 1Hirata Yusuke 2Nishiwaki Kiyoji 3Kono Nozomu 45Arai Hiroyuki 46Miki Hiroaki 1*1 Department of Cellular Regulation, Research Institute for Microbial Diseases, Osaka University, Suita, Osaka, Japan2 Laboratory of Health Chemistry, Graduate School of Pharmaceutical Sciences, Tohoku University, Sendai, Miyagi, Japan3 Department of Bioscience, School of Science and Technology, Kwansei Gakuin University, Sanda, Hyogo, Japan4 Department of Health Chemistry, Graduate School of Pharmaceutical Sciences, The University of Tokyo, Bunkyo-ku, Tokyo, Japan5 PRIME, Japan Agency for Medical Research and Development, Chiyoda-ku, Tokyo, Japan6 AMED-CREST, Japan Agency for Medical Research and Development, Chiyoda-ku, Tokyo, JapanLambie Eric EditorLudwig-Maximilians-Universitat Munchen, GERMANYThe authors have declared that no competing interests exist. Conceptualization: YF HM. Formal analysis: TI YF. Funding acquisition: TI YF HM. Investigation: TI YF OH DY YH. Project administration: YF HM. Resources: KN NK HA. Supervision: YF HM. Validation: YF HM. Visualization: TI YF. Writing - original draft: TI YF HM. Writing - review & editing: KN NK HA. * E-mail: yfunato@biken.osaka-u.ac.jp (YF); hmiki@biken.osaka-u.ac.jp (HM)26 8 2016 8 2016 12 8 e100627615 2 2016 4 8 2016 © 2016 Ishii et al2016Ishii et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Mg2+ serves as an essential cofactor for numerous enzymes and its levels are tightly regulated by various Mg2+ transporters. Here, we analyzed Caenorhabditis elegans strains carrying mutations in genes encoding cyclin M (CNNM) Mg2+ transporters. We isolated inactivating mutants for each of the five Caenorhabditis elegans cnnm family genes, cnnm-1 through cnnm-5. cnnm-1; cnnm-3 double mutant worms showed various phenotypes, among which the sterile phenotype was rescued by supplementing the media with Mg2+. This sterility was caused by a gonadogenesis defect with severely attenuated proliferation of germ cells. Using this gonadogenesis defect as an indicator, we performed genome-wide RNAi screening, to search for genes associated with this phenotype. The results revealed that RNAi-mediated inactivation of several genes restores gonad elongation, including aak-2, which encodes the catalytic subunit of AMP-activated protein kinase (AMPK). We then generated triple mutant worms for cnnm-1; cnnm-3; aak-2 and confirmed that the aak-2 mutation also suppressed the defective gonadal elongation in cnnm-1; cnnm-3 mutant worms. AMPK is activated under low-energy conditions and plays a central role in regulating cellular metabolism to adapt to the energy status of cells. Thus, we provide genetic evidence linking Mg2+ homeostasis to energy metabolism via AMPK. Author Summary Mg2+ is the second most abundant cation in cells and serves as an essential cofactor for numerous enzymes. To avoid its shortage, cellular and organismal levels of Mg2+ are tightly regulated by the concerted actions of various Mg2+ transporters and channels. In this study, we analyzed Caenorhabditis elegans strains carrying mutations in genes encoding Mg2+ transporters and found that the mutations abrogated Mg2+ homeostasis. Additionally, these worms were sterile because of a developmental defect in the gonads with severely attenuated proliferation of germ cells. These abnormalities were rescued by additional Mg2+ supplementation to the medium, and thus were considered to be due to Mg2+ shortage. We investigated the mechanism of this Mg2+-associated attenuation of gonadal development, and found that disrupting of the function of AMP-activated protein kinase (AMPK) restored gonad elongation. It is well-known that AMPK is activated under low-energy conditions and plays a central role in regulating cellular metabolism to adapt to the energy status of cells. Thus, we demonstrated that Mg2+ homeostasis is intimately connected to energy metabolism via AMPK. http://dx.doi.org/10.13039/501100001691Japan Society for the Promotion of Science261704Ishii Tasuku http://dx.doi.org/10.13039/501100001691Japan Society for the Promotion of Science26460364Funato Yosuke http://dx.doi.org/10.13039/501100001691Japan Society for the Promotion of Science26291042Miki Hiroaki http://dx.doi.org/10.13039/501100001691Japan Society for the Promotion of Science26111007Miki Hiroaki This study was supported by KAKENHI Grant Numbers 261704 (TI), 26460364 (YF), 26291042 and 26111007 (HM) from Japan Society for the Promotion of Science (https://www.jsps.go.jp). The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files. C. elegans cnnm mutants isolated and used in this study are available on Wormbase (http://www.wormbase.org) under the following ID numbers: dcr1 (cnnm-2): WBVar02146596; dcr2 (cnnm-3): WBVar02146597; dcr3 (cnnm-4): WBVar02146598.Data Availability All relevant data are within the paper and its Supporting Information files. C. elegans cnnm mutants isolated and used in this study are available on Wormbase (http://www.wormbase.org) under the following ID numbers: dcr1 (cnnm-2): WBVar02146596; dcr2 (cnnm-3): WBVar02146597; dcr3 (cnnm-4): WBVar02146598. ==== Body Introduction Mg2+ is the second most abundant cation in cells and serves as an essential cofactor for numerous enzymes. In mammals, magnesium levels are primarily regulated by intestinal absorption and renal reabsorption, where the epithelial cell layer permits selective and regulated Mg2+ transport between apical and basolateral surfaces. There are two known pathways for Mg2+ transport through the epithelial cell layers: the paracellular and the transcellular pathways [1]. The transcellular pathway consists of apical entry and basolateral extrusion mediated by Mg2+-permeable cation channels and transporters. TRPM6, a member of the transient receptor potential channel (TRP) family, is a key molecule in the transcellular pathway [2]. TRPM6 localizes at the apical membrane of intestinal epithelial cells and distal convoluted tubule (DCT) cells in the kidney [3] and mediates Mg2+ absorption and reabsorption, respectively. Indeed, mutations in TRPM6 result in recessive familial hypomagnesemia with secondary hypocalcemia [4,5]. In addition, the related channel TRPM7 was also found to play an important role in magnesium homeostasis in mice [6]. These observations implicate TRPM6/7 in the apical entry of Mg2+ into epithelial cells. Another key molecule in the transcellular pathway is the ancient conserved domain protein/cyclin M (CNNM) family. In mammals, the CNNM family consists of 4 integral membrane proteins (CNNM1–4) that possess an evolutionarily conserved domain from bacteria [7]. Recent genomic analyses have revealed that several single-nucleotide polymorphisms in CNNMs are linked to serum magnesium levels [8] and that mutations in CNNM2 are responsible for familial hypomagnesemia [9]. It was reported that CNNM4 extrudes Mg2+ from the basolateral membrane of intestinal epithelial cells and is involved in intestinal Mg2+ absorption [10]. Another family member, CNNM2, is strongly expressed at the basolateral membrane of DCT cells [9,11] and can extrude Mg2+ similarly to CNNM4 [12], suggesting that CNNM2 plays a similar role in basolateral Mg2+ extrusion in kidney DCT cells. Two groups recently reported that CNNMs associate with phosphatase of regenerating liver (PRL), a cancer-associated tyrosine phosphatase [13,14]. One group found that PRL binds to CNNMs and inhibits the Mg2+-transporting function [13], whereas the other group reported that PRL stimulates this activity [14]. Therefore, how PRL affects the function of CNNMs remains unclear. In addition, it is unknown whether other molecules are involved in the regulation of CNNM function. To address these problems, comprehensive screening can be used to identify genes that functionally associate with CNNMs. Caenorhabditis elegans (C. elegans), which is a model organism commonly used for genetic analyses, also absorbs Mg2+ through a similar transcellular mechanism in the intestine. The apical entry step of the transcellular pathway is mediated by two TRPM family channels: GTL-1 and GON-2 [15]. C. elegans has an excretory canal that removes wastes from the body, wherein another C. elegans TRPM channel, GTL-2, plays an important role in magnesium homeostasis [16]. Thus, C. elegans possesses a system for regulating magnesium homeostasis that is similar to that in mammals. Taken together with the genetic tractability of C. elegans, this organism may serve as an ideal experimental model for investigating the regulatory mechanism and functional importance of magnesium homeostasis. In this study, we performed functional analyses of the C. elegans CNNM family and found that cnnm-1; cnnm-3 double mutant worms displayed pleiotropic phenotypes. The sterile phenotype (due to defective gonadogenesis) was rescued by Mg2+ supplementation: oocyte development was restored and mutant worms became fertile. Detailed analyses of the gonadal phenotype revealed that the inactivating mutation of aak-2, which encodes the α subunit of AMP-activated protein kinase (AMPK), significantly rescued the gonadogenesis defect in cnnm-1; cnnm-3 mutants, thereby indicating a genetic interaction between CNNM and AMPK. Results cnnm family genes of C. elegans A homology search using BLAST with amino acid sequences of the human CNNM4 protein revealed that the C. elegans genome contains genes encoding 5 previously uncharacterized CNNM family proteins, which possess the functionally essential domains DUF21 and CBS [10,12] (S1 Fig). Each C. elegans CNNM protein showed significant identity with all human CNNM family members (24–47%). To determine the evolutionary relationship between C. elegans CNNM proteins and other CNNM homologs, we constructed a phylogenetic tree (Fig 1A). The vertebrates have 4 paralogs (CNNM1–4), and each of them is orthologous between different vertebrate species (human, mouse, frog, and zebrafish). In contrast, C. elegans CNNM-1−5 emerged independently of the vertebrate CNNMs. To investigate the in vivo functions of C. elegans CNNM family proteins, we obtained and generated mutant alleles for all cnnm family members (Fig 1B). These mutations abolish the function of each CNNM protein because the mutant proteins lack functionally essential part of either the DUF21 or CBS domains (S1 Fig, see its legend for details). 10.1371/journal.pgen.1006276.g001Fig 1 The C. elegans cnnm family. (A) Phylogenetic tree of CNNM family. Amino acid sequences of C. elegans CNNMs, along with human, mouse, frog, zebrafish, and fruit fly CNNM orthologues, were aligned using Clustal W (version 2.1, http://clustalw.ddbj.nig.ac.jp/), and the phylogenetic tree was generated with neighbor-joining method by using Clustal W2—phylogeny (http://www.ebi.ac.uk/Tools/phylogeny/clustalw2_phylogeny/). Vertebrate CNNM orthologues are grouped, and C. elegans CNNMs are also grouped. Bar, 0.1 substitutions per amino acid. (B) Schematic illustrations of the cnnm family genes. Exons and introns are indicated by boxes and lines, respectively. The regions encoding the evolutionarily conserved DUF21 and CBS domains are indicated with black and gray boxes, respectively. Sequence names are shown in parentheses. The illustrations were generated using the Exon-Intron Graphic Maker by Nikhil Bhatla (http://www.wormweb.org/exonintron). Lines, deletion; arrowhead, point mutation; arrow, Mos1 insertion. Bar, 0.5 kb. Pleiotropic phenotypes of cnnm-1; cnnm-3 mutant worms Observation of these mutant worms revealed no obvious abnormalities except for in the cnnm-3 mutant worms, a few (5.3%) of which were sterile (Fig 2A). We speculated that functional redundancy among the cnnm family members may have masked the mutant phenotype. Thus, we generated double mutants for all possible combinations by crossing all single mutants. We found that cnnm-1; cnnm-3 and cnnm-2; cnnm-3 mutant worms were severely (100%) and moderately (22%) sterile, respectively (Fig 2A). Because of the completely sterile phenotype, we analyzed cnnm-1; cnnm-3 mutants in subsequent experiments. 10.1371/journal.pgen.1006276.g002Fig 2 Pleiotropic phenotypes of cnnm-1; cnnm-3 mutant worms. (A) Quantification of the sterile phenotype for each genotype. Eggs collected by synchronous laying were grown for 4 days and then examined for the presence of embryos in their uteri by microscopic observation. Worms lacking embryos were scored as sterile. More than 100 worms were analyzed for each genotype. cnnm(0): cnnm-1; cnnm-2; cnnm-3; cnnm-4; cnnm-5 mutant. (B) Nomarski images of wild-type and cnnm-1; cnnm-3 mutant worms grown for 3 days. High-magnification views of the boxed areas are also shown (right). Arrow indicates the vulva. Bar, 30 μm. (C) Mixed stage worms from L2 to adult (L4 to adult worms were most abundant) were examined for alae formation and subjected to body size measurement (n = 30 in each experiment). The body size at the transition from L4 to the adult molt was estimated by calculating the mean body size of the smallest three worms with alae and the largest three worms without alae. The data are shown as the means of three experiments. Error bars indicate SEM. p values were determined by ANOVA, followed by two-tailed multiple Student’s t-test with Tukey’s correction. *p < 0.05 versus wild-type. (D) Lifespan of worms with the indicated genotype. For each genotype, 45 synchronized L4/young adult worms were transferred to fresh plates (15 worms per plate) and were then scored daily for survival. The graph represents data combined from at least two experiments. Mean lifespan (± SEM) of worms is also indicated in parentheses. p values were determined by log rank (Mantel-Cox) test, and the Bonferroni method was then used to correct for multiple comparisons. ****p < 0.0001. The cnnm-1; cnnm-3 mutant worms were significantly smaller than the wild-type N2 worms grown for the same amount of time (Fig 2B). Therefore, we compared the body sizes of stage-matched worms. cnnm-1; cnnm-3 mutant worms did not form the vulva (Fig 2B), the eversion of which determines the adult stage [17]. Therefore, we focused on the presence of alae, the longitudinal ridges present in adult worms but not in earlier L2−4 larvae as the marker to confirm whether the worms had reached the adult stage [18,19]. We examined alae formation and body sizes of mixed stage worms from L2 to adult, and then estimated the body size at the transition from L4 to adult molt, which was determined as the mean value of body size of the three smallest worms with alae and three largest worms without alae. The results showed that the body size of cnnm-1; cnnm-3 mutant worms was smaller than that of wild-type worms (Fig 2C). At 64 h when all wild-type worms had just reached the adult stage, approximately half of the cnnm-1; cnnm-3 mutant worms were alae-positive, indicating developmental delay. These phenotypes of cnnm-1; cnnm-3 mutant worms were rescued by the introduction of either cnnm-1 or cnnm-3 genomic DNA (Fig 2A and 2C), confirming that these abnormalities were caused by mutations in cnnm-1 and cnnm-3. We also found that the color of the intestine in cnnm-1; cnnm-3 mutant worms was dark (Fig 2B). This was also observed in mutant worms of daf-2, which encodes the insulin-like receptor [20]. Because daf-2 mutations are well-known to extend lifespan, we next examined the lifespan of cnnm-1; cnnm-3 mutant worms (Fig 2D). Consistent with previous studies, daf-2 mutant worms showed much longer lifespans than wild-type worms. In contrast, cnnm-1; cnnm-3 mutant worms had shorter lifespans than wild-type worms, which were similar to those of daf-16 mutant worms; these worms were used here as an example of short-lived mutants [20]. Introduction of either cnnm-1 or cnnm-3 genomic DNA to cnnm-1; cnnm-3 mutant worms not only rescued the short-life phenotype, but also showed moderately longer lifespans than wild-type worms, particularly in the case of worms introduced with cnnm-1 (Fig 2D). Overexpression of either cnnm-1 or cnnm-3 in wild-type worms also prolonged the lifespan (S2 Fig), indicating the important role of these genes in determining lifespan. We also analyzed whether the short-life phenotype of cnnm-1; cnnm-3 mutants could be exacerbated by further mutations in other CNNM family genes. For this, we generated cnnm-1; cnnm-2; cnnm-3; cnnm-4; cnnm-5 quintuple mutant worms (hereafter, cnnm(0)), which possessed mutations in all five C. elegans cnnm family genes, by mating. As expected, cnnm(0) worms showed complete sterility (Fig 2A) with a lifespan similar to that of cnnm-1; cnnm-3 mutants (Fig 2D). Effects of Mg2+ supplementation on cnnm-1; cnnm-3 mutant worms Because mammalian CNNM family proteins are involved in Mg2+ transport [10,12–14], the phenotypes of cnnm-1; cnnm-3 mutant worms may be related to abnormalities in magnesium homeostasis. Therefore, we tested the effects of Mg2+ supplementation in the media, and found that 76.6% and 100% of cnnm-1; cnnm-3 mutant worms became fertile following supplementation with 1 mM and 3 mM of Mg2+, respectively (Fig 3A). In contrast, the small body size and short lifespan were not affected by Mg2+ supplementation (Fig 3B and 3C). Supplementation of culture plates with Ca2+ did not affect any of the phenotypes. Collectively, these results suggest that the sterile phenotype of cnnm-1; cnnm-3 mutant worms is related to altered magnesium homeostasis. As the cnnm-1; cnnm-3 mutant worms did not produce oocytes or form a vulva (Fig 2B), which are formed during gonadal development, we next examined gonadal development in cnnm-1; cnnm-3 mutant worms. 10.1371/journal.pgen.1006276.g003Fig 3 Effects of Mg2+ supplementation on cnnm-1; cnnm-3 mutant worms. (A) cnnm-1; cnnm-3 mutant worms were grown on plates supplemented with the indicated concentrations of Mg2+ or Ca2+ for 4 days and then examined for sterility. More than 50 worms were analyzed for each experimental condition. (B) Wild-type and cnnm-1; cnnm-3 mutant worms were grown on plates supplemented with 1 mM Mg2+ or Ca2+ for at least two generations. Mixed stage worms from L2 to adult (L4 to adult worms comprise the most) were examined for alae formation and subjected to body size measurement (n = 30 per each experiment). Body size at the transition from L4 to adult molt was estimated as in Fig 2C. The data are shown as the means of three experiments. Error bars indicate SEM. p values were determined by ANOVA, followed by two-tailed multiple Student’s t-tests with Tukey’s correction. *p < 0.05. (C) 45 synchronized L4/young adult worms of the indicated genotype were transferred to fresh plates supplemented with 1 mM Mg2+ or Ca2+ (15 worms per plate), and then scored for survival daily. The graph represents data combined from at least two experiments. Mean lifespan (± SEM) of worms is also indicated in parentheses. p values were determined by log rank (Mantel-Cox) test, and the Bonferroni method was then used to correct for multiple comparisons. ****p < 0.0001. cnnm-1 and cnnm-3 are required for postembryonic gonadal development At hatching, the primordial gonad in C. elegans is composed of four cells: Z1–Z4 [21]. During larval development, Z1 and Z4 cells give rise to the somatic gonad including the distal tip cells (DTCs), uterus, sheath cells, and spermathecae, whereas the Z2 and Z3 cells give rise to the germ line. We examined the extent of gonadal development by expressing GFP under control of the lag-2 promoter, which drives gene expression in Z1/Z4 cells and DTCs [22,23] located at the distal end of two gonadal arms and lead the elongation of the arms to form the U-shaped hermaphrodite gonad. When the mutant worms hatched from eggs, two GFP-positive cells were observed at the appropriate positions, suggesting that primordial gonad development proceeded normally in mutants. However, the primordial gonad of mutants did not elongate, even in the L4/young adult stages (Fig 4A). Immunofluorescence analysis using an anti-PGL-1 antibody, which stains P-granules in germ cells, demonstrated that most of the cnnm-1; cnnm-3 mutant L4/young adult worms had only two germ cells corresponding to the Z2/Z3 cells (Fig 4B). Considering that Mg2+ supplementation restored the fertility of cnnm-1; cnnm-3 mutant worms, these results suggest that CNNM-1 and CNNM-3 promote postembryonic gonadal development by regulating Mg2+ levels. Previous studies demonstrated that mutants of gon-2, which encodes a TRPM channel protein, showed a severe gonadogenesis defect, which was partially restored by Mg2+ supplementation [15,24]. 10.1371/journal.pgen.1006276.g004Fig 4 cnnm-1 and cnnm-3 are required for postembryonic development of the gonad. (A) L4/young adult worms of the indicated genotype, carrying lag-2p::GFP, were examined for gonadal arm extension. Representative images and the high-magnification view of the boxed area are shown (left). In each image, the anterior side of the worm is positioned to the left. Dotted lines indicate the outlines of the gonad. Bar, 20 μm. Worms with gonads of ≥ 60 μm in length of the long axis (3-fold as long as the primordial gonad) were determined as extended and the results (%) are shown on the graph (right). More than 100 worms were analyzed. (B) The worms were stained with DAPI (blue) and anti-PGL-1 antibody (red), and then the number of germ cells (PGL-1-positive cells) per worm was counted. Representative images and the high-magnification view of the boxed area are shown (top). Bar, 15 μm. The graph indicates the distribution of worms (%) with the indicated numbers of germ cells (bottom). In addition to somatic gonad and germ cells, several types of cells, such as the ray cells of the male tail, are known to proliferate during postembryonic stages [25]. Our findings revealed no apparent abnormalities in the tail ray structures of cnnm-1; cnnm-3 mutant males (S3 Fig), suggesting that not all of the postembryonic cell divisions were affected in cnnm-1; cnnm-3 mutant worms. Localization of CNNM-1 and CNNM-3 at the basolateral membrane of intestinal cells To characterize the roles of CNNM-1 and CNNM-3, we first examined their expression patterns by generating transgenic worms expressing GFP under control of the cnnm-1 or cnnm-3 promoters. Unique GFP expression was observed in various tissues, such as the pharynx, hypodermis, rectum, and muscles, but strong expression was commonly observed in the intestine and neurons (Fig 5A). Given this expression pattern, we forced the expression of CNNM-1 or CNNM-3 in the intestine or neurons of cnnm-1; cnnm-3 mutant worms using the intestine-specific ges-1 promoter [26] or neuron-specific aex-3 promoter [27], respectively. The intestinal expression of either CNNM-1 or CNNM-3 nearly completely rescued the sterile phenotype of cnnm-1; cnnm-3 mutants, whereas their expression in neurons showed only a subtle effect (Fig 5B), suggesting that their expression in the intestine is important for gonadal development. The intestinal cells of C. elegans are attached to each other at the borders of the apical membrane via cell-cell junctions known as apical junctions, which have mixed traits of both the adherens junction and the tight junction in mammalian epithelial cells, and thus exhibit apico-basal polarity [28]. We subsequently examined the subcellular localization of CNNM-1 and CNNM-3 in intestinal cells using transgenic worms expressing their respective Venus-fusion proteins, which also rescued the sterile phenotype (Fig 2A). Excluding some large clumps in the cytoplasm, which are often observed following ectopic expression of Venus-fusion proteins, the fluorescent signal of Venus-fusion CNNM1 and CNNM3 was predominantly observed in the basolateral membrane of intestinal cells (Fig 5C). Therefore, both CNNM-1 and CNNM-3 are considered to extrude Mg2+ from intestinal cells to the pseudocoelom. Based on this hypothesis, cnnm-1; cnnm-3 mutant worms were expected to have higher levels of Mg2+ in intestinal cells and lower levels of Mg2+ in the pseudocoelom, which contains other tissues such as the gonad, possibly explaining why Mg2+ supplementation restored gonadal development (Fig 3A). 10.1371/journal.pgen.1006276.g005Fig 5 Localization of CNNM-1 and CNNM-3 at the basolateral membrane of intestinal cells. (A) Fluorescent images of wild-type L1 larvae with cnnm-1p::GFP or cnnm-3p::GFP. Areas of neurons and intestines showing strong GFP signals are marked. Bar, 30 μm. (B) Quantification of sterile phenotype in wild-type, cnnm-1; cnnm-3 mutants, and cnnm-1; cnnm-3 mutants expressing either CNNM-1 or CNNM-3 under the control of the promoters for intestinal (ges-1p) or pan-neuronal (aex-3p) markers. More than 50 worms were analyzed for each genotype. (C) Fluorescent images of intestinal cells from wild-type L1 larvae expressing the Venus-fusion proteins of CNNM-1 or CNNM-3 (left). Schematic representation of the intestinal cell plasma membranes are also shown (right). Apical and basolateral membranes are indicated by dotted and solid lines, respectively. Bar, 10 μm. Assessment of magnesium levels by ICP-MS To assess the predicted regulatory model, we quantified the magnesium levels in wild-type worms and in cnnm-1; cnnm-3 mutant worms using inductively coupled plasma mass spectrometry (ICP-MS). As shown in Table 1, we found that cnnm-1; cnnm-3 mutant worms had higher magnesium levels (143% of wild-type worms). We then physically dissected the intestines from the worms using a scalpel and subjected the intestines to magnesium quantitation. The results indicated much higher levels of magnesium in cnnm-1; cnnm-3 mutant worms (195% of wild-type worms). It should be noted that there was no overlap in magnesium levels in wild-type and mutant worms. It was technically difficult to physically dissect other remaining tissues, which are much smaller than the intestine. Therefore, we estimated magnesium levels in other tissues by calculating the volumes of the total body and intestine, and found that the magnesium level was reduced in other tissues of cnnm-1; cnnm-3 mutant worms (67% of wild-type levels). Taken together with the result that the infertile phenotype was rescued by artificially increasing Mg2+ levels (Fig 3), the abnormalities in cnnm-1; cnnm-3 mutant worms may have been caused by Mg2+ deficiency. 10.1371/journal.pgen.1006276.t001Table 1 Magnesium levels in wild-type and cnnm-1; cnnm-3 mutant worms. wild-type cnnm-1; cnnm-3 p-value Total (ppb/μg) 3.9 ± 0.2 5.6 ± 0.4 0.014 Intestine (ppb/μg) 4.0 ± 0.4 7.8 ± 0.8 0.015 Other tissues (ppb/μg) (estimated value) 3.9 2.6 ND The total and intestinal levels of magnesium were directly determined by ICP-MS using L4/young adult worms. The data are shown as the means (± SEM) of three experiments. Magnesium levels in the other remaining tissues were estimated by calculating the total body volume (wild-type: 1,170 ± 61 nl, n = 10; cnnm-1; cnnm-3: 638 ± 38 nl, n = 10) and intestine (wild-type: 458 ± 22 nl, n = 10; cnnm-1; cnnm-3: 357 ± 13 nl, n = 10). p values were determined by Student’s two tailed t-test. RNAi screening for genes that functionally associated with cnnm-1 and cnnm-3 Next, we searched for genes that are functionally associated with cnnm-1 and cnnm-3. For this purpose, we performed RNAi-based screening, because (i) an RNAi feeding library targeting 86% of the open reading frames of C. elegans is available and widely used in genome-wide screening [29] and (ii) unlike conventional forward genetics approaches, it is not necessary to collect the worms and their progenies after screening, making this method applicable to sterile cnnm-1; cnnm-3 strains. To increase RNAi efficacy, screening was performed using worms carrying the rrf-3 mutation, which renders the worms hypersensitive to RNAi treatment [30]. Two rounds of screening identified 31 genes, of which RNAi treatment reproducibly resulted in elongation of the gonadal arm in more than 50% of cnnm-1; cnnm-3; rrf-3 mutant worms (Table 2 and S4 Fig). These 31 genes are involved in a variety of biological processes, including protein transport, metabolism, mitochondrial function, signal transduction, gene expression, ion transport, immune response, and the cell cycle. Among these, we performed detailed analyses of aak-2, which encodes the α-subunit of AMPK. AMPK is the key energy sensor in most eukaryotic cells and is activated under low-energy conditions such as decreased ATP levels [31]. Most intracellular ATP is known to form complexes with Mg2+, which is required for numerous enzymatic reactions involving ATP [32,33]. We hypothesized that dysregulation of cellular Mg2+ levels in cnnm-1; cnnm-3 mutant worms could affect AMPK activity. 10.1371/journal.pgen.1006276.t002Table 2 Genes of which RNAi suppressed the gonadogenesis defect in cnnm-1; cnnm-3 mutant worms. (%) gonadal arm extended Sequence name Gene name Description 1st round 2nd round C10E2.6 mct-6 Monocarboxylate transporter 100 100 R160.1 dpy-23 AP-2 complex subunit mu2 100 100 C06B8.7 Scavenger receptor cysteine-rich domain 100 100 F55D10.3 glit-1 Thyroglobulin 80 93 T05H4.5 hpo-19 NADH-cytochrome b5 reductase 90 90 C55B7.8 dbr-1 RNA-lariat debranching enzyme 90 87 F02E8.3 aps-2 AP-2 complex subunit sigma2 90 87 T01C8.1 aak-2 AMPK alpha2 80 87 C34C6.6 prx-5 Peroxisomal targeting signal 1 receptor 80 87 R03E1.1 sym-4 WD repeat-containing protein 80 83 F29A7.6 M-phase phosphoprotein 6 100 83 T14G10.7 hpo-5 GPI transamidase component PIG-S 70 83 C07A9.11 ncx-7 Sodium/potassium/calcium exchanger 90 80 R12C12.2 ran-5 RanBP1 domain 80 77 F11E6.5 elo-2 palmitic acid elongase 50 73 F36H12.5 Transcription initiation factor TFIID subunit 3 80 73 F44A6.2 sex-1 Nuclear hormone receptor 60 73 F58B3.4 Nucleolar pre-rRNA processing protein 80 70 C27F2.8 Transmembrane protein 131 80 67 T26A8.4 Zinc finger CCCH domain-containing protein 4 80 67 K04E7.2 pept-1 Oligopeptide transporter 80 67 Y65B4BR.4 wwp-1 NEDD4-like E3 ubiquitin-protein ligase WWP1 50 63 C09H6.3 mau-2 MAU2 chromatid cohesion factor homolog 70 60 C27B7.8 rap-1 Ras-related protein Rap-1b 50 60 F23C8.6 did-2 Charged multivesicular body protein 1b 50 57 F08B12.2 prx-12 Peroxisomal biogenesis factor 12 60 57 ZK418.4 lin-37 LIN37 family protein 70 53 R07H5.8 Adenosine kinase 60 53 Y71H10B.1 Cytosolic purine 5'-nucleotidase 50 53 Y40B1B.7 Coiled-coil domain-containing protein 86 50 53 F46F11.4 ubl-5 Ubiquitin-like protein 5 50 50 Genome wide RNAi screening identified 31 genes that functionally associate with cnnm-1 and cnnm-3. The information is derived from either Wormbase or InterProScan. AMPK mediates the gonadogenesis defect caused by cnnm-1; cnnm-3 mutation AMPK is a heterotrimeric kinase consisting of a catalytic subunit (α) and two regulatory subunits (β and γ). In C. elegans, there are two α subunits, AAK-1 and AAK-2, which are encoded by different genes [34]. Therefore, we examined whether the predicted null mutations for aak-1 and/or aak-2 could suppress the gonadogenesis defect in cnnm-1; cnnm-3 mutant worms (Fig 6A). The additional mutation in aak-2 or in both aak-1 and aak-2 nearly completely restored the gonadal arm extension, while the aak-1 mutation showed only a marginal effect. We also analyzed germ cell proliferation in cnnm-1; cnnm-3; aak-1; aak-2 quadruple mutant worms and found that most contained many germ cells (> 100 cells, Fig 6B). Moreover, when we examined the fertility of quadruple mutant worms by supplementation with various concentrations of Mg2+, fertility was restored at lower concentrations of Mg2+ (Fig 6C, left). We also performed rescue experiments by changing the initiation timing of Mg2+ supplementation. The results showed that fertility was restored even when supplementation was started at later stages (Fig 6C, right). Therefore, the additional mutations in aak-1 and aak-2 significantly augmented the effect of Mg2+ on fertility, implicating that AMPK mediates the effect of Mg2+ on gonadogenesis. 10.1371/journal.pgen.1006276.g006Fig 6 AMPK mediates the gonadogenesis defect of cnnm-1; cnnm-3 mutant worms. (A) L4/young adult worms of the indicated genotype, carrying lag-2p::GFP, were examined for gonadal arm extension. Representative images are shown (left). In each image, the anterior side of the worm is positioned to the left. Dotted lines indicate the outlines of the gonad. Bar, 20 μm. Worms with gonads of ≥ 60 μm in length of the long axis (3-fold as long as the primordial gonad) were determined as extended and the results (%) are shown in the graph (right). More than 100 worms were analyzed. (B) cnnm-1; cnnm-3; aak-1; aak-2 mutant worms were stained with DAPI (blue) and anti-PGL-1 antibody (red), and then the number of germ cells (PGL-1-positive cells) per worm was counted. Representative images are shown (left). Bar, 15 μm. The graph indicates the distribution of worms (%) with the indicated numbers of germ cells (right). (C) Worms of the indicated genotype were grown on plates supplemented with the indicated concentrations of Mg2+ from the F0 generation (left) or grown on plates supplemented with 1 mM Mg2+ starting from the indicated stages (right). More than 50 worms were analyzed for each experimental condition. To identify the cell type that is primarily affected by the cnnm-1; cnnm-3 mutations and responsible for the gonadogenesis defect, we performed tissue-specific RNAi experiments. It has been reported that somatic gonad-specific RNAi can be achieved using a strain that carries both a mutation in the rde-1 gene, which encodes an Argonaute protein required for siRNA maturation [35], and a transgene qIs140[lag-2p::rde-1] that drives the expression of wild-type rde-1 under the control of the lag-2 promoter [36]. We generated cnnm-1; cnnm-3; rde-1; qIs140 strains and performed feeding RNAi experiments of aak-2. As shown in Fig 7, somatic gonad-specific RNAi of aak-2 restored the gonadal arm extension in most (75.9% extended) cnnm-1; cnnm-3; rde-1; qIs140 worms, while worms without the transgene (cnnm-1; cnnm-3; rde-1) failed to show restored extension. These results clearly show the importance of aak-2 function in the somatic gonad. 10.1371/journal.pgen.1006276.g007Fig 7 Primary defect occurs in somatic gonad. L4/young adult worms of the indicated genotype, carrying lag-2p::GFP, were fed with bacteria carrying the dsRNA corresponding to aak-2. In each image, the anterior side of the worm is positioned to the left. Dotted lines indicate the outlines of the gonad. Bar, 20 μm. Worms with gonads of ≥ 60 μm in length of the long axis (3-fold as long as the primordial gonad) were determined as extended and the results (%) are shown in the graph (right). More than 50 worms were analyzed for each condition. AMPK can regulate multiple downstream molecules, including the target of rapamycin complex (TORC)1, which is well-known to be important in cell growth [37]. Therefore, we performed RNAi experiments to investigate the importance of TORC1 in gonadogenesis. Because RNAi of either let-363 or daf-15 is known to cause L3 larval arrest [38], we observed worms in the L2–L3 stages. We found that inhibition of daf-15 (encoding Raptor ortholog) or let-363 (encoding ortholog of mammalian TOR kinase) efficiently suppressed gonadal arm extension in cnnm-1; cnnm-3; aak-1; aak-2 quadruple mutant worms (S5 Fig). In contrast, inhibition of rheb-1 (encoding Rheb ortholog) showed minimal suppression of gonadal arm extension. Biochemical analyses using mammalian cultured cells showed that AMPK inhibited TOR kinase via phosphorylation of TSC (the upstream regulator of Rheb [39]) and Raptor [40], but there is no TSC homolog in C. elegans [41]. Taken together, the results of our RNAi experiments suggest the involvement of dysregulated AMPK-TORC1 signaling in the gonadogenesis defect of cnnm-1; cnnm-3 mutants. Discussion In this study, we showed that cnnm-1; cnnm-3 mutant worms displayed pleiotropic phenotypes, such as infertility due to a gonadogenesis defect, shortened lifespan, and small body size (Fig 2). Among these, the gonadogenesis defect was completely restored by adding Mg2+ to the culture media (Fig 3). These results suggest that abnormal Mg2+ regulation in cnnm-1; cnnm-3 mutant worms affected gonadal development. In contrast, Mg2+ supplementation affected neither the lifespan nor the body size of cnnm-1; cnnm-3 mutants (Fig 3). Whether the lifespan and body size phenotypes are related to altered magnesium homeostasis remains unknown, and further analyses are required to clarify the relationship with CNNM functions. Our elemental analyses suggested that the cnnm-1; cnnm-3 mutant worms had higher levels of Mg2+ in intestinal cells and lower levels of Mg2+ in the pseudocoelom, which contains other tissues such as the gonad (Table 1). This is consistent with the presumed molecular function of CNNM proteins at the basolateral membrane of the intestinal epithelia. In addition, we found that cnnm-1; cnnm-3 mutants exhibited a severe gonadogenesis defect (Fig 4), which was completely restored by additional Mg2+ supplementation to the media. Collectively, these results strongly suggest that the gonadogenesis defect is due to Mg2+ deficiency in the gonad. Because of experimental limitations, we could only estimate magnesium levels in non-intestinal tissues, which showed moderate reduction in cnnm-1; cnnm-3 mutant worms (~67% of wild-type worms). Whether this level of reduction alone can explain the pleiotropic phenotypes of mutant worms is unclear. However, a decrease of total magnesium levels by ~20% can cause proliferation arrest of HEK 293 human cultured cells [10]. More detailed studies of the Mg2+ distribution are required to precisely characterize the mechanism of how Mg2+ deficiency affects worm development. The importance of Mg2+ in the regulation of various cell functions, such as proliferation, was predicted previously [42,43]. A study of chicken DT40 cells lacking TRPM7, a Mg2+-permeable cation channel, revealed the significance of Mg2+ influx in maintaining cell proliferation [44]. However, the mechanism of Mg2+ action is poorly understood. Our previous study showed that intracellular Mg2+ levels significantly affect ATP levels in cultured mammalian cells [13]. Because AMPK is an energy sensor kinase that is directly regulated by cellular ATP levels [39], AMPK was thought to play an important role in mediating the effect of Mg2+. In this study, we performed a genome-wide RNAi screen to identify genes involved in Mg2+-associated regulation of cell proliferation, which yielded 31 candidate genes including aak-2, which encodes a catalytic subunit of AMPK (Table 2). We confirmed the importance of AMPK by showing that additional mutations in aak-1 and aak-2 restored gonadal arm extension in cnnm-1; cnnm-3 mutants (Fig 6). Moreover, in cnnm-1; cnnm-3; aak-1; aak-2 quadruple mutant worms, fertility was restored at lower levels of Mg2+ supplementation (Fig 6). Therefore, we provide genetic evidence linking Mg2+ homeostasis to the AMPK function. In addition, tissue-specific RNAi experiments clearly located the rescue function of AMPK in somatic gonad (Fig 7). Therefore, Mg2+ decrease presumably causes gonadogenesis defect by affecting the AMPK function in somatic gonad, which then affects proliferation of germ cells. One of the important downstream targets of AMPK is TORC1: activated AMPK suppresses the function of TORC1 [37]. Our RNAi experiments implicated daf-15 (encoding Raptor ortholog) and let-363 (encoding ortholog of mammalian TOR kinase) in the restored elongation of gonads by aak-1/aak-2 mutation (S5 Fig), suggesting the importance of AMPK-TORC1 signaling in mediating gonadogenesis. It should be noted that Mg2+ influx through TRPM7 in chicken culture cells is essential for sustained activation of TORC1 and cell proliferation [44]. This Mg2+-dependent TORC1 signaling is presumed to be mediated by phosphoinositide 3-kinase and Akt. Whether it is functionally related to the Mg2+-dependent AMPK-TORC1 signaling in C. elegans remains unknown, but it is intriguing that Mg2+ perturbation convergently affects the same target molecule in distant species. Gonadal development in C. elegans is severely affected by the nutrition status, and starvation forces primordial gonad to stop cell proliferation due to the checkpoint activation. Reportedly, the starvation-induced proliferation arrest of germ cells can be partly rescued by aak-1/2 mutation [45], as in the case of cnnm-1; cnnm-3 mutant worms. Therefore, the arrest of germ cell proliferation in cnnm-1; cnnm-3 mutants are presumed to occur by similar AMPK-dependent mechanism. In more detail, gonad cells are arrested at different stages of the cell cycle by starvation, depending on the cell types: Z1/Z4 cells (somatic gonad cells) are arrested at the G1 phase, while Z2/Z3 cells (germ cells) are arrested at the G2 phase [46–48]. It is not determined at which stage cells are arrested in cnnm-1; cnnm-3 mutants, and thus, it is impossible to further evaluate the similarities between the starved worms and cnnm-1; cnnm-3 mutant worms. However, it is reported that inactivation of GON-2 Mg2+ channel causes G1 arrest in Z1/Z4 cells by upregulating the G1/S checkpoint molecule CKI-1 [46]. Also, in chicken cell culture experiments, TRPM7-deficiency caused downregulation of TORC1 signaling and G1 arrest [44]. Taken together, it is plausible that Mg2+ shortage in the pseudocoelom of cnnm-1; cnnm-3 mutant worms causes G1 arrest in Z1/Z4 cells by the AMPK-TORC1-dependent checkpoint control mechanism, and subsequently causes G2 arrest in Z2/Z3 cells (Fig 8). 10.1371/journal.pgen.1006276.g008Fig 8 Proposed model for the role of C. elegans CNNM proteins. (Left) Mg2+ in the intestinal lumen enters into the intestinal cells through GON-2 and GTL-1, the apically-localized Mg2+-permeable channels. Mg2+ is then extruded from the intestinal cells to the pseudocoelom by CNNM-1 and CNNM-3, the Mg2+ transporters localized at the basolateral membrane. Pseudocoelom is in contact with many tissues/cells including primordial gonad cells (Z1−Z4). (Right) Through GON-2, somatic gonad cells (Z1/Z4) take up Mg2+, which suppresses the AMPK function and in turn augments the TORC1 function, ultimately stimulating the proliferation of adjacent germ cells (Z2/Z3). Finally, besides aak-2, our RNAi screen yielded several other candidate genes, which are involved in various biochemical/biological processes. Further characterizations of the relationship between Mg2+ and these genes may reveal new roles for Mg2+ in the regulation of biological functions. Materials and Methods C. elegans strains and general methods for handling worms All C. elegans strains used in this study were derived from wild-type C. elegans var. Bristol (N2). Unless otherwise indicated, the worms were grown at 20°C on NGM-lite plates seeded with OP-50 E. coli, as described previously [24]. The following mutations, rearrangement, and transgenes were used in this study: LGI cnnm-4(dcr3) and daf-16(mgDf50); LGII rrf-3(pk1426); LGIII daf-2(e1370), cnnm-5(ttTi19567), and aak-1(tm1944); LGIV cnnm-1(gk222902) and nT1[qIs51] (IV; V); LGV rde-1(ne219); LGX cnnm-3(dcr2), cnnm-2(dcr1), and aak-2(ok524); qIs56[lag-2p::GFP]; qIs140[lag-2p::rde-1]; sEx rCesC33D12.2::GFP 14584[rCesC33D12.2rCesC33D12.2::GFP:: GFP]. Strains carrying each mutation and rearrangement were either isolated from the Trimethylpsoralen/UV-mutagenized library (see the “isolation of cnnm mutant strains” section for details), provided by the Caenorhabditis Genetics Center which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440), or obtained from NEMAGENETAG Project funded by the European Community [49]. All isolated/obtained strains were backcrossed at least 4 times with N2 before use. Double homozygous worms for cnnm-1(gk222902) and cnnm-3(dcr2) are completely sterile, and therefore, cnnm-1(gk222902)/+; cnnm-3(dcr2) worms were maintained using the nT1[qIs51] (IV; V) balancer. Synchronization of worm development was achieved by egg laying of gravid adults for 6 h unless otherwise indicated. Generation of plasmids and transgenic C. elegans lines To generate cnnm-1p::GFP, the 4,333-bp fragment of the 5′ region of cnnm-1 was amplified by PCR and then inserted into the GFP expression vector pPD95.77 (kindly provided by A. Fire). To generate the cnnm-1 genomic construct, the cnnm-1 genomic fragment (−4,933 to +7,492 relative to the ATG start codon) was obtained by restriction enzyme digestion from the fosmid WRM0636cE07 (Dnaform). The remaining portion of the cnnm-1 genomic fragment (+7,493 to +8,924) was generated by PCR amplification. For the cnnm-3 genomic construct, the cnnm-3 genomic fragment (−2,935 to +4,090) was obtained by PCR amplification. Thereafter, each genomic fragment was inserted into pBluescript KS (Stratagene). For the cnnm-1::Venus translational fusion construct, the same fosmid fragment as that used for the cnnm-1 genomic construct was linked to the cnnm-1 genomic fragment (+7,493 to +8,610) generated by PCR amplification. The fragments were then inserted into pPD95.79-venus (kindly provided by T. Ishihara). To express the venus-fusion protein of CNNM-3, the cnnm-3 genomic fragment (−2,935 to −1) and the cnnm-3 cDNA were both prepared by PCR, and these fragments were then inserted into pPD95.79-venus. To express cnnm-1 or cnnm-3 under the control of the ges-1 promoter or aex-3 promoter, cnnm-1 or cnnm-3 cDNA was prepared by RT-PCR and then inserted into pDEST-ges-1p and pDEST-aex-3p, kindly provided by H. Kuroyanagi [50,51]. To generate plasmids for feeding RNAi experiments (to target genes that are not included in the Ahringer library), cDNA fragments for let-363 and daf-15 were prepared by RT-PCR with the following primer sets: let-363; 5′-ACTAGTGCCGATAGACAGAACAAAGCAGCC-3′ and 5′-GTGGTACCGGACAAGCCATTCAACACCTTC-3′; daf-15; 5′-GTGCTAGCCCTCGTTTGCAGAACGTTTGAC-3′ and 5′-AGGTACCCCAGTTGAGCTCTCCGAGCACAG-3′. Amplified fragments were then inserted into L4440 (kindly provided by A. Fire). DNA fragments were inserted by conventional methods utilizing the restriction enzymes and ligases, with the exception of the expression construct for venus-fused CNNM-3, which was generated using the Gibson assembly method [52]. The DNA sequences of all PCR products were confirmed by sequencing. To generate transgenic lines, plasmids were injected into N2 or cnnm-1(gk222902)/nT1[qIs51]; cnnm-3(dcr2) along with rol-6(su1006) [53] or rab-3p::mCherry (Addgene) as an injection marker. Isolation of cnnm mutant strains The mutant strains of cnnm-2(dcr1), cnnm-3(dcr2), and cnnm-4(dcr3) in this study were isolated from the Trimethylpsoralen/UV-mutagenized library by performing nested PCR as described previously [54]. The primer sets used for screening were as follows: cnnm-2 first round; 5′-TGTCCCGTTTGATGGAAAAT-3′ and 5′-TTTGGAACTATCGTGCCTCC-3′; cnnm-2 second round; 5′-CGAGGATGGTAGAAATGCTCA-3′ and 5′-TACCTGTGGCATCATGGTTG-3′. cnnm-3 first round; 5′-TTGATTAGCGGCAATAAGGG-3′ and 5′-ATATGCCAAAATGGCTTTCG-3′; cnnm-3 second round; 5′-GCTCACCATTCAACGATTCA-3′ and 5′-ATGAACTCACGAGGTGTCGG-3′. cnnm-4 first round; 5′-CATTTTTCAGCGAGCCTTTC-3′ and 5′-CCCATCTTCTTCCGAATCAA-3′; cnnm-4 second round; 5′-CTTTGCCTCGGTTTATCTGC-3′ and 5′-AGACGTGAATGGCCTTGTTC-3′. The cnnm-1(gk222902) and the cnnm-5(ttTi19567) alleles were generated by the C. elegans Reverse Genetics Core Facility at the University of British Columbia and the NEMAGENETAG Project, respectively. Germ cell counts Germ cells were stained as previously described [55] with slight modifications. L4/young adults were permeabilized using the freeze–crack method and sequentially fixed in cold methanol for 10 min and in cold acetone for 10 min. The samples were blocked with 2% bovine serum albumin in PBS-T (PBS containing 0.05% Tween 20) for 30 min at room temperature, and then incubated overnight at 4°C with mouse anti-PGL-1 antibody K76 (1:20 dilution), developed by S. Strome [56] and provided by the Developmental Studies Hybridoma Bank. This was followed by incubation with Alexa Fluor 568 goat anti-mouse IgG (1:2000 dilution, Invitrogen) for 2 h at room temperature. Coverslips were mounted on a microscopic glass slide. Next, the samples were observed using a microscope, and PGL-1-positive cells were counted as germ cells. Gonadal arm extension To analyze gonadal arm extension, lag-2p::GFP was used to visualize the Z1/Z4 cells and DTCs. The worms were analyzed at the L4/young adult stage, unless otherwise noted, using a microscope. Based on visual observation of the lag-2p::GFP signals and the phase contrast view, we determined the outline of the gonad, and worms with gonads of ≥ 60 μm in length of the long axis (3-fold length of the primordial gonad) were defined as extended. Lifespan assay A lifespan assay was performed as described previously [20] with slight modifications, starting with L4/young adults. To remove contamination with progeny, worms were transferred to fresh NGM-lite plates seeded with OP-50 every 2 days until day 8, after which only those worms on plates where progeny was observed were transferred. Survival was monitored daily. Worms that did not move, respond to nose touch with a platinum picker, or exhibit pumping were determined as dead and were removed. Worms that crawled off the plate, had a protruded vulva, or died by internal hatching were excluded. Body size measurement Mixed stage worms from L2 to adult (L4 to adult worms were most abundant) were anesthetized with M9 buffer containing 50 mM NaN3 and were mounted on a 3% agarose pad on a microscopic glass slide. Thereafter, the worms were examined for alae formation using a microscope and then photographed for body size measurement. The area of the worms was directly measured from the images using Image J (NIH software). The body size at the transition from L4 to adult molt was determined by calculating the mean body size of the smallest three worms with alae and the largest three worms without alae. Male tail observation Observation of the male tail was performed as previously described [57]. Adult worms were anesthetized with 10 mM sodium azide and transferred to a 5% agar pad. The worms were turned over with a pick to the ventral side up and immediately covered with a coverslip. Genome wide RNAi screening Feeding RNAi was performed as described previously [58]. In total, 15,357 bacterial RNAi feeding strains from the Ahringer library [29] were tested as follows in the first round screening (n = 5–10 worms per strain), using bacteria carrying the empty vector L4440 as the negative control. cnnm-1(gk222902)/nT1[qIs51]; cnnm-3(dcr2); rrf-3(pk1426); qIs56[lag-2p::GFP] gravid adults were bleached, and synchronized P0 worms at the L1 stage were transferred to RNAi plates. F1 progeny lacking the balancer nT1[qIs51] (without pharyngeal GFP expression) were phenotypically scored at the L4/young adult stage as described in the “Gonadal arm extension” section. The wells were scored as positive if gonadal arms extended in more than 50% of worms in the F1 generation. Some wells showed larval arrest or sterile phenotypes in the P0 generation, and therefore, these wells were scored in the P0 generation. The first round screening led to the identification of 119 positive wells, which was followed by sequence analyses to identify the RNAi clone in each well. Because some wells contained multiple clones, we re-transformed each of the 135 sequence-verified clones into the HT115 E. coli strain and performed the second round screening as described above (n = 30 worms per clone). ICP-MS To measure magnesium levels in whole worms, 300 synchronized L4/young adult worms were incubated for 30 min with washing buffer containing 110 mM HNO3 (semiconductor grade, Wako) and 187 mM NH3 (ultrapure grade, Kanto Chemical), which corresponds to approximately 300 mOsm/l and pH 7.0–8.0 at room temperature, and were then washed 5 times with washing buffer. Subsequently, worms were boiled at 95°C for 5 min and sonicated using Bioruptor (UCD-250HSA; Cosmo Bio). The homogenates were completely dried by incubation at 98°C, and then subjected to treatment with 100 μl of 40% HNO3 at 95°C for 2 h. The solution was diluted to 1 ml with ddH2O and magnesium levels were determined using ICP-MS (7700x; Agilent), according to the manufacturer’s instructions. The magnesium levels were normalized to total protein levels, which were determined using the BCA assay kit (Thermo Scientific). A blank sample was prepared using the same procedure without worms. To measure magnesium levels in the intestine, approximately 300 synchronized L4/young adult worms were cut with a scalpel just behind the pharynx in a drop of washing buffer. The extruded intestine was cut away from the remnants of the body, and the isolated intestines were then washed twice with washing buffer. Magnesium levels were analyzed as described above. Volume calculation Body volume was measured as described previously [59,60] with slight modifications. L4/young adults were transferred to M9 buffer containing 50 mM NaN3 on coverslips and then photographed. The total body volume of the worms was calculated by assuming that the body shape was composed of two cones (from the tip of the nose to the anterior end of the intestine, and from the posterior end of the intestine to the tip of the tail) and a cylinder (remaining body part) and by measuring each length and radius. Intestine volume was calculated by assuming that the shape was cylindrical. Microscopy Fluorescence images were acquired using an inverted microscope (IX81; Olympus) equipped with a laser scanning confocal imaging system (FluoView FV1000; Olympus). Nomarski images were collected concurrently or alone using the same microscope using Nomarski optics. A multiline argon laser and image analysis system (FV10-ASW; Olympus) were also used for image acquisition. To analyze gonadal arm extension, lag-2p::GFP fluorescence was observed using a stereo microscope (SZX7; Olympus) equipped with a U-RFL-T 100W mercury lamp (U-RFL-T; Olympus). Statistics All statistical analyses were performed using GraphPad Prism 6 software (GraphPad Software) and are presented as the mean ± SEM. p values were obtained by Student’s two-tailed t-test for Table 1 and by ANOVA, followed by two-tailed multiple Student’s t-test with Tukey’s correction for Figs 2C and 3B. For lifespan assays (Figs 2D, 3C and S2 Fig), we used the log rank (Mantel-Cox) test. Bonferroni correction was applied to multiple comparisons of lifespans. Supporting Information S1 Fig Alignment of amino acid sequences of C. elegans CNNM family proteins. Asterisk (*) and colon (:) denote identical residues and conserved substitutions, respectively. The sequence of H. sapiens CNNM4 is also shown. The regions for functionally essential DUF21 and CBS domains, and the amino acids for which coding nucleotide sequences were directly lost/changed by each genetic alteration, are highlighted. cnnm-1(gk222902) contains a point mutation that led to a premature stop codon in place of arginine residue 255 in the DUF21 domain. cnnm-2(dcr1) contains a deletion of 162 nucleotides from the inside of exon 11 to the inside of exon 12, leading to the loss of 40 amino acids in the CBS domain. cnnm-3(dcr2) contains a deletion of 289 nucleotides that include the splice acceptor site of intron 4 and 183 nucleotides in the following exon 5. This deletes 61 amino acids in the DUF21 domain and causes additional deletion/alterations because of incorrect splicing. cnnm-4(dcr3) contains a deletion of 173 nucleotides that include the entire exon 5 and the splice acceptor and donor sites in adjacent introns. This deletes exon 5-encoded 41 amino acids in the DUF21 domain, and causes frameshifts if RNA splicing occurs by directly linking exon 4 and exon 6. cnnm-5(ttTi19567) contains the Mos 1 sequence inserted in exon 2, resulting in a truncated product that lacks both the DUF21 and CBS domains. (TIF) Click here for additional data file. S2 Fig Overexpression of CNNM can extend lifespan. For each genotype, 45 synchronized L4/young adult worms were transferred to fresh plates (15 worms per plate) and then scored daily for survival. The graph represents data combined from at least two experiments. Mean lifespan (± SEM) of worms is also indicated in parentheses. p values were determined by log rank (Mantel-Cox) test, and the Bonferroni method was then used to correct for multiple comparisons. ****p < 0.0001. (TIF) Click here for additional data file. S3 Fig cnnm-1; cnnm-3 mutant male worms show normal tail ray development. Nomarski images of the ventral view of adult wild-type and cnnm-1; cnnm-3 mutant mail tails. The anterior side of the worm is positioned to the left. Bar, 20 μm. (TIF) Click here for additional data file. S4 Fig Representative images showing the effect of RNAi screening. Representative images of L4/young adult worms of cnnm-1; cnnm-3; rrf-3 mutants with Ex[lag-2p::GFP], which were fed with bacteria carrying the dsRNA corresponding to the indicated genes (the results of the top 5 genes in second round screening are shown). In each image, the anterior side of the worm is positioned to the left. Dotted lines indicate the outlines of the gonad. As a negative control, the worms were fed with bacteria carrying the empty vector L4440. Bar, 20 μm. (TIF) Click here for additional data file. S5 Fig Importance of TORC1 signaling. cnnm-1; cnnm-3; aak-1; aak-2 mutant worms, carrying lag-2p::GFP, were fed with bacteria carrying the dsRNA corresponding to the indicated genes. Because RNAi of either let-363 or daf-15 is known to cause L3 larval arrest [38], we observed the worms at the L2–L3 stages. In each image, the anterior side of the worm is positioned to the left. Dotted lines indicate the outlines of the gonad. Bar, 20 μm. Worms with gonads of ≥ 60 μm in length of the long axis (3-fold as long as the primordial gonad) were determined as extended and the results (%) are shown in the right graph. More than 50 worms were analyzed for each condition. (TIF) Click here for additional data file. S1 Dataset Numerical data for figures and tables. The underlying numerical data for each figure or table are shown in separate sheets. (XLSX) Click here for additional data file. We thank T. Hayashi, H. Yamamoto, and Y. Tsumori (Osaka University, Japan) for their technical assistance. ==== Refs References 1 Schweigel M , Martens H . Magnesium transport in the gastrointestinal tract . Front Biosci . 2000 ;5 : D666 –D677 . 10922297 2 Dimke H , Hoenderop JG , Bindels RJ . Molecular basis of epithelial Ca2+ and Mg2+ transport: insights from the TRP channel family . J Physiol . 2011 ;589 : 1535 –1542 . 10.1113/jphysiol.2010.199869 21041532 3 Voets T , Nilius B , Hoefs S , van der Kemp AW , Droogmans G , Bindels RJ , et al TRPM6 Forms the Mg2+ Influx Channel Involved in Intestinal and Renal Mg2+ Absorption . J Biol Chem . 2004 ;279 : 19 –25 . 14576148 4 Schlingmann KP , Weber S , Peters M , Niemann Nejsum L , Vitzthum H , Klingel K , et al Hypomagnesemia with secondary hypocalcemia is caused by mutations in TRPM6, a new member of the TRPM gene family . Nat Genet . 2002 ;31 : 166 –170 . 12032568 5 Walder RY , Landau D , Meyer P , Shalev H , Tsolia M , Borochowitz Z , et al Mutation of TRPM6 causes familial hypomagnesemia with secondary hypocalcemia . Nat Genet . 2002 ;31 : 171 –174 . 12032570 6 Ryazanova LV , Rondon LJ , Zierler S , Hu Z , Galli J , Yamaguchi TP , et al TRPM7 is essential for Mg(2+) homeostasis in mammals . Nat Commun . 2010 ;1 : 109 21045827 7 Wang CY , Shi JD , Yang P , Kumar PG , Li QZ , Run QG , et al Molecular cloning and characterization of a novel gene family of four ancient conserved domain proteins (ACDP) . Gene . 2003 ;306 : 37 –44 . 12657465 8 Meyer TE , Verwoert GC , Hwang SJ , Glazer NL , Smith AV , van Rooij FJ , et al Genome-wide association studies of serum magnesium, potassium, and sodium concentrations identify six Loci influencing serum magnesium levels . PLoS Genet . 2010 ;6 : e1001045 10.1371/journal.pgen.1001045 20700443 9 Stuiver M , Lainez S , Will C , Terryn S , Günzel D , Debaix H , et al CNNM2, encoding a basolateral protein required for renal Mg2+ handling, is mutated in dominant hypomagnesemia . Am J Hum Genet . 2011 ;88 : 333 –343 . 10.1016/j.ajhg.2011.02.005 21397062 10 Yamazaki D , Funato Y , Miura J , Sato S , Toyosawa S , Furutani K , et al Basolateral Mg2+ extrusion via CNNM4 mediates transcellular Mg2+ transport across epithelia: a mouse model . PLoS Genet . 2013 ;9 : e1003983 10.1371/journal.pgen.1003983 24339795 11 de Baaij JH , Stuiver M , Meij IC , Lainez S , Kopplin K , Venselaar H , et al Membrane topology and intracellular processing of cyclin M2 (CNNM2) . J Biol Chem . 2012 ;287 : 13644 –13655 . 10.1074/jbc.M112.342204 22399287 12 Hirata Y , Funato Y , Takano Y , Miki H . Mg2+-dependent interactions of ATP with the cystathionine- β-synthase (CBS) domains of a magnesium transporter . J Biol Chem . 2014 ;289 : 14731 –14739 . 10.1074/jbc.M114.551176 24706765 13 Funato Y , Yamazaki D , Mizukami S , Du L , Kikuchi K , Miki H . Membrane protein CNNM4-dependent Mg2+ efflux suppresses tumor progression . J Clin Invest . 2014 ;124 : 5398 –5410 . 10.1172/JCI76614 25347473 14 Hardy S , Uetani N , Wong N , Kostantin E , Labbé DP , Bégin LR , et al The protein tyrosine phosphatase PRL-2 interacts with the magnesium transporter CNNM3 to promote oncogenesis . Oncogene . 2015 ;34 : 986 –995 . 10.1038/onc.2014.33 24632616 15 Teramoto T , Lambie EJ , Iwasaki K . Differential regulation of TRPM channels governs electrolyte homeostasis in the C. elegans intestine . Cell Metab . 2005 ;1 : 343 –354 . 16054081 16 Teramoto T , Sternick LA , Kage-Nakadai E , Sajjadi S , Siembida J , Mitani S , et al Magnesium Excretion in C. elegans Requires the Activity of the GTL-2 TRPM Channel . PLoS One . 2010 ;5 : e9589 10.1371/journal.pone.0009589 20221407 17 Sharma-Kishore R , White JG , Southgate E , Podbilewicz B . Formation of the vulva in Caenorhabditis elegans: a paradigm for organogenesis . Development . 1999 ;126 : 691 –699 . 9895317 18 Sulston JE , Horvitz HR . Post-embryonic cell lineages of the nematode, Caenorhabditis elegans . Dev Biol . 1977 ;56 : 110 –156 . 838129 19 Singh RN , Sulston JE . Some Observations On Moulting in Caenorhabditis Elegans . Nematologica . 1978 ;24 : 63 –71 . 20 Kenyon C , Chang J , Gensch E , Rudner A , Tabtiang R . A C. elegans mutant that lives twice as long as wild type . Nature . 1993 ;366 : 461 –464 . 8247153 21 Kimble J , Hirsh D . The postembryonic cell lineages of the hermaphrodite and male gonads in Caenorhabditis elegans . Dev Biol . 1979 ;70 : 396 –417 . 478167 22 Blelloch R , Anna-Arriola SS , Gao D , Li Y , Hodgkin J , Kimble J . The gon-1 gene is required for gonadal morphogenesis in Caenorhabditis elegans . Dev Biol . 1999 ;216 : 382 –393 . 10588887 23 Mathies LD , Schvarzstein M , Morphy KM , Blelloch R , Spence AM , Kimble J . TRA-1/GLI controls development of somatic gonadal precursors in C. elegans . Development . 2004 ;131 : 4333 –4343 . 15294864 24 Sun AY , Lambie EJ . gon-2, a gene required for gonadogenesis in Caenorhabditis elegans . Genetics . 1997 ;147 : 1077 –1089 . 9383054 25 Sulston JE , Albertson DG , Thomson JN . The Caenorhabditis elegans male: postembryonic development of nongonadal structures . Dev Biol . 1980 ;78 : 542 –576 . 7409314 26 Edgar LG , McGhee JD . Embryonic expression of a gut-specific esterase in Caenorhabditis elegans . Dev Biol . 1986 ;114 : 109 –118 . 3956859 27 Iwasaki K , Staunton J , Saifee O , Nonet M , Thomas JH . aex-3 encodes a novel regulator of presynaptic activity in C. elegans . Neuron . 1997 ;18 : 613 –622 . 9136770 28 Cox EA , Hardin J . Sticky worms: adhesion complexes in C. elegans . J Cell Sci . 2004 ;117 : 1885 –1897 . 15090594 29 Kamath RS , Fraser AG , Dong Y , Poulin G , Durbin R , Gotta M , et al Systematic functional analysis of the Caenorhabditis elegans genome using RNAi . Nature . 2003 ;421 : 231 –237 . 12529635 30 Simmer F , Moorman C , van der Linden AM , Kuijk E , van den Berghe PV , Kamath RS , et al Genome-wide RNAi of C. elegans using the hypersensitive rrf-3 strain reveals novel gene functions . PLoS Biol . 2003 ;1 : E12 14551910 31 Hardie DG , Ross FA , Hawley SA . AMPK: a nutrient and energy sensor that maintains energy homeostasis . Nat Rev Mol Cell Biol . 2012 ;13 : 251 –262 . 10.1038/nrm3311 22436748 32 Romani AM . Cellular magnesium homeostasis . Arch Biochem Biophys . 2011 ;512 : 1 –23 . 10.1016/j.abb.2011.05.010 21640700 33 Günther T . Concentration, compartmentation and metabolic function of intracellular free Mg2+ . Magnes Res . 2006 ;19 : 225 –236 . 17402290 34 Apfeld J , O’Connor G , McDonagh T , DiStefano PS , Curtis R . The AMP-activated protein kinase AAK-2 links energy levels and insulin-like signals to lifespan in C. elegans . Genes Dev . 2004 ;18 : 3004 –3009 . 15574588 35 Yigit E , Batista PJ , Bei Y , Pang KM , Chen CC , Tolia NH , et al Analysis of the C. elegans Argonaute family reveals that distinct Argonautes act sequentially during RNAi . Cell . 2006 ;127 : 747 –757 . 17110334 36 Martynovsky M , Wong MC , Byrd DT , Kimble J , Schwarzbauer JE . mig-38, a novel gene that regulates distal tip cell turning during gonadogenesis in C. elegans hermaphrodites . Dev Biol . 2012 ;368 : 404 –414 . 10.1016/j.ydbio.2012.06.011 22732572 37 Hardie DG . AMPK-sensing energy while talking to other signaling pathways . Cell Metab . 2014 ;20 : 939 –952 . 10.1016/j.cmet.2014.09.013 25448702 38 Hara K , Maruki Y , Long X , Yoshino K , Oshiro N , Hidayat S , et al Raptor, a binding partner of target of rapamycin (TOR), mediates TOR action . Cell . 2002 ;110 : 177 –189 . 12150926 39 Inoki K , Zhu T , Guan KL . TSC2 mediates cellular energy response to control cell growth and survival . Cell . 2003 ;115 : 577 –590 . 14651849 40 Gwinn DM , Shackelford DB , Egan DF , Mihaylova MM , Mery A , Vasquez DS , et al AMPK phosphorylation of raptor mediates a metabolic checkpoint . Mol Cell . 2008 ;30 : 214 –226 . 10.1016/j.molcel.2008.03.003 18439900 41 Martin TD , Chen XW , Kaplan RE , Saltiel AR , Walker CL , Reiner DJ , et al Ral and Rheb GTPase activating proteins integrate mTOR and GTPase signaling in aging, autophagy, and tumor cell invasion . Mol Cell . 2014 ;53 : 209 –220 . 10.1016/j.molcel.2013.12.004 24389102 42 Rubin H . Central role for magnesium in coordinate control of metabolism and growth in animal cells . Proc Natl Acad Sci U S A . 1975 ;72 : 3551 –3555 . 171666 43 Rubin H . The logic of the Membrane, Magnesium, Mitosis (MMM) model for the regulation of animal cell proliferation . Arch Biochem Biophys . 2007 ;458 : 16 –23 . 16750508 44 Sahni J , Scharenberg AM . TRPM7 ion channels are required for sustained phosphoinositide 3-kinase signaling in lymphocytes . Cell Metab . 2008 ;8 : 84 –93 . 10.1016/j.cmet.2008.06.002 18590694 45 Fukuyama M , Sakuma K , Park R , Kasuga H , Nagaya R , Atsumi Y , et al C. elegans AMPKs promote survival and arrest germline development during nutrient stress . Biol Open . 2012 ;1 : 929 –936 . 10.1242/bio.2012836 23213370 46 Hong Y , Roy R , Ambros V . Developmental regulation of a cyclin-dependent kinase inhibitor controls postembryonic cell cycle progression in Caenorhabditis elegans . Development . 1998 ;125 : 3585 –3597 . 9716524 47 Fukuyama M , Rougvie AE , Rothman JH . C. elegans DAF-18/PTEN mediates nutrient-dependent arrest of cell cycle and growth in the germline . Curr Biol . 2006 ;16 : 773 –779 . 16631584 48 Watanabe S , Yamamoto TG , Kitagawa R . Spindle assembly checkpoint gene mdf-1 regulates germ cell proliferation in response to nutrition signals in C. elegans . EMBO J . 2008 ;27 : 1085 –1096 . 10.1038/emboj.2008.32 18309291 49 Vallin E , Gallagher J , Granger L , Martin E , Belougne J , Maurizio J , et al A genome-wide collection of Mos1 transposon insertion mutants for the C. elegans research community . PLoS One . 2012 ;7 : e30482 10.1371/journal.pone.0030482 22347378 50 Kuroyanagi H , Kobayashi T , Mitani S , Hagiwara M . Transgenic alternative-splicing reporters reveal tissue-specific expression profiles and regulation mechanisms in vivo . Nat Methods . 2006 ;3 : 909 –915 . 17060915 51 Kuroyanagi H , Ohno G , Sakane H , Maruoka H , Hagiwara M . Visualization and genetic analysis of alternative splicing regulation in vivo using fluorescence reporters in transgenic Caenorhabditis elegans . Nat Protoc . 2010 ;5 : 1495 –1517 . 10.1038/nprot.2010.107 20725066 52 Gibson DG , Young L , Chuang RY , Venter JC , Hutchison CA 3rd, Smith HO . Enzymatic assembly of DNA molecules up to several hundred kilobases . Nat Methods . 2009 ;6 : 343 –345 . 10.1038/nmeth.1318 19363495 53 Mello CC , Kramer JM , Stinchcomb D , Ambros V . Efficient gene transfer in C.elegans: extrachromosomal maintenance and integration of transforming sequences . EMBO J . 1991 ;10 : 3959 –3970 . 1935914 54 Kubota Y , Kuroki R , Nishiwaki K . A fibulin-1 homolog interacts with an ADAM protease that controls cell migration in C. elegans . Curr Biol . 2004 ;14 : 2011 –2018 . 15556863 55 Kawasaki I , Shim YH , Kirchner J , Kaminker J , Wood WB , Strome S . PGL-1, a predicted RNA-binding component of germ granules, is essential for fertility in C. elegans . Cell . 1998 ;94 : 635 –645 . 9741628 56 Strome S , Wood WB . Generation of asymmetry and segregation of germ-line granules in early C. elegans embryos . Cell . 1983 ;35 : 15 –25 . 6684994 57 Suzuki Y , Yandell MD , Roy PJ , Krishna S , Savage-Dunn C , Ross RM , et al A BMP homolog acts as a dose-dependent regulator of body size and male tail patterning in Caenorhabditis elegans . Development . 1999 ;126 : 241 –250 . 9847238 58 Kamath RS , Martinez-Campos M , Zipperlen P , Fraser AG , Ahringer J . Effectiveness of specific RNA-mediated interference through ingested double-stranded RNA in Caenorhabditis elegans . Genome Biol . 2001 ;2 : RESEARCH0002 11178279 59 Choe KP , Strange K . Evolutionarily conserved WNK and Ste20 kinases are essential for acute volume recovery and survival after hypertonic shrinkage in Caenorhabditis elegans . Am J Physiol Cell Physiol . 2007 ;293 : C915 –C927 . 17596296 60 Banerjee S , Versaw WK , Garcia LR . Imaging Cellular Inorganic Phosphate in Caenorhabditis elegans Using a Genetically Encoded FRET-Based Biosensor . PLoS One . 2015 ;10 : e0141128 10.1371/journal.pone.0141128 26484766

PMC005xxxxxx/PMC5001715.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756438010.1371/journal.pone.0157311PONE-D-16-02463Research ArticleBiology and life sciencesCell biologyCellular typesAnimal cellsBlood cellsWhite blood cellsNK cellsBiology and life sciencesCell biologyCellular typesAnimal cellsImmune cellsWhite blood cellsNK cellsBiology and life sciencesImmunologyImmune cellsWhite blood cellsNK cellsMedicine and health sciencesImmunologyImmune cellsWhite blood cellsNK cellsBiology and Life SciencesCell BiologyCellular TypesAnimal CellsBlood CellsWhite Blood CellsMacrophagesBiology and Life SciencesCell BiologyCellular TypesAnimal CellsImmune CellsWhite Blood CellsMacrophagesBiology and Life SciencesImmunologyImmune CellsWhite Blood CellsMacrophagesMedicine and Health SciencesImmunologyImmune CellsWhite Blood CellsMacrophagesBiology and Life SciencesCell BiologyCellular TypesAnimal CellsBlood CellsWhite Blood CellsT CellsBiology and Life SciencesCell BiologyCellular TypesAnimal CellsImmune CellsWhite Blood CellsT CellsBiology and Life SciencesImmunologyImmune CellsWhite Blood CellsT CellsMedicine and Health SciencesImmunologyImmune CellsWhite Blood CellsT CellsResearch and Analysis MethodsHistochemistry and Cytochemistry TechniquesImmunohistochemistry TechniquesResearch and Analysis MethodsImmunologic TechniquesImmunohistochemistry TechniquesMedicine and Health SciencesClinical MedicineClinical ImmunologyImmune DeficiencyBiology and Life SciencesImmunologyClinical ImmunologyImmune DeficiencyMedicine and Health SciencesImmunologyClinical ImmunologyImmune DeficiencyBiology and Life SciencesBiochemistryLipidsCholesterolBiology and Life SciencesBiochemistryProteinsLipoproteinsMedicine and Health SciencesVascular MedicineAtherosclerosisCombined B, T and NK Cell Deficiency Accelerates Atherosclerosis in BALB/c Mice Immunodeficiency and Atherosclerosis in BALB/c MiceCheng Fei 1Twardowski Laura 2Reifenberg Kurt 3Winter Kerstin 1Canisius Antje 4Pross Eva 1Fan Jianglin 5Schmitt Edgar 6Shultz Leonard D. 7Lackner Karl J. 4Torzewski Michael 2*1 Dr. Margarete Fischer-Institute of Clinical Pharmacology, Stuttgart and University of Tübingen, Tübingen, Germany2 Department of Laboratory Medicine, Robert-Bosch-Hospital, Stuttgart, Germany3 Center for Preclinical Research, German Cancer Research Center, Heidelberg, Germany4 Institute of Clinical Chemistry and Laboratory Medicine, University Medical Center, Johannes Gutenberg University Mainz, Mainz, Germany5 Department of Molecular Pathology, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, Yamanashi, Japan6 Institute for Immunology, University Medical Center, Johannes Gutenberg University Mainz, Mainz, Germany7 The Jackson Laboratory, Bar Harbor, Maine, United State of AmericaBonsal Geetha P. EditorTulane University, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Conceived and designed the experiments: FC LT KR MT. Performed the experiments: FC LT KR KW AC EP. Analyzed the data: FC LT EP MT. Contributed reagents/materials/analysis tools: JF ES LDS KJL MT. Wrote the paper: FC LT KR MT. * E-mail: michael.torzewski@rbk.de26 8 2016 2016 11 8 e015731119 1 2016 29 5 2016 © 2016 Cheng et al2016Cheng et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.This study focused on the unique properties of both the Ldlr knockout defect (closely mimicking the human situation) and the BALB/c (C) inbred mouse strain (Th-2 slanted immune response). We generated two immunodeficient strains with severe combined B- and T-cell immunodeficiency with or without a complete lack of natural killer cells to revisit the role of adaptive immune responses on atherogenesis. C-Ldlr-/- Rag1-/- mice, which show severe combined B- and T-cell immunodeficiency and C-Ldlr-/- Rag1-/- Il2rg-/- mice, which combine the T- and B-cell defect with a complete lack of natural killer cells and inactivation of multiple cytokine signalling pathways were fed an atherogenic Western type diet (WTD). Both B6-Ldlr-/- and C-Ldlr-/- immunocompetent mice were used as controls. Body weights and serum cholesterol levels of both immunodeficient strains were significantly increased compared to C-Ldlr-/- controls, except for cholesterol levels of C-Ldlr-/- Rag1-/- double mutants after 12 weeks on the WTD. Quantification of the aortic sinus plaque area revealed that both strains of immunodeficient mice developed significantly more atherosclerosis compared to C-Ldlr-/- controls after 24 weeks on the WTD. Increased atherosclerotic lesion development in C-Ldlr-/- Rag1-/- Il2rg-/- triple mutants was associated with significantly increased numbers of macrophages and significantly decreased numbers of smooth muscle cells compared to both C-Ldlr-/- wild type and C-Ldlr-/- Rag1-/- double mutants pointing to a plaque destabilizing effect of NK cell loss. Collectively, the present study reveals a previously unappreciated complexity with regard to the impact of lymphocytes on lipoprotein metabolism and the role of lymphocyte subsets in plaque composition. Robert Bosch Stiftung (DE)Torzewski Michael This work was supported by the Robert Bosch Foundation, Germany. MT received the funding. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Manipulation of the immune system by a variety of means has been used to unravel the roles of cellular and humoral immunity in the pathogenesis of atherosclerosis [1]. Although the absence of autoantibodies and T lymphocytes did not influence [2] or only played a minor role [3] in the extent of aortic atherosclerotic lesions in Apoe-deficient mice, studies in Ldlr deficient mice suggest that T lymphocytes exert pro-inflammatory effects early in atherogenesis [4]. In contrast, subsets of B lymphocytes and/or antibodies are protective against atherosclerosis both in Apoe-[5] and Ldlr-[6] deficient mice. Unlike conventional T cells that recognize peptide antigens presented by the classical MHC class I or class II molecules, natural killer T (NKT) cells are specific for glycolipid antigens presented by the MHC class I-like molecule CD1 [7]. CD1d-restricted NKT cells are pro-atherogenic in both Apoe- [8] and Ldlr- [8–10] deficient mice. Depending on the local milieu of cytokines, CD4+ T lymphocytes differentiate into a T-helper (Th) 1 or Th 2 lineage. Cytokines associated with a pro-inflammatory Th-1 response (interferon (IFN)-γ and interleukin (IL)-12) play a role in atherogenesis while Th2 associated cytokines including IL-10 are protective against lesion development [11]. It is also known that severe hypercholesterolemia induces a switch from Th1 to Th2 development in Apoe-deficient mice resulting not only in the formation of IgG1 autoantibodies to oxidized LDL, but also in the appearance of Th2-type cytokines in the atherosclerotic lesions [12]. The impact of a Th1 versus Th2 immune response on atherogenesis was further detailed in the context of Apoe-deficient mouse strains, which either had a C57BL/6 (B6) background, producing predominantly Th1 helper cells, or which had a BALB/c (C) background, leaning toward Th2 dominated immune responses [13]. However, as Apoe-deficient mice exhibit an impaired immune response during infection with Listeria monocytogenes [14] general concerns over using this model to evaluate immune response in atherosclerosis have been raised, and it has been demonstrated that Ldlr-deficient mice did not exhibit any potentially confounding perturbations in immune response [4]. In addition, since atherosclerotic lesions in Ldlr-deficient mice do not develop spontaneously as in Apoe-deficient animals but are inducible under Western type diet (WTD), and since the serum lipoprotein profile of Ldlr-deficient animals is characterized by high-level LDL rather than by chylomicrons and VLDL (as in Apoe-deficient mice) [15–17], Ldlr-deficient mice represent a more physiological model of atherogenesis compared to Apoe-deficient mice. In the present study, we studied the unique properties of both Ldlr-deficiency (closely mimicking the human situation) and the BALB/C mouse strain (Th2 slanted immune response) and generated two immunodeficient strains with severe combined B- and T-cell immunodeficiency with or without a complete lack of natural killer cells to revisit the role of adaptive immune responses on atherogenesis. Materials and Methods Mouse Strains and Dietary Induction of Atherosclerosis The B6-Ldlr-/- (official designation: B6.129S7-Ldlrtm1Her/J, Stock No 2207) knockout strain was purchased from The Jackson Laboratory (TJL, Bar Harbor, Maine, USA). Generation of the BALB/c-Ldlrtm1Her (internal designation C-Ldlr-/-) mice has been described in Spencer et al[18] and homozygous C.129S7-Ldlrtm1Her (C.Ldlr-/-) breeder mice were obtained from the animal care facility of the University of Washington. C-Ldlr-/- mice carry the atherosclerotic Ldlrtm1Her allele on the inbred BALB/c background which is prone to Th2 immune responses. C.129S7-Ldlrtm1Her Rag1tm1Mom (internal designation C-Ldlr-/- Rag1-/-) double mutants were generated by crossing the C-Ldlr-/- strain to C-Rag1-/- mice [19] (generously provided by the Department of Immunology of the University Medical Center, Mainz) and intercrossing the resulting C-Ldlr+/- Rag1+/- F1 mice. C-Ldlr-/- Rag1-/- mice combine atherogenesis susceptibility with a severe combined B- and T-cell immunodeficiency. C.129S7-Ldlrtm1Her Rag1tm1Mom Il2rgtm1Wjl (internal designation C-Ldlr-/- Rag1-/- Il2rg-/-) triple knockout mutants were generated by crossing C-Ldlr-/- mice to BALB/c.Cg-Rag1tm1Mom Il2rgtm1Wjl mice (internal designation C-Rag1-/- Il2rg-/-), raised by L.D.S at The Jackson Laboratory, crossing the resulting C-Ldlr+/- Rag1+/- Il2rg+/- hybrids to the C-Rag1-/- Il2rg-/- parental strain until homozygosity was reached at the Rag1 and Il2rg loci and intercrossing the resulting C-Ldlr+/- Rag1-/- Il2rg-/- mice. C-Ldlr-/- Rag1-/- Il2rg-/- mice are prone to atherosclerosis, are deficient in mature T-cells, B-cells and natural killer cells and have inactivation of the IL-2, IL-4, IL-7, IL-9, IL-15, and IL-21 cytokine signalling pathways [20]. Genetic authenticity of all strains was monitored commercially (KBioscience, Hoddesdon, UK) using the SNP-based marker set previously developed by The Jackson Laboratory [21]. All mice were maintained at the Central Laboratory Animal Facility of the University Medical Center of Mainz, Germany, under specific pathogen free conditions in accordance with standard animal care requirements and maintained on a 12/12 hour light-dark cycle. Water and food were given ad libitum. Female C-Ldlr-/- Rag1-/- and C-Ldlr-/- Rag1-/- Il2rg-/- knockout mice, as well as female C-Ldlr-/- and B6-Ldlr-/- mice were placed on a pro-atherogenic Western-type diet (WTD, Ssniff Spezialdiäten GmbH, Soest, Germany) at an age of eight weeks. The WTD contained 21% (wt/wt) fat and 0.15% (wt/wt) cholesterol and was administered for a time span of either 12 or 24 weeks, respectively. All animal work performed in this study was conducted according to the national guidelines and was reviewed and confirmed by an institutional review board/ethics committee headed by the local animal welfare officer (Prof. Kempski) of the University Medical Center Mainz, Germany. The animal experiments were finally approved by the responsible national authority, which is the National Investigation Office Rheinland-Pfalz (Koblenz, Germany). The Approval ID assigned by this authority is AZ 23 177-07/G 07-1-003. Lipoprotein and Analysis of Murine Sera Murine sera were diluted 1:3 prior to quantitative cholesterol and triglyceride analyses. Quantitative cholesterol determinations were conducted using a colorimetric assay (CHOP-PAP, Roche® Diagnostics, Mannheim, Germany). Triglycerides were determined by quantifying free glycerine originating from hydrolytic cleavage (GPO-PAP, Roche Diagnostics). Additionally, lipoproteins were isolated by small-volume sequential ultracentrifugation with a Beckman TLA100.2 rotor as described. Isolated individual density fractions and whole sera were resolved by electrophoresis in 1% agarose gels. The gels were dried and stained with Fat Red 7B to identify lipoproteins containing neutral lipid [22]. Lastly, plasma lipoprotein profiles were analyzed by fast protein liquid chromatography (FPLC) gel filtration as described previously [23]. Tissue Preparation and Quantification of Atherosclerotic Lesions Mice fed with WTD were sacrificed by exposure to carbon dioxide and peritoneal cavities were opened. The lung, liver, small intestine, kidney, spleen, heart and aorta were resected en bloc and fixed in 4% PBS-buffered formaldehyde. Hearts were sequentially cut into a total of 30 (3 μm thick) sections around the aortic sinus. Out of 30 sections, every fifth slide was stained with trichrome and computer-assisted measurement of plaque size was performed as described previously [24]. The remaining aortic sinus and spleen sections were used for further immunohistochemistry and histochemistry. Immunohistochemistry and Histochemical Analyses The sections of paraffin-embedded tissues were deparaffinized and antigen-retrieval was achieved by heating the sections in target retrieval solution (Dako Corporation, Denmark)., Slides were treated with 3% H2O2 to block endogenous peroxidase activity. Immunostaining with murine monoclonal antibodies was performed using the Vector M.O.M. immunodetection kit (Vector Laboratories, Burlingham, Calif.). Staining with rat antibodies was performed using the VECTASTAIN Elite ABC Kit (Vector Laboratories), and staining with rabbit antibodies was performed using the Dako REAL EnVision Detection System, rabbit/mouse kit (Dako Corporation, Denmark). After blocking, slides were incubated with primary antibodies listed in Table 1. Reaction products were identified by immersing the slides in diaminobenzidine tetrachloride (DAB) to yield a brown reaction product. The slides were then counterstained with hematoxylin and mounted. Negative controls included replacement of the primary antibody by irrelevant isotype-matched antibodies. Collagen content was analyzed by picrosirius red and polarized light microscopic imaging. Immunohistochemical or picrosirius red staining was quantified by Photoshop-based image analysis as described [25,26]. The ratio of the positively stained area to the total lesion area (percent-positive area) or the number of positively stained cells per mm2 lesion, respectively, was calculated. All quantitative morphometric and immunohistochemical data were collected independently by two experienced operators blinded to the mice genotypes. 10.1371/journal.pone.0157311.t001Table 1 Primary Antibodies for Immunohistochemistry. Antibodies Name Source Company α-Actin human Muscle Actin (HHF35) mAb Mouse Dako cCaspase3 Cleaved Caspase-3 (Asp175) Antibody Rabbit Cell Signaling CD2 Anti-CD2 polyclonal antibody Rabbit Bioss CD56 NCAM1/CD56 polyclonal antibody Rabbit Proteintech CD57 Anti-CD57 rabbit clonal antibody Rabbit DB Biotech F4/80 Anti-mouse F4/80 (CI:A3-1) mAb Rat AbD Serotec Geminin Geminin Polyclonal Antibody Rabbit Proteintech VCAM1 Anti-VCAM1 antibody (EPR5047) Rabbit Abcam mAb: monoclonal antibody Statistical Analyses Data were analysed with Prism 5.0. Most of the outcome parameters determined in this study (macrophage, T-, NK- and smooth muscle (SMC) cellularity as well as collagen content of atherosclerotic lesions) did not follow a normal distribution as judged by Shapiro-Wilk tests. These parameters are thus presented as box-plots with median, interquartile range, minimum and maximum diagrams and their statistical analyses have been performed with the non-parametric Mann-Whitney U tests. Body weights and serum lipid concentrations were found to follow a normal distribution. These data are thus presented as mean (± SD) and were analyzed by t-test of significance. Differences between the mouse genotypes were considered as significant for p-values < 0.05. Results Serum Lipids and Lipoproteins Starting at eight weeks of age female C-Ldlr-/- Rag1-/- and C-Ldlr-/- Rag1-/- Il2rg-/- knockout mice, as well as female C-Ldlr-/- and B6-Ldlr-/- controls were administered an atherogenic WTD for 12 or 24 weeks, respectively. Fig 1A shows the group sizes, body weights, serum cholesterol and triglyceride levels of all genotypes and of the various dietary groups. Body weights and serum cholesterol levels of both C-Ldlr-/- Rag1-/- and C-Ldlr-/- Rag1-/- Il2rg-/- immunodeficient strains were significantly increased compared to C-Ldlr-/- controls (p<0.01), respectively, except for cholesterol levels of C-Ldlr-/- Rag1-/- double mutants after 12 weeks on the WTD. Moreover, cholesterol levels of C-Ldlr-/- Rag1-/- Il2rg-/- triple mutants were higher than those of C-Ldlr-/- Rag1-/- double mutants after 24 weeks on the WTD (p<0.05). Serum triglyceride levels of both C-Ldlr-/- Rag1-/- and C-Ldlr-/- Rag1-/- Il2rg-/- immunodeficient strains were significantly increased compared to C-Ldlr-/- controls after 12 weeks on the WTD (p<0.01 and p<0.05, respectively). However, there, were no significant differences in triglyceride levels after 24 weeks on the WTD. 10.1371/journal.pone.0157311.g001Fig 1 Lipid and Lipoprotein analysis of murine sera. A, Group size (n), body weights,serum cholesterol and serum triglyceride concentrations of C-Ldlr-/- controls, C-Ldlr-/- Rag1-/- and C-Ldlr-/- Rag1-/- Il2rg-/-mice after 12 and 24 weeks (w) on the WTD, respectively. Data are presented as means ± standard deviations. *, ** indicate statistically significant differences (*p < 0.05, ** p<0.01). B, Agarose gel electrophoresis of serum lipoproteins (left panel). Serum (2 μL) was electrophoresed on a 1% agarose gel and stained for neutral lipids (α- and ß-migrating lipoproteins) with Fat Red 7B. Representative lipoprotein profiles (right panel). Pooled plasma (n = 6) collected from mice fed a WTD for 12 weeks was used and analyzed by FPLC as described in the Materials and Methods. To analyze the effect of immunodeficiency on lipoproteins, serum lipoprotein profiles were examined by both agarose gel electrophoresis followed by staining with Fat Red 7B after (Fig 1B, left panel) and FPLC (Fig 1B, right panel). It appears that apoB-containing particles (chylomicrons, VLDLs and LDLs) were increased in triple mutant mice but HDL contents were unchanged compared with other two groups. Atherosclerosis Lesion Development Mice were sacrificed after 12 or 24 weeks on the WTD. We first evaluated whether atherosclerotic lesion development could be observed in mice on the BALB/c background (Fig 2A, left panel). As expected, quantification of the plaque areas (median/interquartile range) of the aortic sinus revealed significant more lesion development in B6-Ldlr-/- mice (12 weeks: 230343/184909, 24 weeks: 702682/217326 μm2) compared to C-Ldlr-/- mice (12 weeks: 9195/16769, 24 weeks: 40206/32221 μm2). Of note, however, there was sufficient lesion development in C-Ldlr-/- mice after both 12 and 24 weeks on the WTD. 10.1371/journal.pone.0157311.g002Fig 2 Atherosclerotic lesion development. A, Quantification of cross-sections of the aortic sinus area of B6-Ldlr-/- and C-Ldlr-/- mice (left panel), C-Ldlr-/-, C-Ldlr-/- Rag1-/- and C-Ldlr-/- Rag1-/- Il2rg-/- mice (right panel) after 12 weeks (12 w), and 24 weeks (24 w) on the WTD, respectively. Lesion plaque areas were presented as boxplots with median, interquartile range, minimum, and maximum. *, ** indicate statistically significant differences (* p < 0.05, ** p<0.01), n. s. = not significant. B, Representative specimens of the aortic sinus area of C-Ldlr-/- (upper left panel), B6-Ldlr-/- (upper right panel), C-Ldlr-/- Rag1-/- (lower left panel) and C-Ldlr-/- Rag1-/- Il2rg-/- (lower right panel) mice after 24 weeks on the WTD (trichrome staining). Next, we compared the two immunodeficient strains with the C-Ldlr-/- controls. As demonstrated in the right panel of Fig 2A and in the lower panels of Fig 2B, substantial lesions developed in the aortic sinus area of all animals and the lesions increased in size with time. There were no quantitative differences between the cohorts after 12 weeks on the WTD. Quantification of the plaque areas (median/interquartile range) of the aortic sinus after 24 weeks on the WTD, however, revealed that C-Ldlr-/- Rag1-/- (55916/53488 μm2) and C-Ldlr-/- Rag1-/- Il2rg-/- (44522/34736 μm2) immunodeficent mice developed significantly more atherosclerosis compared to C-Ldlr-/- controls (25285/24975 μm2, p<0,01, Fig 2A, right panel). Representative photomicrographs of the atherosclerotic lesions observed are shown in Fig 2B. Phenotype Analysis of Atherosclerotic Lesions Since immunodeficiency obviously influenced atherosclerosis lesion progression in BALB/c mice, we next compared lesion composition (percent of macrophages, SMCs, number of T- and NK cells, and collagen per aortic sinus area) in the immunodeficient mutants and C-Ldlr-/- controls. However, due to extreme outliers and high interindividual variations at this early time point, we abstained from further phenotype analysis of atherosclerotic lesions at 12 weeks and confined to phenotype analysis after 24 weeks on the WTD (Figs 3 and 4). The macrophage and SMC cellularity as well as collagen content were not significantly different between C-Ldlr-/- Rag1-/- double mutants and C-Ldlr-/- controls (Fig 3A). Interestingly, increased atherosclerotic lesion development in C-Ldlr-/- Rag1-/- Il2rg-/- triple mutants was associated with significantly more macrophages (53,6/22,3%, Fig 3A, left panel), significantly less SMCs (8,5/5,3%, p<0.01, Fig 3A, middle panel) without significantly differences in collagen (1,6/1,9%, Fig 3A, right panel) as compared to both C-Ldlr-/- wild type (macrophages 23,6/18,4%, p<0.01; SMCs 16,5/4,7%, p<0.01; collagen 1,3/1,0%) and C-Ldlr-/- Rag1-/- double mutants (macrophages 22,1/14,6%, p<0.01; SMCs 17,4/9,6%, p<0.01; collagen 2,5/2,1%) suggesting a plaque destabilizing effect of immunodeficiency. These differences, however, were not attributable to different proliferation and /or apoptosis as determined by immunohistochemistry (Fig 4A). Likewise, VCAM-1, a member of the adhesion molecule superfamily, was not expressed differentially in the three animal groups (Fig 4B). 10.1371/journal.pone.0157311.g003Fig 3 Phenotypic analysis of atherosclerotic lesions. A, Analysis of atherosclerotic lesions after 24 weeks (24 w) on the WTD. Atherosclerotic lesions of the aortic sinus of C-Ldlr-/- controls, C-Ldlr-/- Rag1-/- and C-Ldlr-/- Rag1-/- Il2rg-/- mice were quantified for macrophages (left panel), SMCs (middle panel), and collagen (right panel). Percent-positive area for macrophages, SMCs, and collagen was quantified by Photoshop-based image analysis. Data are presented as boxplots with median, interquartile range, minimum, and maximum. B, Representative examples of atherosclerotic lesion composition after 24 weeks on the WTD. Atherosclerotic lesions of the aortic sinus of C-Ldlr-/- controls, C-Ldlr-/- Rag1-/- and C-Ldlr-/- Rag1-/- Il2rg-/- mice were stained with rat anti-mouse F4/80 mAb (left panels) for quantification of macrophages, mouse anti-smooth muscle α-actin (middle panels) for quantification of SMCs, and picrosirius red with subsequent polarization (right panels) for quantification of collagen. The vessel lumen is to the upper left corner. The demarcation between intima and media is indicated by arrowheads. Note that the adventitial tissue (*) also polarizes after picrosirius red staining (internal positive control). 10.1371/journal.pone.0157311.g004Fig 4 Proliferation, apoptosis and VCAM-1 expression in atherosclerotic lesions. A, Proliferation and apoptosis. Left panels, Representative immunohistochemical staining with rabbit polyclonal antibodies against geminin (upper panel) or cleaved caspase 3 (lower panel) in atherosclerotic lesions of the aortic sinus. The aortic lumen is to the upper left corner. Right panels, The number of positively stained cells per atherosclerotic lesion area (mm2) was determined for both geminin (upper panel) and cleaved caspase 3 (lower panel). Note high interindividual variations without any significant differences between the different mouse strains. B, VCAM-1 expression. Representative immunohistochemical staining with rabbit monoclonal antibody against VCAM-1 in atherosclerotic lesions of the aortic sinus. Note VCAM-1 expression in endothelial cells with no significant differences between the three animal groups. The aortic lumen is to the upper left corner. To evaluate the potential role of T and NK cells, the lesional cellularity of CD2+ (surface antigen of the T lymphocyte lineage, not shown), CD56+ (expressed on NK cells and a subset of T cells) and CD57+ (expressed on a subset of NK cells and a subset of T lymphocytes) cells was investigated (Fig 5). However, on average, there were only insignificant numbers or no cells except for three mice with higher numbers of CD57+ cells (Fig 5B). Basically, Nk cells were not detectable in both double and triple mutants (Fig 5A, inserts). 10.1371/journal.pone.0157311.g005Fig 5 NK and NKT cells in atherosclerotic lesions and spleens of C-Ldlr-/- control mice. Representative immunohistochemical staining with a rabbit polyclonal antibody against CD56 (Proteintech, Manchester, UK) (upper panels) or CD57 (Acris Antibodies, Hiddenhausen, Germany) (lower panels) of an atherosclerotic lesion located in the aortic sinus (left) and the spleen as a positive control (right) (A). Note that Nk cells were detectable neither in double nor in triple mutants (inserts). The number of positively stained cells (asterisks) per atherosclerotic lesion area (mm2) was determined (B). The aortic lumen is to the upper left corner. Discussion In the present study, a combined B and T cell immunodeficiency with or without a complete lack of natural killer cells in Th2-prone C-Ldlr knockout mice significantly accelerates atherosclerotic lesion progression after 24 weeks on a WTD. Furthermore, a combined B and T cell immunodeficiency with a complete lack of natural killer cells has the potential to change atherosclerotic cellularity with more macrophages and fewer SMCs. Strikingly, these effects were accompanied by significantly higher body weights as well as serum cholesterol and triglyceride levels in the immunodeficient mice of both strains. Increased serum lipids were found to be caused by elevated apoB-containing lipoproteins such as VLDLs and LDLs as shown in FPLC results. It is currently unknown; however, how immunodeficiency affects lipoprotein metabolism such as increased production of VLDL particles or delayed catabolism of these particles in the liver. In spite of this, immunodeficiency did not influence systemic cytokine levels. At a first glance, these results appear to be contradictory to several previous studies investigating the impact of immunodeficiency on atherogenesis [2–4]. An alternate point of view, however, modifies this first impression and points to some new interesting issues concerning immunodeficiency and atherogenesis. First of all, the significantly higher serum cholesterol and triglyceride levels in the immunodeficient mice of both strains are striking. These differences may be attributed, at least in part, to the BALB/c background. Besides C57BL/6 mice, BALB/c mice are widely used throughout biomedical research with a high susceptibility to atherogenesis following a high fat diet. Thus, depletion of NKT cells resulted in an obese phenotype with insulin resistance, glucose intolerance, body weight gain and lipid accumulation [27]. Another study showed increased triglyceride content, liver injury and inflammation upon Type I NKT depletion [28]. It has previously been demonstrated that thymic NKT cells of BALB/c mice express higher levels of IL-4, which has been linked to increased protection from metabolic dysregulation compared to C57BL/6 mice [29]. Given the different Th1/Th2 balance of C57BL/6 and BALB/c mice, the loss of IL-4 may lead to a more severe phenotype in the latter. Preliminary results of cytokine analyses in the murine sera, however, suggest that the proposed cytokine effects are mediated locally rather than systemically, e. g. by adipose-derived NKT cells [30]. Quite similar observations have been also made on the C57BL/6 background. Thus, adipose tissue invariant NKT cells protected against diet-induced obesity and metabolic disorder through regulatory cytokine production [30]. In another previous study investigating the role of NKT cells in an adoptive transfer model of atherosclerosis using immunodeficient, atherosclerosis-susceptible B6-Rag1-/- Ldlr-/- mice as recipients, significantly higher plasma total cholesterol and triglyceride levels were observed in the recipients of B6-CD1d-/- splenocytes compared with the recipients of B6 splenocytes at the terminal 12-week time point [10]. Furthermore, it cannot be excluded that a lack of NK cells directly raises serum lipoproteins as the latter have been shown to be internalized by NK cells through specific LDL-receptor mediated uptake [31–33]. The significantly higher body weights and serum cholesterol levels in the immunodeficient mice of both strains may underlie the accelerated atherosclerotic lesion progression. Another impact may be attributed to the B cell immunodeficiency, as innate B-1 cells have been demonstrated to play a protective role in atherogenesis both in Apoe- [5,34] and Ldlr-deficient [6] mice through the secretion of naturally occurring antibodies. Of course, T cells also play a role in atherogenesis. An earlier study using a class I MHC deficient C57BL/6J strain has already shown that a lack of suppressor T cells was associated with an increase in lesion formation [35]. Since then, a plethora of studies provided dual roles for T lymphocytes in both suppressing and promoting atherosclerosis depending on different Th-cell polarization (Th1, Th2, Th17, and Treg), predictive for the specific type of immune response [11]. A matter of particular interest is the recent observation that depletion of FOXP3+ regulatory T cells promotes hypercholesterolemia and atherosclerosis in Ldlr-/- deficient mice [36]. As our animal models orchestrate several of these possible mechanisms simultaneously, it is difficult to address causality for the unexpected altered lipoprotein levels. This should be subject to more sophisticated adoptive transfer animal models. As we have demonstrated, the modulating effect of the C-Ldlr-/- Rag1-/- Il2rg-/- triple mutants on atherosclerotic cellularity with more macrophages and fewer SMCs was neither attributable to different proliferation and/or apoptosis nor to different adhesion capacities of the endothelium mediated by different VCAM-1 expression. Rather, this phenomenon argues for a more sophisticated view of the impact of NK cells on atherogenesis. Schulz et al. showed in their recent publication that NK cell cytotoxicity against autologous macrophages remains elevated if cultured with IL-10 [37]. The authors suggest that NK cells can delete macrophages and potentially other immature APC under inflammatory conditions that could be a possible mechanism of macrophage increase in triple KO atherosclerotic lesions. Compared to control and double KO mice (24 weeks on WTD), triple KO mice obviously developed atherosclerotic lesions with morphological features of fatty streak-like lesions. These types of early lesions do not develop fibrous caps and always have a very high presentation of macrophages. If so, lack of NK cells in triple KO mice leads to slowing down atherogenesis, development of fatty streak-like lesions, and as a result—increase of macrophage content in this experimental group consistent with the observation that lesion area of double and triple mutants is not vastly different (Fig 2A) as against macrophage cellularity (Fig 3A). Accordingly, NK cells have been shown to promote atherosclerotic lesion development in Ldlr-/- and Apoe-/- mice, respectively [38,39]. Whatever the precise mechanisms are, it seems unlikely that this modulating effect can be attributed to NK cells within the atherosclerotic lesions as we found only very limited numbers of NK and NKT cells in the Ldlr-/- control mice using two different antibodies (CD56, CD57). Rather, systemic effects such as the impact of the total NK and NKT cell pool on lipoprotein metabolism (see above) should be considered. It is now widely accepted that immune cells play an important role in atherogenesis. Collectively, the present study reveals a previously unappreciated complexity with regard to the impact of lymphocytes on lipoprotein metabolism on the one hand and the role of certain lymphocyte subsets in plaque composition on the other hand. This complexity may be attributed, at least in part, to the Th2-prone BALB/c background. BALB/c mice homozygous for targeted immunological mutations like Rag1-knockout, Rag2-knockout and the Il2rg-knockout are suitable recipients for human hematopoetic stem cells and therefore the study of atherosclerosis in humanized mice. Supporting Information S1 Table Summary statistics for body weights, serum cholesterol and triglyceride levels of all mouse groups. (XLSX) Click here for additional data file. S2 Table Quantification of the plaque areas of the aortic sinus of all mouse groups. (XLSX) Click here for additional data file. S3 Table Phenotypic analysis of atherosclerotic lesions in C-Ldlr-/-, C-Ldlr-/- Rag1-/- and C-Ldlr-/- Rag1-/- Il2rg-/- mice after 24 weeks on the western type diet. (XLSX) Click here for additional data file. S4 Table Quantification of cell proliferation and apoptosis in atherosclerotic lesions of C-Ldlr-/-, C-Ldlr-/- Rag1-/- and C-Ldlr-/- Rag1-/- Il2rg-/- mice as determined by immunohistochemistry. (XLSX) Click here for additional data file. S5 Table Quantification of NK cell distribution in atherosclerotic lesions of C-Ldlr-/- mice as determined by immunohistochemistry. (XLSX) Click here for additional data file. We thank Carolin Orning for expert technical assistance. ==== Refs References 1 Libby P , Lichtman AH , Hansson GK (2013 ) Immune effector mechanisms implicated in atherosclerosis: from mice to humans . Immunity 38 : 1092 –1104 . 10.1016/j.immuni.2013.06.009 23809160 2 Daugherty A , Pure E , Delfel-Butteiger D , Chen S , Leferovich J , Roselaar SE , et al (1997 ) The effects of total lymphocyte deficiency on the extent of atherosclerosis in apolipoprotein E-/- mice . J Clin Invest 100 : 1575 –1580 . 9294126 3 Dansky HM , Charlton SA , Harper MM , Smith JD (1997 ) T and B lymphocytes play a minor role in atherosclerotic plaque formation in the apolipoprotein E-deficient mouse . Proc Natl Acad Sci U S A 94 : 4642 –4646 . 9114044 4 Song L , Leung C , Schindler C (2001 ) Lymphocytes are important in early atherosclerosis . J Clin Invest 108 : 251 –259 . 11457878 5 Caligiuri G , Nicoletti A , Poirier B , Hansson GK (2002 ) Protective immunity against atherosclerosis carried by B cells of hypercholesterolemic mice . J Clin Invest 109 : 745 –753 . 11901183 6 Major AS , Fazio S , Linton MF (2002 ) B-lymphocyte deficiency increases atherosclerosis in LDL receptor-null mice . Arterioscler Thromb Vasc Biol 22 : 1892 –1898 . 12426221 7 Getz GS , Vanderlaan PA , Reardon CA (2011 ) Natural killer T cells in lipoprotein metabolism and atherosclerosis . Thromb Haemost 106 : 814 –819 . 10.1160/TH11-05-0336 21946866 8 Nakai Y , Iwabuchi K , Fujii S , Ishimori N , Dashtsoodol N , Watano K , et al (2004 ) Natural killer T cells accelerate atherogenesis in mice . Blood 104 : 2051 –2059 . 15113755 9 Aslanian AM , Chapman HA , Charo IF (2005 ) Transient role for CD1d-restricted natural killer T cells in the formation of atherosclerotic lesions . Arterioscler Thromb Vasc Biol 25 : 628 –632 . 15591216 10 VanderLaan PA , Reardon CA , Sagiv Y , Blachowicz L , Lukens J , Nissenbaum M , et al (2007 ) Characterization of the natural killer T-cell response in an adoptive transfer model of atherosclerosis . Am J Pathol 170 : 1100 –1107 . 17322392 11 Ait-Oufella H , Sage AP , Mallat Z , Tedgui A (2014 ) Adaptive (T and B cells) immunity and control by dendritic cells in atherosclerosis . Circ Res 114 : 1640 –1660 . 10.1161/CIRCRESAHA.114.302761 24812352 12 Zhou X , Paulsson G , Stemme S , Hansson GK (1998 ) Hypercholesterolemia is associated with a T helper (Th) 1/Th2 switch of the autoimmune response in atherosclerotic apo E-knockout mice . J Clin Invest 101 : 1717 –1725 . 9541503 13 Schulte S , Sukhova GK , Libby P (2008 ) Genetically programmed biases in Th1 and Th2 immune responses modulate atherogenesis . Am J Pathol 172 : 1500 –1508 . 10.2353/ajpath.2008.070776 18467709 14 Roselaar SE , Daugherty A (1998 ) Apolipoprotein E-deficient mice have impaired innate immune responses to Listeria monocytogenes in vivo . J Lipid Res 39 : 1740 –1743 . 9741685 15 Ishibashi S , Brown MS , Goldstein JL , Gerard RD , Hammer RE , Herz J (1993 ) Hypercholesterolemia in low density lipoprotein receptor knockout mice and its reversal by adenovirus-mediated gene delivery . J Clin Invest 92 : 883 –893 . 8349823 16 Piedrahita JA , Zhang SH , Hagaman JR , Oliver PM , Maeda N (1992 ) Generation of mice carrying a mutant apolipoprotein E gene inactivated by gene targeting in embryonic stem cells . Proc Natl Acad Sci U S A 89 : 4471 –4475 . 1584779 17 Zhang SH , Reddick RL , Piedrahita JA , Maeda N (1992 ) Spontaneous hypercholesterolemia and arterial lesions in mice lacking apolipoprotein E . Science 258 : 468 –471 . 1411543 18 Spencer MW , Mühlfeld AS , Segerer S , Hudkins KL , Kirk E , LeBoeuf RC , et al (2004 ) Hyperglycemia and hyperlipidemia act synergistically to induce renal disease in LDL receptor-deficient BALB mice . Am J Nephrol 24 : 20 –31 . 14671436 19 Mombaerts P , Iacomini J , Johnson RS , Herrup K , Tonegawa S , Papaioannou VE (1992 ) RAG-1-deficient mice have no mature B and T lymphocytes . Cell 68 : 869 –877 . 1547488 20 Cao X , Shores EW , Hu-Li J , Anver MR , Kelsall BL , Russell SM , et al (1995 ) Defective lymphoid development in mice lacking expression of the common cytokine receptor gamma chain . Immunity 2 : 223 –238 . 7697543 21 Petkov PM , Ding Y , Cassell MA , Zhang W , Wagner G , Sargent EE , et al (2004 ) An efficient SNP system for mouse genome scanning and elucidating strain relationships . Genome Res 14 : 1806 –1811 . 15342563 22 Koike T , Kitajima S , Yu Y , Li Y , Nishijima K , Liu E , et al (2009 ) Expression of human apoAII in transgenic rabbits leads to dyslipidemia: a new model for combined hyperlipidemia . Arterioscler Thromb Vasc Biol 29 : 2047 –2053 . 10.1161/ATVBAHA.109.190264 19778946 23 Fan J , Unoki H , Kojima N , Sun H , Shimoyamada H , Deng H , et al (2001 ) Overexpression of lipoprotein lipase in transgenic rabbits inhibits diet-induced hypercholesterolemia and atherosclerosis . J Biol Chem 276 : 40071 –40079 . 11477088 24 Paigen B , Morrow A , Holmes PA , Mitchell D , Williams RA (1987 ) Quantitative assessment of atherosclerotic lesions in mice . Atherosclerosis 68 : 231 –240 . 3426656 25 Lehr HA , Mankoff DA , Corwin D , Santeusanio G , Gown AM (1997 ) Application of photoshop-based image analysis to quantification of hormone receptor expression in breast cancer . J Histochem Cytochem 45 : 1559 –1565 . 9358857 26 Torzewski M , Ochsenhirt V , Kleschyov AL , Oelze M , Daiber A , Li H , et al (2007 ) Deficiency of glutathione peroxidase-1 accelerates the progression of atherosclerosis in apolipoprotein E-deficient mice . Arterioscler Thromb Vasc Biol 27 : 850 –857 . 17255533 27 Martin-Murphy BV , You Q , Wang H , De La Houssaye BA , Reilly TP , Friedman JE , et al (2014 ) Mice lacking natural killer T cells are more susceptible to metabolic alterations following high fat diet feeding . PLoS One 9 : e80949 10.1371/journal.pone.0080949 24465369 28 Miyagi T , Takehara T , Uemura A , Nishio K , Shimizu S , Kodama T , et al (2010 ) Absence of invariant natural killer T cells deteriorates liver inflammation and fibrosis in mice fed high-fat diet . J Gastroenterol 45 : 1247 –1254 . 10.1007/s00535-010-0272-y 20596733 29 Lai D , Zhu J , Wang T , Hu-Li J , Terabe M , Berzofsky JA , et al (2011 ) KLF13 sustains thymic memory-like CD8(+) T cells in BALB/c mice by regulating IL-4-generating invariant natural killer T cells . J Exp Med 208 : 1093 –1103 . 10.1084/jem.20101527 21482696 30 Lynch L , Nowak M , Varghese B , Clark J , Hogan AE , Toxavidis V , et al (2012 ) Adipose tissue invariant NKT cells protect against diet-induced obesity and metabolic disorder through regulatory cytokine production . Immunity 37 : 574 –587 . 10.1016/j.immuni.2012.06.016 22981538 31 De Sanctis JB , Blanca I , Bianco NE (1995 ) Expression of different lipoprotein receptors in natural killer cells and their effect on natural killer proliferative and cytotoxic activity . Immunology 86 : 399 –407 . 8550077 32 Juompan L , Fournie GJ , Benoist H (1994 ) LDL and acetyl-LDL inhibit the NK activity and are taken up by CD56+ lymphocytes . Biochim Biophys Acta 1224 : 1 –10 . 7524682 33 Maczek C , Recheis H , Böck G , Stulnig T , Jürgens G , Wick G (1996 ) Comparison of low density lipoprotein uptake by different human lymphocyte subsets: a new method using double-fluorescence staining . J Lipid Res 37 : 1363 –1371 . 8808771 34 Chen A , Geng Y , Ke H , Constant L , Yan Z , Pan Y , et al (2014 ) Cutting edge: Dexamethasone potentiates the responses of both regulatory T cells and B-1 cells to antigen immunization in the ApoE(-/-) mouse model of atherosclerosis . J Immunol 193 : 35 –39 . 10.4049/jimmunol.1302469 24899497 35 Fyfe AI , Qiao JH , Lusis AJ (1994 ) Immune-deficient mice develop typical atherosclerotic fatty streaks when fed an atherogenic diet . J Clin Invest 94 : 2516 –2520 . 7989611 36 Klingenberg R , Gerdes N , Badeau RM , Gisterå A , Strodthoff D , Ketelhuth DF , et al (2013 ) Depletion of FOXP3+ regulatory T cells promotes hypercholesterolemia and atherosclerosis . J Clin Invest 123 : 1323 –1334 . 10.1172/JCI63891 23426179 37 Schulz U , Kreutz M , Multhoff G , Stoelcker B , Kohler M , Andreesen R , et al (2010 ) Interleukin-10 promotes NK cell killing of autologous macrophages by stimulating expression of NKG2D ligands . Scand J Immunol 72 : 319 –331 . 10.1111/j.1365-3083.2010.02435.x 20883317 38 Whitman SC , Rateri DL , Szilvassy SJ , Yokoyama W , Daugherty A (2004 ) Depletion of natural killer cell function decreases atherosclerosis in low-density lipoprotein receptor null mice . Arterioscler Thromb Vasc Biol 24 : 1049 –1054 . 14988092 39 Selathurai A , Deswaerte V , Kanellakis P , Tipping P , Toh BH , Bobik A , et al (2014 ) Natural killer (NK) cells augment atherosclerosis by cytotoxic-dependent mechanisms . Cardiovasc Res 102 : 128 –137 . 10.1093/cvr/cvu016 24469537

PMC005xxxxxx/PMC5001716.txt

==== Front PLoS Comput BiolPLoS Comput. BiolplosploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, CA USA 2756372010.1371/journal.pcbi.1005085PCOMPBIOL-D-16-00124Research ArticleBiology and Life SciencesBiochemistryMetabolismMetabolitesComputer and Information SciencesNetwork AnalysisMetabolic NetworksBiology and Life SciencesOrganismsBacteriaPseudomonasPseudomonas PutidaMedicine and Health SciencesPharmacologyPharmacokineticsDrug MetabolismPhysical SciencesChemistryChemical ReactionsOxidation-Reduction ReactionsPhysical SciencesChemistryElectrochemistryOxidation-Reduction ReactionsPhysical SciencesChemistryStoichiometryBiology and Life SciencesCell BiologyCell PhysiologyCell MetabolismBiology and Life SciencesBiochemistryMetabolismMetabolic PathwaysEfficient Reconstruction of Predictive Consensus Metabolic Network Models Consensus Metabolic Network Modelsvan Heck Ruben G. A. 12Ganter Mathias 1Martins dos Santos Vitor A. P. 23*Stelling Joerg 1*1 Department of Biosystems Science and Engineering and Swiss Institute of Bioinformatics, ETH Zurich, Basel, Switzerland2 Laboratory of Systems and Synthetic Biology, Wageningen University, Wageningen, The Netherlands3 LifeGlimmer GmbH, Berlin, GermanyReed Jennifer L. EditorUniversity of Wisconsin-Madison, UNITED STATESThe authors have declared that no competing interests exist. Conceived and designed the experiments: MG JS VAPMdS. Performed the experiments: RGAvH MG. Analyzed the data: RGAvH MG JS VAPMdS. Contributed reagents/materials/analysis tools: RGAvH MG. Wrote the paper: RGAvH MG JS VAPMdS. * E-mail: vitor.martinsdossantos@wur.nl (VAPMdS); joerg.stelling@bsse.ethz.ch (JS)26 8 2016 8 2016 12 8 e100508526 1 2016 29 7 2016 © 2016 van Heck et al2016van Heck et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Understanding cellular function requires accurate, comprehensive representations of metabolism. Genome-scale, constraint-based metabolic models (GSMs) provide such representations, but their usability is often hampered by inconsistencies at various levels, in particular for concurrent models. COMMGEN, our tool for COnsensus Metabolic Model GENeration, automatically identifies inconsistencies between concurrent models and semi-automatically resolves them, thereby contributing to consolidate knowledge of metabolic function. Tests of COMMGEN for four organisms showed that automatically generated consensus models were predictive and that they substantially increased coherence of knowledge representation. COMMGEN ought to be particularly useful for complex scenarios in which manual curation does not scale, such as for eukaryotic organisms, microbial communities, and host-pathogen interactions. Author Summary Many large-scale mathematical models describe metabolism to understand how microbes and other organisms (including humans) function and interact with each other and with their environment. Making these models is extremely time- and effort-intensive; it requires gathering and combining information from many sources, including the organism’s genome sequence, biological databases, scientific literature, and expert advice. The exact procedure and resources used depend on the model creators’ expertise and research interests, such that independently created models for the same organism are often very different and can hardly be compared. However, each model typically contains unique information that is ‘lost’ when working with a different model. To integrate the available knowledge, we developed a computational tool to build consensus metabolic models. Our tool—COMMGEN- combines independently generated models by matching identical parts and resolving differences between inconsistent parts. We apply our tool to four sets of models of different organisms. In all these sets, COMMGEN identified and resolved hundreds of inconsistencies. COMMGEN can be generally applied to standardize and improve models of metabolism, in particular for complex scenarios, such as those involving microbial communities and host-pathogen interactions. http://dx.doi.org/10.13039/501100001711Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen ForschungSystemsX.ch RTD MetaNetXStelling Joerg Sixth Framework Programme (BE)SystemTBMartins dos Santos Vitor A. P. http://dx.doi.org/10.13039/501100004965Sixth Framework ProgrammeINFECTMartins dos Santos Vitor A. P. We gratefully acknowledge financial support from the Swiss Initiative for Systems Biology (SystemsX.ch, project MetaNetX) reviewed by the Swiss National Science Foundation (SNF), the Wageningen university IPOP project, and the European projects INFECT (Project reference: 305340) and EmPowerPutida (Project reference: 635536). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Genome-scale constraint-based metabolic models (GSMs) are curated organism-specific knowledge repositories [1]. They integrate many distinct (bio)chemical entities and typically account for thousands of metabolites, reactions and genes. When assuming that metabolism is in a steady state, GSMs also enable metabolic simulations with applications in genome annotation [2,3], analysis of omics data [4–6], phenotype predictions [7–9], organism comparison [9–12], drug discovery [7,13,14], and metabolic engineering [8,15]. GSMs thereby quantitatively reconstruct the internal metabolic and transport wiring of the modeled organism and thus increase our systems level understanding. Genome-scale metabolic reconstructions consist of metabolites, metabolic reactions (including boundary reactions and a biomass reaction), cellular compartments, and genes [1,16]. The reactions are organized according to the cellular compartments in which they are active. Enzyme-driven (as opposed to spontaneous) reactions are associated with Gene-protein-reaction rules (GPR), which include one or more genes. For multiple genes, the GPR indicates whether alternative isozymes or enzyme complexes catalyze the reaction [17]. A reaction’s equation consists of substrates and products with their corresponding stoichiometries. A reaction’s reversibility describes whether the reaction operates forward, backward, or bi-directionally. The reaction flux bounds specify the reaction’s capacity, that is, the absolute upper and lower bounds of the reaction flux. Transport reactions transfer metabolites between cellular compartments, whereas boundary reactions define nutrient uptake and secretion. The biomass reaction, finally, reflects the molecular composition of a cell or organism and represents cell or organism growth. Together, these entities and their encoding in a GSM aim to represent the current knowledge of the organism’s metabolism. However, even for well-studied organisms such as Saccharomyces cerevisiae or Bacillus subtilis, many uncertainties remain during GSM construction. These uncertainties are typically manually addressed based on expert knowledge and scientific literature, which involves a laborious iterative process that can take several years, for example, for eukaryotes [1]. The main sources of uncertainties are: (i) incomplete and erroneous information from heterogeneous and potentially contradictory data sources such as insufficiently curated and inconsistent gene annotations [18], alternative naming and spelling variants of metabolites (different namespaces) [18–21], and conflicting reaction reversibilities [2,22]; (ii) subjectivity in interpreting literature sources; (iii) integration of qualitative and quantitative data (e.g., inconsistent growth data); and (iv) incompatible levels of detail between and among (reference) databases; for example, databases may represent metabolic pathways by detailed individual reactions or by a single lumped reaction [18], and they may use varying structural definitions for metabolite classes such as lipids and polymers [21,23]. As a consequence, when several GSMs for the same organism are developed independently, they are complementary and only partially overlapping [24,25]. The extent of variation between models for the same organism can be dramatic. For example, the well-established human and yeast GSMs agree only on 3% [18] and 35% [20] of their reactions, respectively, when ignoring electron, proton, and water imbalances. Differences between GSMs resulting from different modeling frameworks and model authors can even be more substantial than biological differences between organisms [26]. Any GSM-driven analysis, which needs to (somewhat arbitrarily) select one GSM when several are available, thus, only operates on a subset of the available information. To represent metabolism more comprehensively, and thereby improve our understanding of a target organism, alternative GSMs of a target organism can be integrated into a so-called consensus model of the respective organism, one per organism. Consensus models have an increased scope (by combining unique parts of initial GSMs) and they are more consolidated (by identifying shared parts of initial GSMs that are likely to be reliable). When discrepancies exist between GSMs, these must be carefully examined to select the most appropriate modeling alternative. However, while consensus models have been generated successfully for several (model) organisms such as budding yeast and human, this required extensive manual curation by communities of domain experts [10,20,24,25,27]. To alleviate this bottleneck and render GSMs truly useful for the understanding of cellular function and evolution, community function, and host-pathogen interactions, semi-automatic consensus model generation approaches have been proposed. It has been shown that the combination of complementary GSMs of the same organism reduces existing gaps in individually reconstructed GSMs [28,29]. These approaches focused mainly on reconciling namespaces (a particularly important challenge for matching metabolites) or on curating the underlying databases [18,21]. Thereby, existing methods address only a small subset of the problems in consensus model generation described above. For example, they do not identify and curate cases when two initial GSMs represent the same metabolic process at different levels of granularity [30]. Here, we present COMMGEN, a tool for COnsensus Metabolic Model GENeration that reconciles two or more distinct GSMs of the same organism beyond a common namespace. COMMGEN automatically identifies similarities, dissimilarities, and complements of the metabolic networks based on an extensive classification of problems that typically arise during GSM integration and on novel algorithms to resolve these problem classes. For several model organisms, we show that semi-automatically created consensus GSMs in a standardized namespace [31] are substantially more consolidated than achievable by a common namespace alone, and that they retain or even improve on the initial GSMs’ predictive capabilities. Because the consensus GSMs contain the information from each initial GSM, they comprehensively represent our best understanding of the organisms’ metabolic networks. Results Our analyses addressed model building, testing and refinement in a stepwise fashion. We started by identifying the classes of inconsistencies that exist between models for four widely different albeit representative microbes. We subsequently set up the framework for COnsensus Metabolic Model GENeration, and tested it on the four case studies for functionality and predictability. Inconsistency classes arising in model merging To systematically resolve inconsistencies between two or more Initial GSMs (IGSMs) to be integrated, we defined three main (coupled) inconsistency categories: metabolites, reactions, and compartments. We explain these categories and the inconsistency classes they contain using examples from four sets of IGSMs that cover gram-positive and gram-negative bacteria as well as yeast (Fig 1a). 10.1371/journal.pcbi.1005085.g001Fig 1 Models used in this study and classification of inconsistencies. (a) Overview of the used initial GSMs. (b) Instances of identical metabolites with different MnXRef identifiers. (c) Non-identical metabolites that perform identical functions in the network context. (d) Alternative modeling of polymers. (e) Nested and encompassing reactions. (f) Alternative usage of redox pairs. (g) Alternative reactions with consequences for redox metabolism. (h) Partially overlapping reactions differing in phosphate products. (i) Lumped vs. non-lumped representation of a pathway. (j) Invalid transport reaction (IR08663). (k) Alternative transport reactions for putrescine. (l) Alternative transport reactions for glycine. (m) Invalid boundary reaction (R841). Circles represent chemical species, arrows chemical reactions, and grey boxes different compartments. Red nodes indicate instances of identical species within the network context whose alternative names are separated by horizontal lines. Rectangular boxes contain the original reaction names, rounded rectangles their corresponding GPRs, where '&' represents a logical AND, and '|' a logical OR. Edges with filled circles represent reversible reactions. Stoichiometric coefficients unequal to one are indicated at their respective arrows. The shown reactions originate from GSMs of four different organisms: B. subtilis (d), as represented in iYO844 [3] (blue) and iBSu1103 [36] (orange); M. tuberculosis (m), as represented in iNJ661 [14] (blue) and GSMN_TB [7] (orange);P. putida (b,c,e,f,h,I,j,k), as represented in iJN746 [33] (blue) and iJP962 [10] (orange); and S. cerevisiae (g,l), as represented in iIN800 [48] (blue) and iMM904 [18,37] (orange) and iND750 [49] (pink). Metabolites IGMs often represent a specific chemical compound differently because metabolite identifiers are ambiguous and they reside in different namespaces [31]. When one simply merges IGSMs, that is, adds the IGSMs’ contents, this leads to redundant pathways (Fig 1b) that may differ in metabolites, gene associations, stoichiometries, and reversibilities. The essential step of identifying and merging different metabolites that represent the same chemical compound in different namespaces has been emphasized previously [29–31]. However, more complicated situations exist when different metabolites actually represent different chemical compounds, but these compounds have the same function in their network context. This typically arises when metabolites are modeled at different granularity, for example, as ‘iron’ and ‘Fe2+’, or ‘glucose’ and ‘alpha-D-glucose’. Common metabolites may also have different chemical sum formulas in different IGSMs, for example, depending on whether functional groups are specified or not (Fig 1c), or when polymers are modeled with a different numbers of subunits (Fig 1d). In such cases, the merging of metabolites has to prevent stoichiometric inconsistencies in the consensus model: if a merged polymer can be produced from fewer subunits than result from its degradation, mass conservation is violated. Hence, a common namespace is not sufficient to identify common metabolites in IGSMs. Reactions A particular biological process is often represented differently in two models because of uncertainties, disagreements, errors, and modeling decisions, resulting in alternative representations of a single reaction or of reaction sets. These alternatives need to be identified and matched to avoid reaction redundancies (Fig 1b) and violations of mass balances due to inconsistent stoichiometries (Fig 1c and 1d). However, inconsistencies may extend beyond namespaces and stoichiometries. They often result from modeling decisions, both in capturing individual reactions, and in the granularity of representation for metabolic processes. Nested reactions, where one reaction is a perfect subset of another reaction with respect to metabolites, are possible consequences. In the example in Fig 1e, the cofactor NADH may be used, but it is not required—for a consensus model, a decision between these alternatives eventually has to be made. Alternative modeling decisions on cofactor usage are common in IGSMs as shown in Fig 1f with a ‘choice’ between using NADH and NADPH and in Fig 1g, where the same chemical conversion can either yield NADP from NADPH or NADPH from NADP. More complex cases to resolve are partially overlapping reactions and lumped reactions, where multiple reactions are artificially represented by fewer reactions. Fig 1h shows an example of two alternative reactions that generate triphosphate or pyrophosphate and monophosphate, respectively; simply merging the two IGSMs would feed the side-products into different pathways because no reaction exists that interconverts these metabolites directly. Such inconsistencies are not only found between IGSMs, where they are expected, but also within IGSMs, as demonstrated in Fig 1i. Hence, it is important to consider the network context of the IGSMs and of the merged GSM. Compartments IGSMs of the same organism may consider different subcellular compartments (Fig 1a), affecting the localization and multiplicity of reactions as well as the incorporated transport reactions. For example, in Fig 1j, the two IGSMs for a gram-negative bacterium have the same net reaction for the import of cysteine into the cytoplasm. In one IGSM this requires one reaction because the periplasm is not explicitly modeled, whereas the more detailed transport in the other IGSM requires two reactions. After identifying this class of inconsistencies, a consensus model can either replace the transporter connecting the extracellular space with the cytoplasm by two reactions, or remove the entire periplasm and retain a single transport reaction. Because transporters and transport reactions are notoriously difficult to identify and characterize [1], IGSMs are often inconsistent in transport reactions. Fig 1k shows an extreme example: a single merging artifact effectively destroys the model of the proton gradient because protons can be transported across the membrane in either direction by simultaneous import and export of putrescine. Inconsistencies in transport reactions can also lead to thermodynamically infeasible cycles [1] such as ATP generation resulting from cycling glycine over the membrane (Fig 1l). Finally, boundary reactions, which are not mass-balanced because they exchange material with the environment, are sometimes lumped with transport reactions for the same chemical compound and thus first require standardization (Fig 1m). Overall, therefore, a broad spectrum of unrelated but interconnected inconsistencies at the metabolite, reaction, and compartment levels need to be identified and resolved for consensus model generation. The COMMGEN framework COMMGEN is a software tool that is designed to address the above problems in consensus model generation, leading to a semi-automatic reconciliation of two or more GSMs for a given organism. In terms of software architecture, COMMGEN operates on GSMs in SBML format [32], the standard modeling language for systems biology (Fig 2a). The IGSMs are first converted into a common chemical naming system using the MnXRef namespace [31]. Next, COMMGEN combines all reactions of the IGSMs into a Basic Consensus Model (BCM). The BCM is used to identify and reconcile inconsistencies between and within the IGSMs, ultimately yielding a Refined Consensus Model (RCM) in SBML format. Because many inconsistencies are interconnected, it is difficult to identify a consensus between IGSMs, to distinguish between conflicting and complementary model parts, and to resolve all inconsistencies automatically. COMMGEN therefore resolves all unambiguous cases automatically, and it guides the user to decide on the remaining cases. COMMGEN records all changes such that the user can automatically repeat the procedure with minimal effort, including manual alterations of previously made choices. 10.1371/journal.pcbi.1005085.g002Fig 2 COMMGEN framework. (a,b) Overview of COMMGEN workflow and available methods. The COMMGEN methods are either fully automatic (+), conditionally or optionally automatic (+/-), or they always require manual intervention (-). (c) Performance of the metabolite matching methods if run without manual intervention, leading to ROC-curves of the classification of metabolites as identical or non-identical based on their network context. Lines correspond to different fractions of the network information being randomly discarded: black, 0%; red, 30%; green, 60%; blue, 90%. The shades indicate the standard deviations in the classification. The data presented here was obtained using the Pseudomonas putida GSMs iJP962 [10] and iJN746 [33]; analysis results for the other sets of GSMs and additional information can be found in S5 Protocol. To identify and address all the different inconsistency classes described above, COMMGEN iteratively applies a set of independent methods (Fig 2b). All methods automatically identify instances of their respective inconsistency classes. Metabolite matching is a core element of model merging. We developed a novel algorithm to identify sets of metabolites that represent the same chemical compound based on their network context, that is, their neighboring metabolites and reactions, thereby addressing the issue of different granularity in IGSMs for metabolites (see Methods for details). Performance tests for P. putida networks revealed very high sensitivity and specificity of the algorithm, even when only a minority of the network is used to infer matching metabolite sets (Fig 2c). Metabolite matching allows COMMGEN subsequently to reconcile the associated reactions: metabolites are merged, through which novel pathways and branching points can be formed, and alternative representations of biochemical reactions become apparent. Specifically, COMMGEN matches sets of reactions in the following categories (see Methods for the respective algorithms): (i) reactions with identical metabolites but different stoichiometries; (ii) nested reactions; (iii) reactions that differ only in redox pairs; (iv) partially overlapping reactions; and (v) lumped reactions. Furthermore, it deals with differences in subcellular compartmentalization by (i) facilitating the removal of transporters; (ii) enabling the removal of entire compartments; (iii) resolving differences in the modeling of boundary reactions; (iv) identifying different transport reactions for the same metabolite across the same membrane; and (v) identifying identical biochemical conversions in different compartments. COMMGEN’s methods differ in the extent to which identified inconsistencies can be resolved automatically (Fig 2b). For some categories, the user can choose to automatically handle inconsistencies, for example, to deal with differences in reaction directionality. Conditionally automatic refers to inconsistency classes where some instances can be addressed automatically, but others cannot: if two matched reactions differ only in stoichiometric coefficients, COMMGEN can automatically select the elementally balanced reaction, but only when exactly one reaction is balanced. Manual intervention is always possible, and it is required when inconsistencies are too complex and diverse for a well-performing heuristic for automation. Manual curation is also advisable when an erroneous choice may substantially impact model performance. For example, a single incorrect match between two metabolites with different chemical sum formulas can have severe consequences for the correctness of model predictions. Hence, although the COMMGEN method for network-based metabolite matching performs extremely well (Fig 2c), we recommend manual confirmation of predicted matches. Model generation with COMMGEN: Case study for P. putida To describe COMMGEN operation in detail and to evaluate the framework’s performance, we focus on consensus model generation for Pseudomonas putida, for which the two GSMs iJP962 [8,10] and iJN746 [33] have been developed independently (Fig 1a). The initial overlap between these two models is surprisingly low: they only have 58% of their genes, 33% of their metabolites and 2% of their reactions in common. Conversion into the MnXRef namespace [31] only increases the common part to 44% for metabolites and 11% for reactions. To quantitatively determine the occurrences of inconsistencies and their resolution, we classify reactions as consensus reactions (shared between the GSMs) and unique reactions. We further categorize unique reactions according to whether they are unrelated to any inconsistency, related to a single inconsistency, or related to multiple inconsistencies (a reaction may appear in the last category because COMMGEN methods are not mutually exclusive in the inconsistencies they identify). Because the identified inconsistencies ultimately depend on namespace consistency, user-defined settings, and user choices, we quantified the resolution of inconsistencies by automatic processing to remove user bias as much as possible. After creating the BCM from the IGSMs and merging the identical reactions, the fraction of consensus reactions was low (11%) and approximately half of the unique reactions were associated with at least one inconsistency (Fig 3a; S1 Protocol). The inconsistencies exemplified in Fig 1 are, thus, not isolated cases; they merely illustrate the main problems in consensus model generation. 10.1371/journal.pcbi.1005085.g003Fig 3 Application of COMMGEN to P. putida GSMs. (a) Automatic inconsistency identification and reconciliation substantially increases consensus and reduces inconsistencies. Reactions are classified into consensus reactions (green) and unique reactions involving no (blue), a single (orange), or multiple (red) inconsistencies. (b, c) Characteristics of the refined consensus model as in (a) without network-based metabolite matching (b), or after manually addressing the remaining inconsistencies (c). (d) Numbers of reversible (‘+’) and irreversible (‘-‘) reactions in the RCM, grouped by the four possible combinations of reversibilities in the IGSMs. (e) Numbers of active and inactive reactions in the RCM, grouped by being active (‘+’) or inactive (‘-‘) in the IGSMs. Next, we employed a four-step automatic process to reconcile inconsistencies between the IGSMs and to converge to an automatically generated RCM (Fig 3a). First, COMMGEN increased the namespace consistency through our network context-based metabolite matching method (note that we manually confirmed the proposed matches such that subsequently identified inconsistencies were not overestimated). This increased the overlap to 53% for metabolites and 16% for reactions. In the second step, COMMGEN addressed the difference in cellular compartments in the P. putida GSMs (Fig 1a). In particular, transport reactions from iJP962 that immediately take up metabolites from the extracellular space into the cytoplasm were split such that they match the transport processes from iJN746, and periplasmic instances of the involved metabolites were added. Next, COMMGEN identified and merged sets of reactions with practically (ignoring protons and water) identical net formula. These sets include reactions that have different GPR rules or different reaction directionalities, or that did not have identical net formulas prior to the splitting of transport reactions or the COMMGEN-based metabolite matching. In this step, we processed inconsistent reaction reversibilities using our previously published method to predict reaction directionalities based on metabolite patterns [2], and we processed inconsistent gene associations by combining the GPR rules with a ‘strict’ heuristic (see S2 Protocol). Finally, COMMGEN identified and merged reactions that involve the same metabolites, but differ in stoichiometric coefficients; directionality and GPR inconsistencies were handled as above. The detailed data shown in Fig 3a emphasize the interdependencies of inconsistencies that may arise in model merging, in particular, that resolving inconsistencies may facilitate subsequent identification of more inconsistencies, resulting in an increased number of identified inconsistent reactions. The four automated steps increased the share of reactions that are consensus reactions originating from both IGSMs from 11% (in the BCM) to 39% (in the RCM), while also substantially reducing the number of reactions associated with inconsistencies (Fig 3a). We evaluated the significance of the metabolite matching step by re-running the process without it, which lead to only 23% consensus reactions (Fig 3b). In addition, we used the automatically generated RCM as the starting point for manual curation guided by COMMGEN methods. This allowed us to reconcile most of the remaining inconsistencies and to obtain a consensus for 50% of the reactions (Fig 3c). In summary, our detailed case study for P. putida therefore provides evidence for the efficiency of the COMMGEN framework, and in particular of its novel methods such as network context-based metabolite matching. Automatically generated consensus models are functional and predictive We next asked, to what extent automated consensus model generation preserved or even extended functionality of the IGSMs, initially focusing on the P. putida models. Our automated method involved the probabilistic prediction of reaction directionalities [2] to resolve reaction inconsistencies, instead of simply setting all reactions with conflicting directionalities to reversible, which would tend to overestimate the organism’s metabolic capabilities. It maintained reaction directions in case of consensus between the IGSMs, although the prediction method is agnostic to matches between models; it constrained directions in many cases when such constraints existed in only one IGSM (Fig 3d). The benefits of this approach are best exemplified with a concrete example (Fig 4a). The P. putida BCM contains a small set of reactions that together allow for non-physiological CO2 fixation. This incorrect CO2 fixation cycle was automatically removed when inconsistent directionalities of a reaction present in both IGSMs were processed, thereby preventing a major error in the RCM. Note that direction prediction also identified a reaction assigned with a direction that is not consistent with the remainder of the network (see also Fig 1i), namely a directed lumped reaction common to both IGSMs, and a bidirectional non-lumped reaction set present in only one model. Another important aspect of model consolidation is the extent to which active reactions in the IGSMs (that is, reactions that can carry metabolic flux in principle) are preserved. As shown in Fig 3e, essentially all active reactions in one of the networks remained active in the RCM, and only reactions that were non-functional in both IGSMs remained inactive. In growth phenotype predictions, the RCM occasionally disagreed with all IGSMs, suggesting ‘new’ metabolic functions. For example, while neither of the IGSMs captured that P. putida can grow on L-quinate as sole carbon source, complementation of reactions in the RCM enabled a biologically consistent model behavior (Fig 4b). These aspects together indicate overall functionality of the automatically generated consensus model. 10.1371/journal.pcbi.1005085.g004Fig 4 Subnetwork analysis for P. putida. (a) Example error of ‘naïve’ iGSM merging where the initial P. putida BCM contains a biologically inaccurate carbon dioxide fixation cycle due to incorrect directionalities in the IGSMs. This error is automatically resolved as COMMGEN assigns reaction directionalities opposite to those shown with dashed reaction arrows. (b) Example for a new metabolic function in the consensus model. P. putida can grow on L-quinate as its sole carbon source. Neither of the initial models captures this behavior, whereas the consensus model provides the necessary, complementary reactions. The performance of GSMs as mathematical models for cellular metabolism is typically evaluated by assessing their ability to correctly predict wild type and mutant growth phenotypes across different growth conditions [34]. We performed corresponding simulations for automatically refined consensus models as well as for their ancestors (IGSMs and BCM) for each of the four evaluated organisms (Fig 1a). Specifically, we computed sensitivity, specificity, accuracy, and Matthew’s correlation coefficient (MCC; unlike accuracy it takes the total numbers of true and false test cases into account) [35] for growth phenotype predictions (see S3 Protocol for details). Fig 5a shows the performance indicators for the IGSMs, the BCMs, and the automatically refined consensus models for each organism. In nearly all metrics, the IGSMs outperformed the BCM (except for P. putida), and they were outperformed by the RCM (except for B. subtilis). For B. subtilis, resolving inconsistencies in the BCM decreased all scores except sensitivity. This can be explained by one IGSM (iBSu1103) being largely based on a predecessor (iYO844); in addition, iBSu1103 was optimized for correct growth predictions using GrowMatch [34,36]. Information from iYO844 can thus include errors that were deliberately removed from iBSu1103 and it can reverse changes made by the performance optimization. Thus, although the prediction profiles of the RCMs largely resemble the IGSM profiles, RCMs on average outperform both the IGSMs and the BCMs, indicating efficiency of the automated consensus model generation methods in COMMGEN even in terms of prediction capabilities. Notably, user choices of the biomass reaction do not influence the performance substantially (Fig 5a), pointing to robustness of the methods as well. 10.1371/journal.pcbi.1005085.g005Fig 5 Performance evaluation of COMMGEN. (a) Evaluation of GSM ability to predict growth phenotypes. Predictive ability of initial GSMs (blue), basic consensus models (red), and automatically created refined consensus model (green) according to the metrics defined in the text. The test data comprised gene knockout data (B. subtilis [3,36], P. putida [8,50], M. tuberculosis [51], S. cerevisiae [49]), biolog data (B. subtilis [3,36], P. putida [8,33]) and auxotrophies (P. putida [50]). See S3 Protocol for details. (b,c) Comparison of manual (yeast consensus model [20] based on the IGSMs iMM904 [37] and iLL672 [38]) and automatic consensus model generation with namespace matching only, or with COMMGEN. (b) Numbers of common reactions and metabolites for manual curation, name space conversion, and automatically created refined consensus model. (c) Incidences of inconsistent reaction classes identified by COMMGEN. Automatic reconciliation is comparable to manual consensus model generation Finally, we wanted to evaluate how automatic consensus model generation compares to its (largely) manual counterpart. We focused on the community approach to establish a yeast consensus model [20] based on the IGSMs iMM904 [37] and iLL672 [38] because this first model reconciliation effort is especially well documented. Fig 5b shows that transfer of the IGSMs into a standardized namespace alone identifies only small subsets of common metabolites and reactions. COMMGEN’s automated reconciliation method, in contrast, achieves nearly the same extent of matching between the IGSMs as reported for the manual curation. The automatically generated RCM showed good performance in mutant phenotype predictions (sensitivity = 0.98, specificity = 0.28, accuracy = 0.87 and MCC = 0.42; note that a comparison to the manual consensus model is impossible because the community effort did not aim at establishing a model suitable for FBA). In addition, COMMGEN directly identifies many inconsistencies between model reactions that result, for example, from different numbers of compartments in the IGSMs (Fig 5c). These would be clear starting points for domain experts for subsequent COMMGEN-assisted manual curation. We believe that the combination of automated procedures with close-to-manual quality and of support for targeted manual curations would substantially enhance future community efforts. Discussion Genome-scale constraint-based metabolic models are both integrated knowledge repositories and predictive mathematical models. In terms of knowledge representation, a consensus model should be more consolidated than individual GSMs due to shared parts, more comprehensive due to unique parts, and more accurate due to reconciliation of inconsistencies in similar parts. A consensus model, however, can propagate errors in the initial models’ unique parts, and it may be less consistent than the initial models, especially when inconsistencies in similar model parts were not identified or reconciled. Inconsistencies in GSMs are typically nested, not mutually exclusive, and therefore difficult to address, which so far prevented the development of methods for the automated generation of consensus models [30]. Manual network reconciliation, the predominant approach applied today, is difficult and cumbersome because the number of inconsistencies between just two or three IGSMs already runs in the thousands. Based on a systematic classification of inconsistencies, COMMGEN automatically identifies and semi-automatically reconciles inconsistencies between and within two or more IGSMs. The inconsistencies could theoretically be reconciled fully automatically, but automated resolution depends on the used reference databases, which vary to a large extent [18]. Therefore, COMMGEN does not entirely remove the need for manual inspection and curation. For example, our framework relies on network similarity between alternative realizations of metabolites and reactions in order to match them. Because the reactions surrounding biomass formation are often implemented very differently in different GSMs, they are not matched. While our implementation lets the user choose one of the IGSM biomass reactions, a manual update seems necessary as long as COMMGEN does not automatically fetch external information that would enable an automatic reconciliation of the biomass reaction. In addition, there exists a trade-off between sensitivity and specificity for the identification of inconsistent reactions, which limits the detection of lumped and non-lumped pathway representations with a different net reaction. Also, the identification of similar or identical reactions in different cellular compartments is difficult to achieve automatically (but an extension of the current framework could progress in this direction by combining the information from metabolite instances in different compartments prior to metabolite matching). COMMGEN thus forms a necessary bridge between full automation and high-quality manual curation for consensus metabolic model generation. Regarding a GSM’s predictive mathematical model character, it is important to note that remaining inconsistencies in a consensus model can have severe effects, for example, when inconsistencies resulting from model merging are not adequately addressed. As a consequence, individual GSMs may outperform a consensus model in terms of predictive ability even though the latter is more representative of the available information. COMMGEN’s aim (and design) is to compare and reconcile IGSMs in order to obtain a high-quality representation of the IGSMs’ combined information. In contrast to model optimization methods such as GrowMatch [34], COMMGEN does not create a model optimized for predictive ability, and it does not use corresponding experimental information. However, our example applications also demonstrated that automatically generated consensus models almost always have higher predictive power than the manually curated IGSMs and that these models can be comparable to manually constructed consensus models as shown for yeast. COMMGEN increases coherence with the actual biological system while maintaining predictive power. This balance is of utmost importance for the usability and reliability of GSMs to elucidate cell function interactions. As demonstrated by our case study for P. putida, we argue that (semi-) automatically generated consensus models provide the basis for additional improvements due to their comprehensiveness and standardized naming system. Gap-filling methods [2,39] may be able to close gaps that are not apparent in the IGSMs. One can use existing methods [2,40] to re-evaluate reaction directionalities, especially for reactions that differed in the IGSMs. Compartment assignment methods [41] can resolve remaining compartmentalization issues and optimization methods [34,42] may alter the model to increase its predictive ability. Finally, a good consensus model is a solid foundation for new models by providing a basis for GSMs of similar organisms, and via its integration into multi-scale whole-cell or tissue models [36]. More generally, the systematic integration of heterogeneous information is an essentially unsolved challenge in (post-)genomic biology. For metabolism, consensus GSMs are formalized means for complementing incomplete information, and for identifying and addressing errors through the comparison of independently generated GSMs for the same organism. COMMGEN automatically identifies and semi-automatically resolves widespread and highly interlinked inconsistencies between initial GSMs, thereby moving beyond existing approaches for manual and computer-aided consensus model generation. It can therefore facilitate the construction of new models by comparing and combining information from automatic model construction tools such as the modelSEED [43] and manual model construction efforts, and facilitate GSM updates using a reference—both tasks are analogous to consensus GSM generation. While we focus here on the reconciliation of multiple GSMs for the same species, we argue that COMMGEN’s methods and standardization are more widely applicable. The identification of similar, yet distinct, biochemical entities can help to compare metabolic capabilities of organisms in detail via their GSMs, or even to compare entire pathway databases. However, dealing with different species will require new, systematic preprocessing steps to map gene sets in different organisms functionally to each other (e.g., via orthology or enzyme classification numbers), which is a topic of future research. In addition, COMMGEN’s methods for identifying redundancies and hierarchical relationships in networks can be used to further advance standardization of terms and ontologies. We therefore expect COMMGEN to be of substantial aid in future integration of knowledge for metabolic networks, to greatly accelerate model-building processes and to thereby improve subsequent high-throughput model-based network analyses. Although COMMGEN will not directly address the domain-specific problems, these capabilities will lay a solid foundation for the systematic, genome-scale comparison of metabolic spaces within and across genera and will have substantial impact for large-scale evolutionary analyses, design of microbial communities, and understanding of host-microbe (pathogen, microbiome) interactions. Methods Genome-scale metabolic models iJN746 and iJP962 were requested from and received by email from the first authors of the corresponding papers. GSMN-TB was downloaded from http://sysbio3.fhms.surrey.ac.uk/. iNJ661 was obtained from the supplementary files of the corresponding paper. The remaining models were taken from the model repository at www.metanetx.org. See S1 Dataset for details. Evaluation of model performance For comparison to experimental data, the models were loaded into the COBRA toolbox [44]. The bounds of the boundary reactions were adjusted based on the medium composition and, where necessary, additional flexibility was provided to individual models. Gene knockout strains were simulated by removing the reactions requiring the encoded protein. To discriminate growth from no growth for wild type strains a default cut-off value (10−6) was used whereas a minimal relative growth rate (30%) to the wild type was used for mutant strains. See S3 Protocol for details. Matching metabolites based on network context In a metabolic network, reaction nodes are only connected to the metabolite and gene nodes that are involved in the corresponding reaction. Similarly, metabolite and gene nodes are only connected to reaction nodes. However, reaction nodes are not informative for the identity of metabolites as two metabolites representing the same chemical compound are non-overlapping in their connected reaction nodes. Therefore, we characterize metabolites by the other metabolite and gene nodes that are connected to the same reactions. We use this information to quantify how similar metabolites from different models are based on their network context. These similarity scores are then compared to the scores of metabolites that are known to match because they are present in both models: pairs of metabolites that score comparable to these shared metabolites may consist of functionally equivalent chemical compounds. We use a user-defined percentile of shared metabolite scores as a threshold to identify similar metabolites. The method is described in the following: We create a Boolean metabolite-to-metabolite matrix Mm (m x m) where a 1 indicates that the two metabolites share a reaction. We create a Boolean gene-to-metabolite matrix Mg (g x m) where a 1 indicates that the metabolite and gene share a reaction. We create an attribute matrix Ma ((m + g) x m) by vertically concatenating Mm and Mg. We normalize Ma by dividing each row by its sum such that the numbers in each row sum up to 1. Thereby, the values in Ma reflect both that a metabolite is connected to a metabolite or gene and how rare (defining) this connection is. We discard rows from Ma that correspond to metabolites and genes that are not included in both models for these cannot aid in the identification of common metabolites between the models. We discard the columns from Ma that correspond to metabolites that are identified to be the same in both GSMs. We create a scoring matrix Ms (m x m) where the number at position i,j corresponds to the Pearson’s correlation coefficient between columns i and j of Ma. We distinguish between similar and non-similar metabolites in Ms using a minimal score. The minimal score equals a user-defined percentile of scores for metabolites that are present in both models. Identification of lumped reactions A lumped reaction is an artificial reaction that represents the net effect of multiple individual reactions. Therefore, if the lumped and non-lumped representations carry flux in opposite directions, steady state is maintained as they cancel each other out. We use this property to identify lumped reactions by linear programming. The method is described in the following: We determine the directionality for each reaction as forward, backward, or reversible. We transform each reaction such that it only runs in the forward direction; backward reactions are reversed and reversible reactions are split into two reactions. We update the stoichiometric matrix S (m x r) accordingly. We remove the boundary reactions from S as these reflect exchanges of metabolites between the organism and the medium. We define the linear programming (LP) problem: max{c′x} s.t. Sirrx=b lb ≤x≤ub We initiate the variables of the LP problem c: Vector (1 x r) containing the objective coefficient for each reaction. We set each value to -1 to penalize flux through each reaction; this ensures that the total flux in the network is minimized. lb: Vector (1 x r) containing the lower bounds of each reaction. As all reactions are forward reactions, every value is set to 0. ub: Vector (1 x r) containing the upper bounds of each reaction. As all reactions are forward reactions, every value is set to 1000. b: Vector (m x 1) containing the desired accumulation or dissipation of each metabolite. Each value in this vector is set to 0 to ensure a steady-state flux distribution. We select a reaction LR with index iLR to be considered as a lumped reaction. We set: c(iLR) = -1000 lb(iLR) = -1. LR is thus now allowed to carry flux in the backward direction, which results in a positive contribution to the objective value. We run the LP problem as defined under step v. The LP problem returns a flux distribution x that either only contains zeros (no non-lumped representation available), or contains a flux distribution such that the flux through LR is maximized in the reverse direction while having a minimal flux through the rest of the network. In the first case, we skip steps ix and x. In the latter case, we identified a set NL1 of corresponding non-lumped reactions. We save the set NL1 for future reference. We modify the LP problem such that any alternative sets NLx may be identified. c(NL1) = 3 x c(NL1) This effectively further penalizes flux through the reactions of NL1 such that it becomes more ‘rewarding’ to use other reactions. We repeat steps viii-x (replace NL1 by NL2, NL3, …) until: No non-zero solution to the problem exists, or The number of reactions in NLx exceeds a user-defined threshold (default: 5), or There is a recurring set NLx. We filter the different sets NLx such that only sets remain that overlap to a pre-defined extent in gene associations with LR. We repeat steps v-xi such that we obtain sets NL for each reaction in the model. Identification of alternative transport Alternative transport reactions result in the transport of a metabolite between two compartments with a different net reaction. We identify metabolites with alternative transport reactions one metabolite at a time. If a metabolite is present in two or more compartments, we identify all transport reactions for this metabolite by selecting reactions where the metabolite is on both sides of the equation. If two of these reactions transport the metabolite between the same two compartments, these reactions are alternative transport reactions. Identification of invalid transport Invalid transport reactions are reactions that transport metabolites between two unconnected compartments. We identify these by forming a list of all compartments that are directly connected through transport reactions in the IGSMs and asking the user to indicate if any of these are invalid. For any of the invalid compartment connections, we identify reactions that contain metabolites from both compartments; these reactions are invalid transport reactions. Identification of alternative compartmentalization We create a separate stoichiometric matrix Scmp (m x r) for each compartment. These matrices only contain reactions of which all metabolites are in the same compartment. Columns (reactions) that are identical between these matrices represent identical reactions with an alternative compartmentalization. Identification of unknown compartment In the MnXRef namespace, metabolites with an unclear compartmentalization are placed in the compartment UNK_COMP. For each reaction that contains a metabolite in UNK_COMP, we identify reactions from the other IGSM(s) that involve all metabolites with known compartmentalization similarly to the identification of alternative stoichiometries. These reactions are then filtered for reactions that also involve the metabolite with the unknown compartmentalization. Identification of invalid boundary reactions Boundary (exchange) reactions are artificial reactions that represent the exchange of metabolites with the medium. They only involve a single metabolite, and have no metabolites on the other side of the equation. In some models these reactions are lumped together with transport reactions that import metabolites from the extracellular compartment. After the MnXRef namespace conversion these reactions are still annotated as boundary reactions, and are thus easily identified in COMMGEN by searching for boundary reactions with non-extracellular metabolites. Removing a compartment To combine GSMs with an alternative compartmentalization, it is sometimes most straightforward to remove a compartment ‘RC’ from a GSM and move its reactions to a different target compartment ‘TC’. We defined four categories of reactions in RC, which are treated differently when RC is removed: (i) Reactions that only involve metabolites from RC are moved to TC; (ii) Multi-compartment reactions that transport a metabolite between RC and TC are removed; (iii) Multi-compartment reactions involving RC and TC that involve a chemical conversion are kept, but all metabolites from RC are placed in TC; (iv) Multi-compartment reactions involving RC and a metabolite other than TC are kept, and all metabolites from RC are placed in TC. Identification of identical net reactions Identical net reactions are reactions that involve the same set of metabolites in the same stoichiometries, but they may be defined in opposing directions. Therefore, we create a double stoichiometric matrix Sdbl (m x 2r) that contains the normal stoichiometric matrix S (m x r), as well as its negative -S (m x r). We then identify columns (reactions) in Sdbl that are identical. Identification of alternative stoichiometries We convert the S (m x r) matrix to a Boolean (0/1) representation Slog (m x r). We then identify columns in Slog that are identical; these correspond to reactions involving the same metabolites, but in different stoichiometries. Identification of alternative redox pairs GSMs often differ in their involvement of redox pairs in any particular reaction. The first step in identifying these inconsistencies is the creation of a list of redox pairs. COMMGEN comes with a list of commonly used redox pairs in the MnXRef namespace, and this list can be expanded by the user. COMMGEN can suggest expansions for this list by selecting metabolite pairs that co-occur frequently (≥ 80% of reactions). We identify reactions that are identical except for their redox pairs by expanding the stoichiometric matrix S (m x r) to Srdx (m+1 x r) by adding an artificial metabolite ‘redox pair’. Then, for each reaction that involves a redox pair, we put the stoichiometric coefficients of the redox metabolites in Srdx to ‘0’, and add a ‘1’ in the ‘redox pair’ row instead. We then use the same approach as for the identification of alternative stoichiometries to identify reactions that only differ in stoichiometries and redox pairs. Identification of nested reactions We convert the S (m x r) matrix to a Boolean (0/1) representation Slog (m x r). For each column (reaction) we then identify other columns that contain nonzero elements on each row where the respective column has a nonzero element. These sets of columns (reactions) are potentially nested reactions. We then confirm these sets by detecting sets where two or more metabolites that are on the same side of the equation for one reaction, are on the same side of the equation for the other reaction. Identification of similar reactions Similar reactions are reactions from different IGSMs that share a predefined number of genes, substrates and products. We identify similar reactions by constructing three sets of pairs of reactions: (i) reactions that originate from different IGSMs, (ii) reactions that share the required number of substrates and products, and (iii) reactions that share the required number of genes. All combinations of two reactions in each of these three sets are considered similar reactions. Implementation and simulation All computational simulations and analyses were performed using MATLAB [45]. Gurobi [46] was used as linear programming solver for flux balance analysis. Namespace conversion COMMGEN uploads SBML files to MetaNetX.org [47], where the namespace conversion into MnXRef [31] is performed, and downloads the resulting model. Because errors may be introduced at this stage (incorrect namespace conversion of individual metabolites) the mapping is presented to the user who can reject incorrect matches. See S4 Protocol for details. File formats and accessibility The COMMGEN version used for this paper is freely available as MATLAB code as S6 Protocol. A current version of COMMGEN can be found at https://gitlab.com/Rubenvanheck/COMMGEN. Supporting Information S1 Dataset Models. This file contains the original models, the input models, the BCMs, and the RCMs for COMMGEN as well as an overview of the changes made between original and input models. (ZIP) Click here for additional data file. S1 Protocol Automatic RCM creation. This file contains the code that was used in order to obtain the data for Fig 3a–3c. (ZIP) Click here for additional data file. S2 Protocol Effect of COMMGEN on gene rules. Upon the merging of reactions differing in gene rules a choice has to be made in how the final gene rule looks. This file shows how the consensus procedure as applied for this study affects the use of ‘OR’ and ‘AND’ operators. (ZIP) Click here for additional data file. S3 Protocol Growth phenotypes. This file contains the scripts and reference data for the prediction of growth and no-growth phenotypes and subsequent creation of Fig 5a. (ZIP) Click here for additional data file. S4 Protocol Example scripts. This file contains two example scripts of how to start with COMMGEN. (ZIP) Click here for additional data file. S5 Protocol Matching of metabolites between models. This file contains the code that was used in order to obtain the ROC curve in Fig 2c. (ZIP) Click here for additional data file. S6 Protocol COMMGEN. This file contains the code for COMMGEN and is recommended to use when running scripts from other additional files. However, we recommend obtaining the current version of COMMGEN from https://gitlab.com/Rubenvanheck/COMMGEN. (ZIP) Click here for additional data file. We thank Sumana Srivatsa for testing COMMGEN as well as Marco Pagni, Thomas Bernard and Sebastien Moretti for support with MetaNetX and MnXRef. ==== Refs References 1 Thiele I , Palsson BØ (2010 ) A protocol for generating a high-quality genome-scale metabolic reconstruction . Nature protocols 5 : 93 –121 . 10.1038/nprot.2009.203 20057383 2 Ganter M , Kaltenbach H-M , Stelling J (2014 ) Predicting network functions with nested patterns . Nature communications 5 . 3 Oh Y-K , Palsson BO , Park SM , Schilling CH , Mahadevan R (2007 ) Genome-scale reconstruction of metabolic network in Bacillus subtilis based on high-throughput phenotyping and gene essentiality data . Journal of Biological Chemistry 282 : 28791 –28799 . 17573341 4 Zur H , Ruppin E , Shlomi T (2010 ) iMAT: an integrative metabolic analysis tool . Bioinformatics 26 : 3140 –3142 . 10.1093/bioinformatics/btq602 21081510 5 Chandrasekaran S , Price ND (2010 ) Probabilistic integrative modeling of genome-scale metabolic and regulatory networks in Escherichia coli and Mycobacterium tuberculosis . Proceedings of the National Academy of Sciences 107 : 17845 –17850 . 6 Colijn C , Brandes A , Zucker J , Lun DS , Weiner B , et al (2009 ) Interpreting expression data with metabolic flux models: predicting Mycobacterium tuberculosis mycolic acid production . PLoS computational biology 5 : e1000489 10.1371/journal.pcbi.1000489 19714220 7 Beste DJV , Hooper T , Stewart G , Bonde B , Avignone-Rossa C , et al (2007 ) GSMN-TB: a web-based genome-scale network model of Mycobacterium tuberculosis metabolism . Genome biology 8 : R89 17521419 8 Puchalka J , Oberhardt MA , Godinho M , Bielecka A , Regenhardt D , et al (2008 ) Genome-scale reconstruction and analysis of the Pseudomonas putida KT2440 metabolic network facilitates applications in biotechnology . PLoS computational biology 4 : e1000210 10.1371/journal.pcbi.1000210 18974823 9 Monk JM , Charusanti P , Aziz RK , Lerman JA , Premyodhin N , et al (2013 ) Genome-scale metabolic reconstructions of multiple Escherichia coli strains highlight strain-specific adaptations to nutritional environments . Proceedings of the National Academy of Sciences 110 : 20338 –20343 . 10 Oberhardt MA , Puchalka J , Martins dos Santos VAP , Papin JA (2011 ) Reconciliation of genome-scale metabolic reconstructions for comparative systems analysis . PLoS Comput Biol 7 : e1001116 10.1371/journal.pcbi.1001116 21483480 11 Bartell JA , Yen P , Varga JJ , Goldberg JB , Papin JA (2014 ) Comparative metabolic systems analysis of pathogenic Burkholderia . Journal of bacteriology 196 : 210 –226 . 10.1128/JB.00997-13 24163337 12 Babaei P , Ghasemi-Kahrizsangi T , Marashi S-A (2014 ) Modeling the differences in biochemical capabilities of pseudomonas species by flux balance analysis: how good are genome-scale metabolic networks at predicting the differences? The Scientific World Journal 2014 . 13 Plata G , Hsiao TL , Olszewski KL , Llinás M , Vitkup D (2010 ) Reconstruction and flux-balance analysis of the Plasmodium falciparum metabolic network . Molecular systems biology 6 . 14 Jamshidi N , Palsson BØ (2007 ) Investigating the metabolic capabilities of Mycobacterium tuberculosis H37Rv using the in silico strain iNJ661 and proposing alternative drug targets . BMC systems biology 1 : 26 17555602 15 Yim H , Haselbeck R , Niu W , Pujol-Baxley C , Burgard A , et al (2011 ) Metabolic engineering of Escherichia coli for direct production of 1, 4-butanediol . Nature chemical biology 7 : 445 –452 . 10.1038/nchembio.580 21602812 16 Bordbar A , Monk JM , King ZA , Palsson BO (2014 ) Constraint-based models predict metabolic and associated cellular functions . Nature Reviews Genetics 15 : 107 –120 . 10.1038/nrg3643 24430943 17 Reed JL , Famili I , Thiele I , Palsson BO (2006 ) Towards multidimensional genome annotation . Nature Reviews Genetics 7 : 130 –141 . 16418748 18 Stobbe MD , Houten SM , Jansen GA , van Kampen AHC , Moerland PD (2011 ) Critical assessment of human metabolic pathway databases: a stepping stone for future integration . BMC systems biology 5 : 165 10.1186/1752-0509-5-165 21999653 19 Hettne KM , Stierum RH , Schuemie MJ , Hendriksen PJM , Schijvenaars BJA , et al (2009 ) A dictionary to identify small molecules and drugs in free text . Bioinformatics 25 : 2983 –2991 . 10.1093/bioinformatics/btp535 19759196 20 Herrgård MJ , Swainston N , Dobson P , Dunn WB , Arga KY , et al (2008 ) A consensus yeast metabolic network reconstruction obtained from a community approach to systems biology . Nat Biotechnol 26 : 1155 –1160 . 10.1038/nbt1492 18846089 21 Radrich K , Tsuruoka Y , Dobson P , Gevorgyan A , Swainston N , et al (2010 ) Integration of metabolic databases for the reconstruction of genome-scale metabolic networks . BMC Syst Biol 4 : 114 10.1186/1752-0509-4-114 20712863 22 Feist AM , Henry CS , Reed JL , Krummenacker M , Joyce AR , et al (2007 ) A genome-scale metabolic reconstruction for Escherichia coli K-12 MG1655 that accounts for 1260 ORFs and thermodynamic information . Molecular systems biology 3 . 23 Kumar A , Suthers PF , Maranas CD (2012 ) MetRxn: a knowledgebase of metabolites and reactions spanning metabolic models and databases . BMC bioinformatics 13 : 6 10.1186/1471-2105-13-6 22233419 24 Thiele I , Swainston N , Fleming RMT , Hoppe A , Sahoo S , et al (2013 ) A community-driven global reconstruction of human metabolism . Nat Biotechnol 31 : 419 –425 . 10.1038/nbt.2488 23455439 25 Rienksma RA , Suarez-Diez M , Spina L , Schaap PJ , Martins dos Santos VA (2014 ) Systems-level modeling of mycobacterial metabolism for the identification of new (multi-)drug targets . Semin Immunol 26 : 610 –622 . 25453232 26 Monk J , Nogales J , Palsson BO (2014 ) Optimizing genome-scale network reconstructions . Nat Biotechnol 32 : 447 –452 . 10.1038/nbt.2870 24811519 27 Thiele I , Hyduke DR , Steeb B , Fankam G , Allen DK , et al (2011 ) A community effort towards a knowledge-base and mathematical model of the human pathogen Salmonella Typhimurium LT2 . BMC systems biology 5 : 8 10.1186/1752-0509-5-8 21244678 28 Aung HW , Henry SA , Walker LP (2013 ) Revising the representation of fatty acid, glycerolipid, and glycerophospholipid metabolism in the consensus model of yeast metabolism . Industrial Biotechnology 9 : 215 –228 . 24678285 29 Chindelevitch L , Stanley S , Hung D , Regev A , Berger B , et al (2012 ) MetaMerge: scaling up genome-scale metabolic reconstructions with application to Mycobacterium tuberculosis . Genome Biol 13 : r6 10.1186/gb-2012-13-1-r6 22292986 30 Stobbe MD , Swertz MA , Thiele I , Rengaw T , van Kampen AH , et al (2013 ) Consensus and conflict cards for metabolic pathway databases . BMC Syst Biol 7 : 50 10.1186/1752-0509-7-50 23803311 31 Bernard T , Bridge A , Morgat A , Moretti S , Xenarios I , et al (2012 ) Reconciliation of metabolites and biochemical reactions for metabolic networks . Briefings in bioinformatics . 32 Hucka M , Finney A , Sauro HM , Bolouri H , Doyle JC , et al (2003 ) The systems biology markup language (SBML): a medium for representation and exchange of biochemical network models . Bioinformatics 19 : 524 –531 . 12611808 33 Nogales J , Palsson BO , Thiele I (2008 ) A genome-scale metabolic reconstruction of Pseudomonas putida KT2440: iJN746 as a cell factory . BMC systems biology 2 : 79 10.1186/1752-0509-2-79 18793442 34 Kumar VS , Maranas CD (2009 ) GrowMatch: An Automated Method for Reconciling In Silico/In Vivo Growth Predictions . Plos Computational Biology 5 . 35 Matthews BW (1975 ) Comparison of the predicted and observed secondary structure of T4 phage lysozyme . Biochimica et Biophysica Acta (BBA)-Protein Structure 405 : 442 –451 . 36 Henry CS , Zinner JF , Cohoon MP , Stevens RL , et al (2009 ) iBsu1103: a new genome-scale metabolic model of Bacillus subtilis based on SEED annotations . Genome Biol 10 : R69 10.1186/gb-2009-10-6-r69 19555510 37 Mo ML , Palsson BO , Herrgård MJ (2009 ) Connecting extracellular metabolomic measurements to intracellular flux states in yeast . BMC Syst Biol 3 : 37 10.1186/1752-0509-3-37 19321003 38 Kuepfer L , Sauer U , Blank LM (2005 ) Metabolic functions of duplicate genes in Saccharomyces cerevisiae . Genome Res 15 : 1421 –1430 . 16204195 39 Thiele I , Vlassis N , Fleming RMT (2014 ) fastGapFill: Efficient gap filling in metabolic networks . Bioinformatics: btu 321 . 40 Noor E , Bar-Even A , Flamholz A , Lubling Y , Davidi D , et al (2012 ) An integrated open framework for thermodynamics of reactions that combines accuracy and coverage . Bioinformatics 28 : 2037 –2044 . 10.1093/bioinformatics/bts317 22645166 41 Mintz-Oron S , Aharoni A , Ruppin E , Shlomi T (2009 ) Network-based prediction of metabolic enzymes' subcellular localization . Bioinformatics 25 : i247 –i1252 . 10.1093/bioinformatics/btp209 19477995 42 Dreyfuss JM , Zucker JD , Hood HM , Ocasio LR , Sachs MS , et al (2013 ) Reconstruction and validation of a genome-scale metabolic model for the filamentous fungus Neurospora crassa using FARM . PLoS computational biology 9 : e1003126 10.1371/journal.pcbi.1003126 23935467 43 Henry CS , DeJongh M , Best AA , Frybarger PM , Linsay B , et al (2010 ) High-throughput generation, optimization and analysis of genome-scale metabolic models . Nat Biotechnol 28 : 977 –982 . 10.1038/nbt.1672 20802497 44 Schellenberger J , Que R , Fleming RM , Thiele I , Orth JD , et al (2011 ) Quantitative prediction of cellular metabolism with constraint-based models: the COBRA Toolbox v2.0 . Nat Protoc 6 : 1290 –1307 . 10.1038/nprot.2011.308 21886097 45 The MathWorks I MATLAB. Natick, Massachusetts, United States. 46 Gurobi Optimization I (2015) Gurobi Optimizer Reference Manual. 47 Ganter M , Bernard T , Moretti S , Stelling J , Pagni M (2013 ) MetaNetX.org: a website and repository for accessing, analysing and manipulating metabolic networks . Bioinformatics 29 : 815 –816 . 10.1093/bioinformatics/btt036 23357920 48 Nookaew I , Jewett MC , Meechai A , Thammarongtham C , Laoteng K , et al (2008 ) The genome-scale metabolic model iIN800 of Saccharomyces cerevisiae and its validation: a scaffold to query lipid metabolism . BMC Syst Biol 2 : 71 10.1186/1752-0509-2-71 18687109 49 Duarte NC , Herrgård MJ , Palsson BO (2004 ) Reconstruction and validation of Saccharomyces cerevisiae iND750, a fully compartmentalized genome-scale metabolic model . Genome Res 14 : 1298 –1309 . 15197165 50 Molina-Henares MA , De La Torre Jus , García-S A , Molina-Henares AJus , Herrera MC , et al (2010 ) Identification of conditionally essential genes for growth of Pseudomonas putida KT2440 on minimal medium through the screening of a genome-wide mutant library . Environmental Microbiology 12 : 1468 –1485 . 10.1111/j.1462-2920.2010.02166.x 20158506 51 Sassetti CM , Boyd DH , Rubin EJ (2003 ) Genes required for mycobacterial growth defined by high density mutagenesis . Molecular microbiology 48 : 77 –84 . 12657046

PMC005xxxxxx/PMC5001717.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 27564373PONE-D-16-1371110.1371/journal.pone.0161857Research ArticlePhysical SciencesMathematicsAlgebraAlgebraic GeometryPhysical SciencesMathematicsApplied MathematicsAlgorithmsResearch and Analysis MethodsSimulation and ModelingAlgorithmsComputer and Information SciencesCryptographyPhysical SciencesMathematicsCryptographySocial SciencesPolitical ScienceNational SecurityEngineering and TechnologyEquipmentEngineering and TechnologyEquipmentSafety EquipmentMedicine and Health SciencesPublic and Occupational HealthSafetySafety EquipmentComputer and Information SciencesInformation TechnologyDatabasesResearch and Analysis MethodsDatabase and Informatics MethodsConstructing Pairing-Friendly Elliptic Curves under Embedding Degree 1 for Securing Critical Infrastructures Constructing Pairing-Friendly EC under Embedding Degree 1 for Securing Critical Infrastructureshttp://orcid.org/0000-0002-5495-6417Wang Maocai 12Dai Guangming 12*Choo Kim-Kwang Raymond 123Jayaraman Prem Prakash 4Ranjan Rajiv 5 1 School of Computer, China University of Geosciences, Wuhan, Hubei, China 2 Hubei Key Laboratory of Intelligent Geo-Information Processing, China University of Geosciences, Wuhan, Hubei, China 3 Department of Information Systems and Cyber Security, University of Texas at San Antonio, San Antonio, Texas, United States of America 4 RMIT University, Melbourne, Australia 5 University of Newcastle, Newcastle, United Kingdom Shi Yongtang Editor Nankai University, CHINA Competing Interests: The authors have declared that no competing interests exist. Conceptualization: MCW GMD. Data curation: MCW KRC. Formal analysis: MCW RR. Funding acquisition: MCW GMD. Investigation: MCW. Methodology: MCW GMD. Project administration: MCW. Resources: MCW. Software: MCW GMD. Supervision: GMD. Validation: MCW GMD PPJ. Visualization: PPJ RR. Writing – original draft: MCW KRC. Writing – review & editing: MCW RR. * E-mail: cugdgm@126.com2016 26 8 2016 11 8 e01618575 4 2016 13 8 2016 © 2016 Wang et al2016Wang et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Information confidentiality is an essential requirement for cyber security in critical infrastructure. Identity-based cryptography, an increasingly popular branch of cryptography, is widely used to protect the information confidentiality in the critical infrastructure sector due to the ability to directly compute the user’s public key based on the user’s identity. However, computational requirements complicate the practical application of Identity-based cryptography. In order to improve the efficiency of identity-based cryptography, this paper presents an effective method to construct pairing-friendly elliptic curves with low hamming weight 4 under embedding degree 1. Based on the analysis of the Complex Multiplication(CM) method, the soundness of our method to calculate the characteristic of the finite field is proved. And then, three relative algorithms to construct pairing-friendly elliptic curve are put forward. 10 elliptic curves with low hamming weight 4 under 160 bits are presented to demonstrate the utility of our approach. Finally, the evaluation also indicates that it is more efficient to compute Tate pairing with our curves, than that of Bertoni et al. http://dx.doi.org/10.13039/501100001809National Natural Science Foundation of China41571403http://orcid.org/0000-0002-5495-6417Wang Maocai http://dx.doi.org/10.13039/501100002858China Postdoctoral Science Foundation2012T50681http://orcid.org/0000-0002-5495-6417Wang Maocai http://dx.doi.org/10.13039/501100002858China Postdoctoral Science Foundation2011M501260http://orcid.org/0000-0002-5495-6417Wang Maocai http://dx.doi.org/10.13039/501100001809National Natural Science Foundation of China61472375Dai Guangming This work was supported by National Natural Science Foundation of China (Grant No. 41571403 and 61472375, http://www.nsfc.gov.cn/) and China Postdoctoral Science Foundation (Grant No. 2012T50681 and 2011M501260, www.chinapostdoctor.org.cn). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body 1 Introduction Many countrieshave thrived on the wealth fromthe information technologies(IT) have enabled, and IT forms the backbone of many aspects of the critical infrastructure sectors [1, 2]. There are 16 critical infrastructure sectors in the U.S. [3]. As noted by both scholars [4–10] and government agencies, such as U.S. Homeland Security, the critical infrastructure represents systems and assets, and it is also defined in detailed [3]. The interconnective of the systems in the critical infrastructure sector, and the increasing sophistication, scale and the persistent nature of cyber attacks against such systems, can potentially result in equipment being forced to operate beyond its intended design and safety limits, resulting in cascading system malfunctions and shut downs such as the collapse of an entire electricity grid; or operating procedures or conditions being manipulated to slow the effort of restoring essential services [11, 12]. It is, therefore, unsurprising that the cyber security of a nation’s critical infrastructure (including assets, networks, and systems) is regarded as a top priority of national security by countries around the world [13–17]. One of the key requirements in critical infrastructure cyber security is information confidentiality, and the cryptography is generally the core technology to provide information confidentiality [18]. Identity-based cryptography(IBC) is a relatively new branch of cryptography, which can directly compute a user’s public key using publicly available information from the user’s identity [19]. Therefore, one does not need to distribute his digital certificate signed by a certificate authority (CA), or query the certificate database to get the other party’s public key when conducting electronic transactions. In other words, IBC resolves the challenges and complexity associated with certificate management and traditional public-key cryptosystem. A limitation of IBC is, however, the computation cost involving in constructing the pairings [20]. IBC has the subject of various research, but it remains a topic of ongoing research interest, and one of the research challenges is the generation of efficient parameters such as pairing-friendly elliptic curves. The existing efficient algorithms to compute Weil and Tate pairings [21, 22] are generally based on Miller’s algorithm [23]on (hyper) elliptic curves. One line of research which focuses on reducing the loop in Miller’s algorithm was initiated by Duursma-Lee [24] and, subsequently extended by Barreto et al. [25] to supersingular abelian varieties. In practice, the cryptographic pairings used to construct these systems are based on the Weil and Tate pairings on elliptic curves over finite fields [26]. Both pairings are a bilinear map from an elliptic curve group on the finite field Fp to the multiplicative group of some extension field Fpk. The parameter k is called the embedding degree of the elliptic curve. The pairing is considered to be secure if both discrete logarithms in the groups E(Fp) and Fpk are computationally infeasible. To optimize the application performance, the parameters p and k should be determined according to this standard that both discrete logarithm problems approximately have the equal difficulty when using the best known algorithms. Moreover, a large prime factor r should be included in the order of the group #E(Fp). For example, if the large prime factor r ≥ 2160, the pairing is generally considered to be safe against existing attacks. Therefore, it is essential to be able to construct elliptic curves efficiently for arbitrary p and k values to differ the security level or to meet the requirement of discrete log in future improvements. This is the gap we attempt to address in this paper. This paper is organized as follows. In the next two sections, we introduce the reader to related literature and Tate pairing, respectively. In Section 4, we describe our approach to constructing pairing-friendly elliptic curves under embedding degree 1 and preliminary evaluation results to demonstrate utility and practicality. Our discussion and concluding remarks are provided in the last two sections. 2 Related Work Constructing elliptic curves with various embedding degrees has been the subject of ongoing research. For example, Cocks and Pinch [27] constructed the curves with arbitrary embedding degree k, but the efficiency is very low because the size q of the field Fp is limited by the subgroup of prime order r with q ≈ r2. Fotiadis and Konstantinou [28] presented two general methods to produce sparse families and applied them to four embedding degrees k, where k. Barreto and Naehrig [29] constructed the curves of prime order with k = 12. Freeman [30] proposed a construction for the curves with embedding degree k = 10. A complete characterization of common elliptic curves of prime order with k = 3, 4, or 6, is provided by Miyaji, Nakabayashi, and Takano [31]. Menezes, Okamoto, and Vanstone [32] illustrated that embedding degree k should be not more 6 in a supersingular elliptic curve, especially k ≤ 3 and k ≠ 2 or k ≠ 3. Some researches [33] reduced the ratio p=log plog r for arbitrary k between the characteristicp of the finite field and the prime order r of the subgroup. However, no concrete examples have been proposed with ρ small enough to construct curves with prime order. In fact, if k = 1, the pairing will become a bilinear map from the elliptic curve group on the finite field Fp to the elliptic curve group on the same finite field Fp. In other words, we would not involve the extension field Fpk when computing the pairing, which is the constraintin pairing-based cryptography applications. Izuta, Nogami and Morikawa [34] proposed a method for generating a certain composite order ordinary pairing-friendly elliptic curve of embedding degree 1. In their method, the order has two large prime factors such as the modulus of RSA cryptography. Lee and Park [35] proposed a new algorithm to construct Brezing-Weng-like elliptic curves having the Complex Multiplication(CM) equation of degree 1, as well as presenting new families of curves with larger discriminants. It is clear from the literature that pairing-friendly elliptic curves under embedding degree 1 are constructed on the base field, rather than the extension field, which can significantly improve the computation efficiency of Tate pairing. This is the gap that this paper attempts to address. More specifically, this paper proposes an effective method to construct pairing-friendly elliptic curves with low hamming weight 4 under embedding degree 1. 3 Tate Pairing In practice, as the theoretical model is unknown, we use the Monte-Carlo method [36] to generate the required data based on a fixed theoretical model. Weil pairing was first introduced into cryptography by Menezes, which was used to study the elliptic curve discrete logarithm problem on certain elliptic curves [32]. Extending on the work of Menezes, Frey introduced Tate pairing to cryptography [37], which is now widely used to design pairing-based cryptosystems because Tate pairing is twice as efficient as Weil pairing. Let E be an elliptic curve over a finite field Fp, and r be a positive integer which is co prime to p. In most applications, r is a prime and r|#E(Fp). Let k be a positive integer such that the field Fpk. contains the r-th roots of unity, and k is called the embedding degree. Then Tate pairing is a mapping [38]: t(P,Q):E(Fpk) [r]×E(Fpk)/rE(Fpk)→Fpk*/(Fpk*)r According to the definition of Tate pairing, if the embedding degree k ≠ 1, then the computation of Tate pairing is related to the extension field Fpk, and the computation process will be time-consuming. However, if the embedding degree k = 1, the computation of Tate pairing only runs on the base field Fpk rather than the extension field Fpk. This will greatly improve the computation efficiency of Tate pairing. In Tate pairing, both the point P and the point Q are from two different subgroups with the same order r as subgroup E(Fpk[r]) and (Fpk*)r respectively. That is to say, if k = 1, then the point P and the point Q come from two different subgroup G1 and G2 of E(Fp) with the same order r, and G1 ∩ G2 = ∅. However, “How to construct the elliptic curve which includes two different groups with the same order r when r is a large prime with r ≥ 2160?” and “How to find the two different groups?” are two key challenges in designing pairing-friendly elliptic curves under embedding degree 1. In this paper, we propose an effective algorithm to construct pairing-friendly elliptic curves under embedding degree 1. In our algorithm, it can be ensured that both the point P and the point Q are from two different subgroups with the same order r, which enables the computation of Tate pairing to run only on the base field. 4 Constructing Pairing-friendly Elliptic Curves In this section, a new method to generate pairing-friendly elliptic curves is proposed, which comprises three algorithms as follows. The first algorithm is used to generate a large prime of low hamming with weight 4. The second algorithm is used to generate the finite field p, the order u of a non-supersingular elliptic curve over Fp, the order r of a point on the elliptic curve. The last algorithm is used to construct pairing-friendly elliptic curves under embedding degree 1. 4.1 The Construction Method In the common method [31, 35]to construct elliptic curves, the equation u = p + 1 ± W is used to generate the parameters of the elliptic curves. This equation provides a means to determine the order #E of an elliptic curve E according to the characteristic p of the finite field Fp. However, the order #E generated using the equation is generally unable to meet the security requirement. Therefore, it is a challenge to generate a suitable elliptic curve using the common method. Moreover, even if a suitable elliptic curve can be generated, it will take a long time. For example, in the method of Izuta, Nogami and Morikawa [34], it will take about 20 hours to generate an elliptic curve. In our method, we present a new equation p = u ± W + 1 to generate the parameters of elliptic curves. On first glance, the new equation may appear similar to the common equation. However, in the new equation, the order #E is known, and we need to obtain p from the order #E (rather than the order #E from p). Thus, we only need to determine the characteristic p of the finite field Fp from the order #E of an elliptic curve E, and our algorithm 2 describes the process required to generate p from the order #E. In other words, we can generate an elliptic curve under arbitrary order, while the order #E of an elliptic curves E can be trivially obtained using u = r * r (from the security requirement), where r is a large prime, and r has a low hamming with weight 4 (based on our algorithm 1). As the order of the subgroup is a large prime of low hamming with weight 4, the efficiency of generating elliptic curves is significantly improved. More specifically, our method requires about 200 ms to generating a suitable elliptic curve. Theorem 1 If E is a non-super singular elliptic curve over Fp with order u, D is the CM discriminant for p, according to the discriminant condition 4p = W2 + DV2 and u = p + 1 ± 1, then p=u±X+1 where W = X ± 2, V = Y. Proof. It is well known that the CM discriminant D for p meets the Eqs (1) and (2) for every non-super singular elliptic curve over Fp with order u. 4p=W2+DV2(1) u=p+1±W(2) The Eq (3) can be gotten from the Eq (2). 4u=4p+4±4W(3) The Eq (4) can be gotten by replacing 4p with the Eq (1) in the Eq (3). 4u=W2+DV2+4±4W(4) The Eq (4) can be written as the Eq (5). 4u=(W±2)2+DV2(5) The Eq (5) can be written as the Eq (6). 4u=X2+DY2(6) where X = W ± 2 and Y = V. Therefore, the Eq (1) can be converted to the Eq (7) with X = W ± 2. 4p=(X±2)2+DY2(7) The Eq (8) can be gotten from the Eqs (6) and (7) 4p=4u±4X=4(8) The Eq (8) can be be written as the Eq (9) p=u±X+1(9) This ends the proof. Theorem 1 provides a method to calculate the characteristic p of the finite field Fp according to the order u of an elliptic curve. That is to say, for any elliptic curve with the order u expected, we can easily calculate the characteristic p of the finite field Fp according to the Eq (9). This is a new way, which can generate an elliptic curve under any order we expected. In Miller algorithm of computing Tate pairing, if some bit of the binary representation for the order r of subgroup is ‘1’, operators would be needed to compute multiplication and inverse operations [39]. Otherwise, (i.e. if the binary bit is ‘0’), no additional operator is needed. It is clear that the process to compute Tate pairing will be more efficient if the binary representation of the order r has fewer ‘1’ bits and more ‘0’ bits. This forms the basis of the three relative algorithms. 4.2 Algorithm 1 Algorithm 1 outlines the method to generate a large prime of low hamming with weight 4. In other words, there are only two ‘1’ bits in addition to the highest bit and the lowest bit in the binary representation for the large prime. The large prime will be used as the order r of subgroup in algorithms 2 and 3. In algorithm 1, the input parameter is the length m(m ≥ 160) of the binary representation for the large prime, the output result is the large prime r of low hamming with weight 4. Algorithm 1. Generating a large prime of low hamming with weight 4. Input: The length m(m ≥ 160) of the binary representation for the large prime; a positive integer t for the number of trials. Output: The large prime r of low hamming with weight 4. step 1 Choose random s, t in the interval (0, m − 1) to ensure 0 < s < t < m − 1; step 2 r ← 20 + 2s + 2t + 2m − 1; step 3 Compute v and an odd value w, such that r − 1 = 2v w step 4 For j from 1 to t do step 4.1 Choose random a in the interval 0 < a < r; step 4.2 Set b ← aw mod r step 4.3 If b = 1 or b = r − 1, goto step 4.6; step 4.4 For i from 1 to v − 1 do step 4.4.1 Set b ← b2 mod r step 4.4.2 If b = r − 1 goto step 4.6; step 4.4.3 if b = 1, goto step 1; step 4.4.4 Next i. step 4.5 goto step 1; step 4.6 Next j; step 5 Output r. 4.3 Algorithm 2 Algorithm 2 describes the method to generate the finite field p, the order u of a non-supersingular elliptic curve over Fp, and the order r of a point on the elliptic curve according to the length m(m ≥ 160) of the finite field p. Algorithm 2. Generating the finite field p, the order u of a non-supersingular elliptic curve over Fp, and the order r of a point on the elliptic curve. Input: The length m(m ≥ 160) of the finite field p. Output: The finite field p, the order u of a non-super singular elliptic curve over Fp, the order r of a point on the elliptic curve. step 1 Generate a large prime r of low hamming with weight 4 using algorithm 1; step 2 Compute the order u of a non-supersingular elliptic curve u = r2; step 3 Assign D = 3, set X = r, Y = r, such that the values of both X and Y satisfy the condition 4u = X2 + DY2; step 4 Compute p = r2 + r + 1 according to p = u ± X + 1 when u = r2, X = r; step 5 If p is not a prime, goto Step 1; step 6 Output the finite field p, the order u of a non-supersingular elliptic curve over Fp, the order r of a point on a elliptic curve. We would also remark that “the IEEE Standard Specifications for Public-Key Cryptography” [40] recommends that in the construction of a curve with prescribed CM, if D = 3, the coefficients a0 and b0 of E should be 0 and 1 respectively. 4.4 Algorithm 3 Algorithm 3 presents the method to construct pairing-friendly elliptic curves under embedding degree 1. We assume that there are two different subgroups with the same order r on the elliptic curve generated by algorithm 3, where r is a large prime. In algorithm 3, the input parameter is the length m(m ≥ 160) for the subgroup order, and the output results are a, b and the prime p as the parameters of the elliptic curve y2 ≡ x3 + ax + b mod p, low hamming prime r as the order of subgroup, point P1 as the base point for generating subgroup G1 while calculating Tate pairing, where rP1 = 0, and point P2 as the base point for generating subgroup G2 while calculating Tate pairing where rP1 = 0, rP2 = 0 and G1 ∩ G2 = ∅. Algorithm 3 is designed to be convenient for users generating pairing-friendly elliptic curves under embedding degree 1, as the only input parameter is the length of the binary representation for the order r of the subgroup. Algorithm 3 runs by calling algorithm 2, which in turn calls algorithm 1. Algorithm 3. Constructing pairing-friendly elliptic curves. Input: The length m(m ≥ 160) for the subgroup order. Output: a, b and the prime p denote the parameters of the elliptic curve y2 ≡ x3 + ax + b mod p, low hamming order r denotes the order of subgroup, point P1(rP1 = 0) and point P2(rP2 = 0). step 1 Generate the finite field p, the order u of a non-supersingular elliptic curve over Fp, the order r of a point on the elliptic curve using algorithm 2; step 2 Select an integer ζ with 0 < ζ < p; step 3 Set a ← 0 and b ← b0ζ mod p; step 4 Locate a point P1 with order r on the curve y2 ≡ x3 + ax + b mod p. step 5 If the output of Step 4 is in the wrong order, goto Step 2. step 6 Locate a point P2 with order r on the curve y2 ≡ x3 + ax + b mod p, where P2 ∉ {kP1|k ∈ {1, 2…, r}}. step 7 The output p, a, b as the parameters of the elliptic curve y2 ≡ x3 + ax + b mod p, the large prime r with low hamming weight as the order of subgroup, the point P1 as the base point for generating subgroup G1 while calculating Tate pairing, where rP1 = 0, and the point P2 as the base point for generating subgroup G2, while rP2 = 0 and G1 ∩ G2 = ∅. The elliptic curve generated by algorithm 3 can potentially include two different subgroups G1 and G2, with large prime order r with low hamming weight for computing Tate pairing. Because the order r of subgroup is a public parameter, these parameters generated by the algorithms presented in the paper do not impact on the security of Pairing-based cryptosystems(PBC). 4.5 Preliminary Findings We implement the construction described in Section 4.1 using Pentium 4 PC (CPU 3.06GHz), and the findings are as follows. Algorithm 1: r = 730750818686719107034401070324602422792720220161 = 2159 + 2124 + 228 + 20 Algorithm 2: p = 53399675901131022287481940452568268554861807611629722703698801201 2286237103997609758897031086083 r = 730750818686719107034401070324602422792720220161 u = r2 = 53399675901131022287481940452568268554861807611556647621830129 2905251836033673007336104310865921 Algorithm 3: Table 1 describes 10 elliptic curves generated by algorithm 3 under the above p, r, u. 10.1371/journal.pone.0161857.t001Table 1 10 pairing-friendly curves with low hamming weight 4 under given p, r, u(r with 160 bits). Parameters b P1 P2 The 1st group 5582 (92216901,324171638614811955738700451351453938462743355913304125667979195175749628071873615636670182400781) (2900911840,470565327089465608766717290556724343611875827253253971039274229090627830093258979017255917746829) The 2nd group 411 (6456,200783653312643253000427685185361672824889578029877838466376294032581210908578307946592006236274) (7718449758221,246594700063950682499345743858581550409793007029702423620631032309187268322219527336805985703980) The 3rd group 6888558 (63,483447298606802197007667782086007062212874881060089832042154800397367248862591909523693105053579) (503,77335678175724311469109067552943334864235153595640796029200821503606472539285202409938146494560) The 4th group 1852511737533 (28136114,242158699775814654792914165109030127513135019773154553510244588647142304534134418272078033955490) (86590,121551762235427048604306704938720769730830601113589433595621413717086224050169417598237928322315) The 5th group 111158 (9069952,191884782129835076896430782729279489410474467097317768476761535512656282281715989056396995028939) (40,362063866070142646287391901715038404328540944674222712928339139412213864564948793769019222805261) The 6th group 7134 (352,373120637403567989697297819791130549970377887442049625481567602466046997349176087351878502769389) (1171827216,497218853134580777525964760410326161237125053509830407941623317513914592421492486740293952308203) The 7th group 562 (7751170,170089631638765123245772026413070351656555635734890000712753208456308661954335842959630945562639) (9123978,491696588250751128316690742989376247219892973435085382978639044870863996676582978948360971631311) The 8th group 1105557501121 (96209917051711,525449385155426365101090731761951180537095809586740545805755386839131992568764934554139827350205) (12835,74127495296715015118497558174623723156473745792295184182056487700402725864604342423009775785761) The 9th group 814 (76110676327,455230770668674635001073148949612110022716252695998918206294806414038955424454903144922555015137) (856171875,173617915786825757928929966543011205848033233611185463361858494144333951770711174509831733937961) The 10th group 1110977 (935542646001,425605723028492010195922479602517755514636860155519341225132168354176911010027713634301519279864) (211058810,424408607333334853102700723328774094337451473846030924917501392093461859795155755424956700868152) 5 Discussion In the Miller algorithm, for every bit of the order r of the subgroup, we would need to compute 16 multiplication and 7 inverse operations. If the bit is 1, however,we would need to compute 11 multiplication and five inverse operations. For the order r of the subgroup with 160 bits in ordinary PBCs, there are 80 ‘1’ bits on average. Therefore, we would need to compute 3,429 multiplication and 1,515 inverse operations. It is pleasing to note that using the parameters in our approach, we only need 2,593 multiplication and 1,135 inverse operations, as shown in Table 2. 10.1371/journal.pone.0161857.t002Table 2 Efficiency analysis. The ordinary PBC PBC with parameter in the paper Every bit (160 bits) Every bit with 1 (79 bits) Every bit (160 bits) Every bit with 1 (3 bits) Multiple Inverse Multiple Inverse Multiple Inverse Multiple Inverse 16 7 11 5 16 7 11 5 2560 1120 869 395 2560 1120 33 15 Total Multiple:3429 Inverse:1515 Multiple:2593 Inverse:1135 An inverse operation is estimated to be 5.18 multiplication operations [39], and implementing our method outlined in this paper will save 24.9% of the time required to compute the Tate pairing: 2593+1135*5.183429+1515*5.18=0.751=75.1% To demonstrate the practicality of the new method we proposed,using the parameters with 160 bits presented in Table 1, we implement a proof-of-concepton a Pentium 4 PC (CPU 3.06GHz) in Table 3, using the parameters with 160 bits presented in Table 1. 10.1371/journal.pone.0161857.t003Table 3 Comparative summary of Tate pairing computations. Parameters The result from Bertoni et al. [39] The result from this paper Platform PentiumIII @ 1GHz Pentium IV@ 3.06GHz Length of prime 160 bits 160 bits Low Hamming Weight 3 4 Time for a Tate pairing 41ms 12.93ms As shown in Fig 1., our implementation takes 12.93 ms to compute a pairing. We then compared with the findings from Bertoni et al. [39], as shown in Table 3. In the latter, the large prime of the order of the subgroup is 160 bits, but with a Hamming weight equal to 3 and the embedding degree of 2. As shown in Table 3, our algorithm is more computationally efficient compared to that of Bertoni et al. 10.1371/journal.pone.0161857.g001Fig 1 The result of computing Tate pairing on the first group curve. The first 9 lines gives the parameters of the first group curves. Then the result of e(P, Q), e(2P, Q), e(P, 2Q), e(3P, Q), e(P, 3Q), e(P, Q)2 and e(P, Q)3 are given and the bilinear property is verified. The computation results depicted in Fig 1. can also be verified using the bilinear characteristic of Tate pairing, as explained below: t(P,2Q)=t(2P,Q)=t(P,Q)2 t(P,3Q)=t(3P,Q)=t(P,Q)3 Recall that in Tate pairing, if the embedding degree k ≠ 1, then the computation of Tate pairing is related to the extension field Fpk, which is very time consuming. Building on Miller’s algorithm, we present an effective algorithm to construct pairing friendly elliptic curves with low hamming weight 4 under embedding degree 1, which enables the computation of Tate pairing only on the base field. 6 Conclusion Ensuring information confidentiality in critical infrastructures will be increasingly important in our increasingly interconnected world. In this paper, we studied the generation method of pairing-friendly elliptic curves for identity-based cryptography(IBC), with the aim to significantly improve the computation efficiency of IBC. We demonstrated how pairing-friendly elliptic curves can be efficiently conducted, both in theory and practice which can be deployed in critical infrastructure systems, such as cyber-physical systems with limited resources [40]. In our approach,pairings computing requires only the base field, rather than the extension field. More specifically, in this paper, we described and conducted a preliminary analysis of the new method to construct pairing-friendly elliptic curves under embedding degree 1. Unlike the existed traditional CM methods,the parameters are not randomly generated in our method. The parameters are computed under a given expression, which significantly improves the efficiency of generating elliptic curve. Moreover, in our algorithm, the only input parameter is the binary length of the large prime r, and then all parameters of the elliptic curve can be rapidly generated. Our method consists of three algorithms, namely: an algorithm to generate low hamming prime r according to the expected length of the large primer, which is also used as the order of the subgroup; an algorithm to calculate the character p of the finite field Fp and the order u of the elliptic curve according to the prime r; and an algorithm to generate the pairing-friendly elliptic curves and the two different points P1 and P2 on the elliptic curve with the same order r. It also ensures G1 ∩ G2 = ∅, where G1 is the subgroup generated by P1 and G2 is the subgroup generated by P2, G1 and G2 are two different subgroups of E with the same order r. The paper also provided 10 elliptic curves with low hamming, weight 4 under 160 bits generated using our algorithms, which demonstrated the utility of our method. Then, we demonstrated the practicality of our method by implementing the method using Tate pairing. Our curves can be applied in real word such as Internet of Things(IoT), Electronic Commerce(EC) and Copyright Protection(CP). In fact, in all fields, which are involved in public key cryptography, the proposed method can be applied to implement digital signature, key management and authentication protocol [41–43]. The future work includes two aspects. The first aspect is to optimize Miller’s algorithm to improve the computation efficiency of Tate pairing. The other aspect is to apply the elliptic curves constructed by our method to the practical cryptosystem. Supporting Information S1 File Pairing-friendly elliptic curves under embedding degree 1 with 160 bits. There are 10 group pairing-friendly elliptic curves under embedding degree 1 with 160 bits. In every group, the parameters of p, r, #E, b, P, Q are given. The parameters of a is equal 0 in all groups. (PDF) Click here for additional data file. S2 File Pairing-friendly elliptic curves under embedding degree 1 with 190 bits. There are 10 group pairing-friendly elliptic curves under embedding degree 1 with 160 bits. In every group, the parameters of p, r, #E, b, P, Q are given. The parameters a is equal 0 in all groups. (PDF) Click here for additional data file. ==== Refs References 1 Dong MX , Ota K , Laurence T , et al LSCD: A Low-Storage Clone Detection Protocol for Cyber-Physical Systems . IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems , 2016 , 35 (5 ):712 –723 . 10.1109/TCAD.2016.2539327 2 Wang L. , Tao J. , Ranjan R. , et al G-Hadoop: MapReduce across distributed data centers for data-intensive computing . Future Generation Computer Systems , 2013 , 29 (3 ):739 –750 10.1016/j.future.2012.09.001 3 Kepler D , Heasley P Implementation of EO 13636 and PPD-21: Draft Report and Recommendations . National Infrastructure Advisory Council , 2013 4 Zhao J. , Wang L. , Tao J. , etc. A security framework in G-Hadoop for big data computing across distributed Cloud data centres . Journal of Computer and System Sciences , 2014 , 80 (5 ):994 –1007 . 10.1016/j.jcss.2014.02.006 5 Chen D. , Liu Z. , Wang L. ,etc. On-demand service hosting on production grid infrastructures . MONET , 2013 , 18 (5 ): 651 –663 . 6 Wang L. , Kurze T. , Tao J. , etc. Natural Disaster Monitoring with Wireless Sensor Networks: A Case Study of Data-intensive Applications upon Low-Cost Scalable Systems . The Journal of Supercomputing , 2013 , 66 (3 ): 1178 –1193 . 7 Wang L. , Fu C . Research Advances in Modern Cyberinfrastructure . Generation Comput , 2010 , 28 (2 ): 111 –112 . 10.1007/s00354-009-0077-9 8 Zhang W. , Wang L. , Liu D. , etc. Towards building a multi-datacenter infrastructure for massive remote sensing image processing . Concurrency and Computation: Practice and Experience , 2013 , 25 (12 ): 1798 –1812 . 10.1002/cpe.2966 9 Wang L. , Chen D. , Hu Y. , etc. Towards enabling Cyberinfrastructure as a Service in Clouds . Computers & Electrical Engineering , 2013 , 39 (1 ): 3 –14 . 10.1016/j.compeleceng.2012.05.001 10 Raymond C A conceptual interdisciplinary plug-and-play cyber securityframework In Kaur H & Tao X , editors, ICTs and the Millennium Development Goals A United Nations Perspective , pp. 81 –99 , New York, USA : Springer , 2014 . 11 Raymond Choo K . The cyber threat landscape: Challenges and future research directions . Computers & Security , 2011 , 30 (8 ): 719 –731 . 10.1016/j.cose.2011.08.004 12 Eisentrager K, Lauter K, Montgomery P. Fast elliptic curve arithmetic and improved Weil pairing evaluation. Proc. of the 2003 RSA conference on The cryptographers’ track, Berlin, Germany, pp. 343–354, 2003. 13 Raymond Choo K . High tech criminal threats to the national information infrastructure . Information Security Technology Report , 2010 , 15 (3 ): 104 –111 . 10.1016/j.istr.2009.09.001 14 Cornish P, Livingstone D, Clemente D, et al. Cyber Security and the UK’s Critical National Infrastructure. A Chatham House Report, 2011. 15 Yang Y, Lu J, Raymond C, et al. On Lightweight Security Enforcement in Cyber-physical Systems. In Proceedings of International Workshop on Lightweight Cryptography for Security & Privacy (LightSec 2015), Bochum, Germany, Lecture Notes in Computer Science, Springer-Verlag, 2015. 16 Dong M , Ota K , Laurence T , et al LSCD: A Low-Storage Clone Detection Protocol for Cyber-Physical Systems . IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems , 2016 , 35 (5 ):712 –723 . 10.1109/TCAD.2016.2539327 17 Dong M , Li H , Ota K , et al Rule caching in SDN-enabled mobile access networks . IEEE Network , 2015 , 29 (4 ):40 –45 . 10.1109/MNET.2015.7166189 18 Mao W . Modern Cryptography: Theory and Practice . Prentice Hall , 2003 19 Joye M. , Neven G . Identity-based Cryptography . IOS Press , 2008 20 Moody D. , Peralta R. , Perlner R. , etc. Report on Pairing-based Cryptography . Journal of Research of the National Institue of Standards and Technology , 2015 , 120 :11 –27 . 10.6028/jres.120.002 21 Rahuman A , Athisha G . Reconfigurable Architecture for Elliptic Curve Cryptography Using FPGA . Mathematical Problems in Engineering , 2013 , 2013 :1 –8 . 10.1155/2013/675161 22 Dai G , Wang M , Peng L , et al Implementation and Optimization for Tate pairing . Intelligent automation and soft computing , 2011 , 17 (5 ):607 –617 . 10.1080/10798587.2011.10643174 23 Miller V . The Weil pairing and its efficient calculation . Journal of Cryptology , 2004 , 17 (4 ):235 –261 . 10.1007/s00145-004-0315-8 24 Duursma I, Lee H S Tate Pairing Implementation for Hyperelliptic Curves y2 = xp − x + d. Proc. of Advances in Cryptology-Asiacrypt, 2003, Heibelberg,Germany, pp. 111–123. 25 Barreto P , Galbraith S , Eigeartaigh C , et al Efficient pairing computation on supersingular abelian varieties . Designs, Codes and Cryptography , 2007 , 42 (3 ): 239 –271 . 10.1007/s10623-006-9033-6 26 Barreto P , Galbraith S , Eigeartaigh C , et al An efficient key-policy attribute-based encryption scheme with constant ciphertext length . Mathematical Problems in Engineering , 2013 , 2013 :1 –7 . 27 Cocks C, Pinch R. Identity-based cryptosystems based on the Weil pairing. unpublished manuscript, 2001. 28 Fotiadis G, Konstantinou E. More Sparse Families of Pairing-Friendly Elliptic Curves. Proc. Proc. of 13th International Conference on Cryptology and Network Security, Heraklion, Greece, pp.384–399, 2014. 29 Barreto P, Naehrig M. Pairing-friendly elliptic curves of prime order. Proc. Of Selected Areas in Cryptography—12th International Workshop, Kingston, Canada, pp. 319–331, 2005. 30 Freeman D Constructing Pairing-Friendly Elliptic Curves with Embedding Degree 10. Proc. Of algorithmic number theory, Berlin, Germany, pp.452–465, 2006. 31 Miyaji A , Nakabayashi M , Takano S . New explicit conditions of elliptic curve traces for FR-reduction . IEICE Transactions on Fundamentals , 2001 , vol. E84-A , no.5 , pp.1234 –1243 . 32 Menezes A , Okamoto T , Vanstone S . Reducing elliptic curve logarithms to logarithms in a finite field . IEEE Transactions on Information Theory , 1993 , 39 (5 ): 1639 –1646 . 10.1109/18.259647 33 Brezing F , Weng A Elliptic curves suitable for pairing based cryptography . Designs, Codes, and Cryptography , 2005 , 37 (1 ):133 –141 . 10.1007/s10623-004-3808-4 34 Izuta T, Nogami Y, Morikawa Y. Ordinary Pairing Friendly Curve of Embedding Degree 1 Whose Order Has Two Large Prime Factors. Proc. of IEEE Region 10 Conference on TENCON 2010, Fukuoka, JAPAN, pp.769–772, 2010. 35 Lee H, Park C. Generating Pairing-Friendly Curves with the CM Equation of Degree 1. Proc. of 3rd International Conference on Pairing-Based Cryptography, Palo Alto, California, USA, pp.66–77, 2009. 36 Pollard J . A monte carlo method for factorization . BIT Numerical Mathematics , 1975 , 15 (3 ): 331 –334 . 10.1007/BF01933667 37 Frey G and Ruck H . A remark concerning m-divisibility and the discrete logarithm in the divisor class group of curves . Mathematics of Computation , 1994 , 62 (206 ): 865 –874 . 10.2307/2153546 38 Zhang F, Naini R and Susilo W. An efficient signature scheme from bilinear pairings and its application. Proc. Of Advances in Cryptography-PKC’04, Heidelberg, Germany, pp. 277–290, 2004. 39 Bertoni G, Chen L, Fragneto P, et al. Computing Tate Pairing on Smartcards. Proc. Of Proceedings of Workshop on Cryptographic Hardware and Embedded Systems, Edinburgh, Scotland, 2005. 40 IEEE-SA Standards Board. IEEE standard specifications for public-key cryptography, 2000. 41 Perera C , Ranjan R , Wang L , et al Big Data Privacy in the Internet of Things Era . IT Professional , 2015 , 17 (3 ): 32 –39 . 10.1109/MITP.2015.34 42 Kolodziej J , Khan S , Wang L , et al Security, energy, and performance-aware resource allocation mechanisms for computational grids . Future Generation Computer Systems , 2014 (31 ):77 –92 . 43 Wei J , Cai W , Wang L , et al A secure information service for monitoring large scale gridse . Parallel Computing , 2007 , 33 (7–8 ): 572 –591 .

PMC005xxxxxx/PMC5001718.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756457310.1371/journal.pone.0161797PONE-D-16-24537Research ArticleBiology and Life SciencesNeuroscienceBrain MappingFunctional Magnetic Resonance ImagingMedicine and Health SciencesDiagnostic MedicineDiagnostic RadiologyMagnetic Resonance ImagingFunctional Magnetic Resonance ImagingResearch and Analysis MethodsImaging TechniquesDiagnostic RadiologyMagnetic Resonance ImagingFunctional Magnetic Resonance ImagingMedicine and Health SciencesRadiology and ImagingDiagnostic RadiologyMagnetic Resonance ImagingFunctional Magnetic Resonance ImagingResearch and Analysis MethodsImaging TechniquesNeuroimagingFunctional Magnetic Resonance ImagingBiology and Life SciencesNeuroscienceNeuroimagingFunctional Magnetic Resonance ImagingBiology and Life SciencesNeuroscienceSensory PerceptionVisionBiology and Life SciencesPsychologySensory PerceptionVisionSocial SciencesPsychologySensory PerceptionVisionBiology and Life SciencesBehaviorBiology and Life SciencesPhysiologySensory PhysiologyVisual SystemEye MovementsMedicine and Health SciencesPhysiologySensory PhysiologyVisual SystemEye MovementsBiology and Life SciencesNeuroscienceSensory SystemsVisual SystemEye MovementsBiology and Life SciencesAnatomyHeadEyesMedicine and Health SciencesAnatomyHeadEyesBiology and Life SciencesAnatomyOcular SystemEyesMedicine and Health SciencesAnatomyOcular SystemEyesResearch and Analysis MethodsResearch AssessmentReproducibilityEngineering and TechnologySignal ProcessingSignal FilteringBiology and Life SciencesAnatomyBrainVisual CortexMedicine and Health SciencesAnatomyBrainVisual CortexInfluences of High-Level Features, Gaze, and Scene Transitions on the Reliability of BOLD Responses to Natural Movie Stimuli Sources of Reliable Cortical Response with Natural VisionLu Kun-Han 23Hung Shao-Chin 1Wen Haiguang 23Marussich Lauren 13Liu Zhongming 123*1 Weldon School of Biomedical Engineering, Purdue University, West Lafayette, IN, United States of America2 School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, United States of America3 Purdue Institute for Integrative Neuroscience, Purdue University, West Lafayette, IN, United States of AmericaZhang Nanyin EditorPenn State University, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: KHL SCH ZML. Data curation: KHL SCH LM. Formal analysis: KHL HW. Funding acquisition: ZML. Investigation: KHL SCH ZML. Methodology: KHL HW. Project administration: KHL ZML. Software: KHL ZML. Supervision: ZML. Validation: KHL ZML. Visualization: KHL. Writing – original draft: KHL ZML. Writing – review & editing: KHL ZML. * E-mail: zmliu@purdue.edu26 8 2016 2016 11 8 e016179718 6 2016 11 8 2016 © 2016 Lu et al2016Lu et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Complex, sustained, dynamic, and naturalistic visual stimulation can evoke distributed brain activities that are highly reproducible within and across individuals. However, the precise origins of such reproducible responses remain incompletely understood. Here, we employed concurrent functional magnetic resonance imaging (fMRI) and eye tracking to investigate the experimental and behavioral factors that influence fMRI activity and its intra- and inter-subject reproducibility during repeated movie stimuli. We found that widely distributed and highly reproducible fMRI responses were attributed primarily to the high-level natural content in the movie. In the absence of such natural content, low-level visual features alone in a spatiotemporally scrambled control stimulus evoked significantly reduced degree and extent of reproducible responses, which were mostly confined to the primary visual cortex (V1). We also found that the varying gaze behavior affected the cortical response at the peripheral part of V1 and in the oculomotor network, with minor effects on the response reproducibility over the extrastriate visual areas. Lastly, scene transitions in the movie stimulus due to film editing partly caused the reproducible fMRI responses at widespread cortical areas, especially along the ventral visual pathway. Therefore, the naturalistic nature of a movie stimulus is necessary for driving highly reliable visual activations. In a movie-stimulation paradigm, scene transitions and individuals’ gaze behavior should be taken as potential confounding factors in order to properly interpret cortical activity that supports natural vision. http://dx.doi.org/10.13039/100000025National Institute of Mental HealthR01MH104402Liu Zhongming This work was supported by NIH R01MH104402, National Institute of Mental Health, to ZL. Data AvailabilityAll relevant data are within the paper.Data Availability All relevant data are within the paper. ==== Body Introduction Conventional imaging studies in visual neuroscience focus on mapping functional brain regions that process specific visual information such as location, motion, and color [1]. These studies often utilize highly controlled experimental paradigms with artificial or static stimuli, which are, however, too narrowly focused to reflect real-life visual experiences. Naturalistic visual environments are dynamic and complex, containing numerous visual features entangled in space and time [2]. Nevertheless, humans are able to readily and actively explore the surroundings, through neural network activities that give rise to an efficient and effective interplay of visual processing, attention, and behavior (e.g. gaze) [3]. To understand the network basis of this interplay, it is desirable to use natural-vision paradigm for functional neuroimaging, such that brain activity can be recorded and mapped while human subjects are engaging themselves in dynamic and realistic visual environments. Some recent studies have begun to characterize brain activity during naturalistic movie stimulation [4–6]. Initial findings demonstrate that extended cortical activities are highly reproducible within and between subjects freely watching a commercial movie [5,6]. Depending on the movie stimulus, cortical areas showing reproducible activity may cover the entire visual cortex, as well as attention and oculomotor networks [3]. Such intra- and inter-subject reproducibility has been consistently observed with functional magnetic resonance imaging (fMRI) [5,7], magnetoencephalography [8,9], electrocorticography [10,11], neuronal spike trains and local field potentials [12]. These findings support a central notion that high reproducibility is a robust and perhaps fundamental characteristic of brain activity across a wide range of spatial and temporal scales during naturalistic visual experiences. This notion further allows one to map neural substrates of natural vision by simply assessing the reproducibility of brain activity measured with fMRI [6] or other neuroimaging modalities. However, the origins of reproducible neural responses to naturalistic visual stimulation are incompletely understood [2]. In particular, it is important to identify and separate the confounding factors that may drive the neural response reproducibility, but bearing little or no relevance to the natural visual content. Specifically, movies possess low-level visual features that are statistically irregular in space and time, and high-level features that drive perception and cognition. To what degree do the high- or low-level features drive the reproducible brain response remains elusive. Russ and Leopold have shown that reproducibility in cortical activity is explained by both low-level (e.g. contrast and luminance) and high-level features (e.g. faces, animals, and biological motion) [2]. Hasson et al. have shown that disrupting the temporal structure of a movie would reduce the reproducibility of cortical responses [13]. As such, the high-level features resulting from integrating information in time may drive, at least in part, the reliable cortical activity. However, it is unknown whether and where cortical response reproducibility may still remain, in the absence of any high-level features when both spatial and temporal structures are disrupted to render a movie stimulus perceptually meaningless. In addition, commercial movies typically contain instantaneous scene transitions due to film editing [14]. The scene transitions may appear as transient stimulus events that may induce reproducible responses while being unrelated or loosely related to the natural context. Moreover, when subjects are freely engaged to a movie stimulus, they may receive inconsistent retinal inputs that depend on their individual gaze behaviors [3]. The effects of the varying gaze behavior on brain activity and its reproducibility also remains to be addressed. In this study, we aimed to address these questions by using data collected with concurrent fMRI and eye tracking from human subjects receiving repeated movie stimuli. Fourier shuffling was applied to the movie to create a spatiotemporally scrambled control stimulus such that the low-level visual features were dissociated from the high-level natural content in the intact movie. By comparing the reproducibility in fMRI activity during the scrambled vs. intact stimuli, we attempted to disentangle the individual contributions of the low and high-level features to the reproducibility of the movie evoked response. We also assessed the contributions of individuals’ gaze behavior to fMRI activity and its reproducibility within and across subjects, and further evaluated the extent to which the reproducibility in fMRI activity was confounded by scene transitions in the movie. Results obtained with the above analyses are expected to help refine the design, analysis, and interpretation of present and future natural vision imaging studies. Materials and Methods Subjects Sixteen human subjects (10 women, ages 20−31) participated in this study in accordance with a research protocol approved by the Institutional Review Board at Purdue University. All subjects were healthy volunteers with normal vision. Informed written consent was obtained from every subject. Among the 16 subjects, three subjects’ data were excluded from the subsequent analyses either because they fell asleep with eyes closed or had excessive head movement. Stimuli An uninterrupted segment of movie (5:37 minutes in length, from 163:06 to 168:50 minutes in The Good, the Bad, and the Ugly, directed by Sergio Leone) was converted to gray scale to provide a naturalistic visual stimulus. From this movie, a scrambled movie was further created to keep the same statistical distribution of pixel intensity as the intact movie, yet rendering itself perceptually meaningless. Specifically, we used 3D Fast Fourier Transform (FFT), as implemented by the function ‘fftn’ in MATLAB, to convert the intact movie into the frequency domain, in which the phase was randomly shuffled while the magnitude remained unchanged; then inverse FFT was applied to the phase-shuffled data to generate a scrambled version of the original movie. As illustrated in Fig 1, the intact and scrambled movies were matched in low visual features, whereas the contrast between them was specific to the presence vs. absence of high-level natural content independent of low-level visual properties. 10.1371/journal.pone.0161797.g001Fig 1 Illustration of the naturalistic vs. scrambled visual stimuli. Example frames from the intact movie (A) and the scrambled movie (B). The scrambled movie was created by shuffling the phase of the intact movie in the frequency domain (C). Paradigm Among the 13 subjects, seven underwent four sessions of fMRI acquisition with visual stimulation: two sessions for the intact movie, and two sessions for the scrambled movie. These subjects were instructed to freely view the movie stimuli. Four subjects underwent two repeated fMRI sessions for the intact movie stimulation, while they were instructed to fixate at a cross-hair (0.8 × 0.8 degrees in width and height) at the screen center. Two subjects underwent eight sessions of fMRI acquisition: two sessions for the intact movie (free-viewing), two sessions for the intact movie (fixation), two sessions for the scrambled movie (free-viewing), and two sessions for the scrambled movie (fixation). Every stimulation session started with a blank gray screen presented for 42 seconds, followed by the intact or scrambled movie presented for 5 minutes and 37 seconds, ended with the blank screen again for 30 seconds. No sound was played during the movie. The order of the above stimulation sessions was randomized and counterbalanced across subjects. MRI All experiments were conducted in a 3T MRI system (Signa HDx, General Electric, Milwaukee). A 16-channel receive-only surface phase-array coil (NOVA Medical, Wilmington) was used throughout every experiment. T1-weighted anatomical images were acquired with a spoiled gradient recalled acquisition (SPGR) sequence (256 sagittal slices with 1 mm thickness and 1×1 mm2 in-plane resolution, TR/TE = 5.7/2ms, flip angle: 12°). FMRI data were acquired with the standard single-shot, gradient-recalled (GRE) echo-planar imaging (EPI) sequence (38 interleaved axial slices with 3.5 mm thickness and 3.5 × 3.5 mm2 in-plane resolution, TR = 2000ms, TE = 35ms, flip angle = 78°, field of view = 22 × 22 cm2, in-plane acceleration by a factor of two based on sensitivity encoding (SENSE), no acceleration in the slice direction). Eye-Tracking The visual input was presented using the MATLAB-based Psychophysics Toolbox [15,16], and they were delivered to the subjects through a binocular goggle system (NordicNeuroLab, Norway) mounted on the head coil. The display resolution was 800 × 600. An MR-compatible monocular eye-tracking camera was integrated into the goggle system to monitor the left eye under infrared illumination. The eye-camera data were transmitted to an eye-tracking system (ViewPoint, Arrington Research, USA), which captured the eye movement and tracked the subject’s gaze location at 30 Hz during the stimulation session. To ensure reliable and accurate eye tracking, the system was recalibrated prior to every stimulation session. Through the goggle system, the visual field covered by the movie was about 26.9-by-20.3 degrees. When the subjects were instructed to fixate at the screen center, the fixation was not ensured by any fixation task, but confirmed retrospectively with eye-tracking data. Pre-processing MRI and fMRI data were preprocessed by using FSL [17] and AFNI [18]. Briefly, T1-weighted anatomical images were non-linearly registered to the MNI brain template. T2*-weighted functional image series were corrected for slice timing, registered to the first volume within each series to account for head motion, masked out non-brain tissues, aligned to the T1-weighted structural MRI, registered to the MNI template and resampled into 3×3×3mm3 voxels. After discarding the first six volumes, the fMRI data were temporally de-trended by using a third-order polynomial function to model the slow signal drift, and spatially smoothed by using a 3-D Gaussian filter with 6mm full width at half maximum (FWHM). As the focus of this study was on brain activity during sustained and dynamic movie stimuli, we only used the data from 12 seconds into the movie to the end of the movie. Data from other periods were all excluded from subsequent analyses to avoid the transient fMRI response shortly after the gray screen was replaced by the movie presentation. Inter- and intra-subject reproducibility in fMRI activity As previously published elsewhere [5], the map of visually evoked activations was obtained by assessing the zero-lag Pearson’s cross correlation in every voxel’s time series between two repeated presentations of the identical movie for the same or different subjects. Specifically, the intra-subject correlation was measured for every voxel and each subject, whereas the inter-subject correlation was measured for every voxel and each pair of different subjects. The intra- and inter-subject correlation coefficients were converted to z scores through the Fisher’s r-to-z transform with the degree of freedom set as the number of time points minus one. Further, the intra- or inter-subject reproducibility was tested for statistical significance in the group level. The intra- or inter-subject reproducibility and its significance were evaluated separately for the intact or scrambled movie in a free-viewing or fixation condition. For the intra-subject reproducibility, the z-transformed correlation coefficient (or z score) between sessions from the same subject was averaged across subjects; the voxel-wise statistical significance was evaluated by using one-sample t-test, with p < 0.0033, DOF = 8, multiple comparison corrected by controlling the false discovery rate (FDR) < 0.03 [19] (FDR is also denoted as q in this paper). Note that the same method for the correction for multiple comparison was used for all other statistical tests in the study, unless it was described otherwise. For the inter-subject reproducibility, the voxel-wise cross correlation was obtained for each pair of different subjects. The pair-wise correlations were not all independent because each subject could be involved in more than one subject pair. This lack of independence invalidated the use of parametric statistical tests, such as the above-mentioned one-sample t test. Instead, we used a non-parametric resampling based statistical inference as implemented in ISC-toolbox (www.nitrc.org/projects/isc-toolbox/) and described in details elsewhere [20]. Briefly, an empirical distribution was obtained for inter-subject cross-correlations given the null hypothesis that such correlations were trivial and non-significant. To do so, every subject’s signals were circularly shifted in time by a random and different amount so that the signals from different subjects were no longer aligned in time. Following the temporal resampling, cross correlations were computed between the resampled time series at the same or different voxels from different subjects, effectively yielding a null resampling distribution with 10 million samples of trivial inter-subject cross correlations. This resampling-based statistical inference also took into account the auto-correlated nature of the fMRI signal. Without resampling, all inter-subject cross correlations were tested against this empirical null distribution, yielding the p values for significance test, corrected for multiple comparison with FDR < 0.03. Note that either physiological (respiratory or cardiac) fluctuations or head motion parameters were not regressed out from the fMRI signals prior to the calculation of the intra- and inter-subject reproducibility, because such task-unrelated fluctuations were not expected to contribute to the cross correlation between signals from different fMRI sessions. Assessing the effects of gaze behavior In the natural vision paradigm, the interpretation of reproducible fMRI activity in terms of fluctuating visual features may be of concern due to the presence of multiple confounding factors that may contribute to the fMRI signal, and drive or disrupt the signal reproducibility. We first addressed the degree to which fMRI activity and its reliability could be accounted for by the subjects’ varying gaze behavior monitored with eye tracking. Specifically, every subject’s gaze locations were estimated from instantaneously captured eye images. The time series of the horizontal (x) and vertical (y) variations were first filtered by moving average within a 1s window, in order to minimize the influence of the extreme values caused by eye blinking. After removing the mean from the x and y time series, we characterized the gaze behavior by calculating the instantaneous saccadic amplitude (in degrees), which was the spanning visual angle between two consecutive gaze locations based on a gaze displacement model, as described elsewhere [21,22]. The saccadic amplitude signal was further demeaned, and was convolved with the canonical hemodynamic response function (HRF), as implemented in SPM (www.fil.ion.ucl.ac.uk/spm/). The result were further resampled to match the fMRI sampling times after applying an anti-aliasing low-pass filter with the cutoff frequency at 0.25 Hz. As such, a single regressor was derived from the time-varying saccade amplitude as an explanatory variable for the fMRI signal. The zero-lag cross correlation between the fMRI signal and the saccade-amplitude regressor was evaluated for every voxel. The correlation coefficients were transformed to the z scores, and tested across subjects for significance by using one-sample t-test (p < 0.011, DOF = 17, FDR < 0.03). To assess the gaze effects on the reproducibility of fMRI activity, the gaze-related regressors were excluded from each voxel time series by using a linear nuisance variable regression (NVR). The difference in the intra-subject reproducibility with vs. without the NVR was tested for significance by using paired t-test (p < 0.03, DOF = 8, uncorrected). Alternatively, the effects of gaze behavior were addressed by comparing the intra-subject reproducibility of fMRI activity when the subjects were freely watching the movie (n = 9) versus watching the same movie but with their eyes fixated (n = 6). The latter served as a control experiment without gaze behavioral variation either within or across subjects. Between these two conditions, the voxel-wise difference in the z-transformed intra-subject correlation was tested for statistical significance by using two-sample t-test (p < 0.03, DOF = 13, uncorrected). Assessing the effects of scene transitions We further addressed the effects of scene transitions due to film editing. The scene transitions manifested themselves as transient and discontinuous changes in visual input. The occurrence times of scene transitions were first detected if the sum of the absolute pixel difference between two consecutive movie frames exceeded a given threshold, and then visually affirmed. A train of binary events was defined as a time series of ones and zeroes, with one indicating the occurrence of a transition event. The binary events were further weighted by the estimated motion between two consecutive movie frames at the transition time based on a block-matching algorithm implemented as a built-in function (vision.BlockMatcher) in MATLAB (the block size = 9) [23]. It measures the movement fields in pixel blocks between two images, regardless of the perceptual meanings of these images. The motion-weighted transition events was further convolved with the HRF, and then filtered and resampled to construct a regressor for the fMRI signal. We used this regressor to evaluate the effects of scene transitions on fMRI activity (p < 0.0017, DOF = 17, FDR < 0.03) and its reproducibility (p < 0.001, DOF = 8, FDR < 0.03) using the similar correlation and regression analyses as used for addressing the effects of gaze behavior. While the above analyses were done with the volumetric data, the results were displayed on both the MNI volume template and the cortical surface templates [24]. Results FMRI reproducibility reveals visually evoked activations We used a long, continuous, dynamic, and complex movie to investigate the neural substrates underlying natural vision. For such a stimulus paradigm, it was difficult to define accurate response predictors required for the conventional fMRI analysis. Using the method proposed by Hasson et al. [5], we localized the visually evoked responses to areas where the fMRI time series was cross correlated between two repeated presentations of the identical movie for the same (intra-subject) or different (inter-subject) subjects. The intra- and inter-subject reproducibility in fMRI activity was statistically evaluated in both individual and group levels to yield the activation maps induced by the movie stimulation. The intact movie gave rise to significant intra- and inter-subject reproducibility in fMRI activity at extended occipital areas that covered almost the entire visual cortex (Fig 2). The reproducibility was stronger within subjects (Fig 2A) than across subjects (Fig 2B), whereas the inter-subject reproducibility was significant even at lateral geniculate nuclei (LGN), and also extended to the parietal and frontal lobes (Fig 2B). These results corroborate the previous finding that naturalistic visual stimulation induces highly reproducible and widely extended cortical activity [5,6], and further extend this finding by showing reproducible thalamic responses during free watching of a movie. 10.1371/journal.pone.0161797.g002Fig 2 Brain activations with the intact movie were found at regions that showed significant intra-subject (A) and inter-subject (B) correlations in cortical activity during free movie watching. The mapping results were based on data from nine subjects. From a single subject, the fMRI signals from two voxels within the primary visual cortex (V1) and the lateral occipito-temporal gyrus (V4) are shown as examples to illustrate the intra-subject reproducibility in cortical activity. The color indicates the cross correlation. Spatiotemporal scrambling eliminates reproducible brain activity Next we asked whether highly reproducible cortical activity was unique to natural vision, or it might also result from irregularly fluctuating visual stimulation that carried no naturalistic content in every movie frame. To address this question, we scrambled the movie to render it perceptually meaningless while preserving low-level visual features, such as luminance and spatial/temporal frequency. When subjects freely watched the scrambled movie, the intra-subject reproducibility of the fMRI signal was not significant anywhere in the brain (Fig 3A), while the inter-subject reproducibility was barely significant only within the primary visual cortex (V1) (Fig 3B). 10.1371/journal.pone.0161797.g003Fig 3 Cortical activations with the scrambled movie were reduced and confined to V1, as revealed by the intra-subject (A) and inter-subject (B) reproducibility of the fMRI signal. The mapping results were based on data from nine subjects. From a single subject, the fMRI signals from two voxels within the primary visual cortex (V1) and the lateral occipito-temporal gyrus (V4) are shown as examples to illustrate the relatively low intra-subject reproducibility in cortical activity. The color indicates the cross correlation. We further questioned whether the lack of reproducible activity during the scrambled movie might be attributed to variable gaze trajectories within and across subjects. The absence of natural content could affect the subjects’ involuntary attention and cause different gaze patterns during the two repetitions of the scrambled movie. Indeed, the intra-subject cross correlation in the gaze trajectory was significantly less (p<0.03) for the scrambled movie than for the intact movie (Fig 4A). However, the variable gaze pattern could not entirely account for the reduced degree and extent of reliable cortical responses, because the reproducibility in the fMRI signal was confined to V1 even when eyes were fixated at the screen center (Fig 4C) compared to the free-viewing condition (Fig 4B). 10.1371/journal.pone.0161797.g004Fig 4 Effects of varying gaze behavior on cortical reliability when watching the scrambled movie. (A) For most subjects (except one), the horizontal and vertical variations of the gaze location were more consistent for the intact (non-scrambled) movie than for the scrambled movie. Error bars represent the standard deviation across subjects. (B) For an example subject, when the subject did not fixate at the screen center, there was a lack of reliable responses (C) The scrambled movie induced reliable responses within V1 when they fixated at the screen center. In both (B) and (C), the color indicates the intra-subject cross correlation in voxel time series (*: p < 0.0005, **: p < 0.05, ***: p < 0.03). These results were based on eye-tracking data from nine subjects who freely watched both the intact and scrambled movies. Therefore, highly reproducible and widely extended fMRI response to a continuous, complex, and dynamic visual stimulus requires the stimulus to be naturalistic, whereas a non-natural stimulus with matched low-level visual properties induces a much less degree and extent of time-locked and reproducible cortical activity. Gaze behavior affects oculomotor activity during natural vision The changing scene content in a naturalistic movie may drive varying gaze behavior as monitored with eye tracking. The varying gaze location, in turn, may alter the retinal input and thus the induced cortical activity. To assess the dependence of cortical fMRI reliability on gaze behavior during natural vision, we measured the zero-lag cross correlation between the fMRI signal and the fluctuation of the saccade amplitude derived from instantaneous eye movements. We found that the saccade-amplitude fluctuation was significantly correlated with the fMRI activity notably in the part of the V1 that represented the peripheral visual field, as well as in other cortical areas along both dorsal and ventral streams, including intra-parietal sulcus (IPS), frontal eye fields (FEF), supplementary eye fields (SEF), dorsal-lateral prefrontal cortex (DLPFC), and insula (Fig 5A). 10.1371/journal.pone.0161797.g005Fig 5 Varying gaze behavior contributed to fMRI activity and its reproducibility during natural movie stimulation. (A) Cross correlation between the fMRI signal and the time-varying saccade amplitude during free movie viewing (p < 0.011, FDR < 0.03). (B) The upper panel shows the difference (without regression minus with regression) in intra-subject correlation when taking the four gaze behaviors as the nuisance variables (p < 0.03, uncorrected). The lower panel shows the difference in intra-subject correlation when taking free-viewing the intact movie as the experimental group (n = 9) while taking the eyes-fixated movie watching (n = 6) as the control group (p < 0.03, uncorrected). The color shows the difference in cross correlation following the t-test. The gaze behavior was found to vary not only across sessions and subjects (Fig 4A), but also within a session, being more consistent for movie frames showing salient objects or human faces. This variation could affect the level and extent of activity reproducibility when the subjects were freely watching the movie. To assess this effect, we compared the intra-subject reproducibility in the fMRI signal with and without regressing out the individual variation in gaze movement. The contrast (without regression minus with regression) was significant primarily in peripheral V1 (Fig 5B, top panel). To further address the gaze-related effects, we compared the intra-subject reproducibility in fMRI signals between the free-viewing (n = 9) and fixation conditions (n = 6), while the gaze behavior varied within and across subjects in the former condition but not in the latter. The contrast map (free-viewing minus fixation) between these two conditions showed localized gaze-related effects, qualitatively similar to those observed with the nuisance variable regression (Fig 5B, bottom panel); but it revealed more areas, including IPS and DLPFC, in addition to peripheral V1. Therefore, When free viewing a movie, the gaze behavior affects brain activity primarily in peripheral V1, as well as in cortical regions likely involved in controlling or responding to saccades [25,26]. However, the gaze behavior does not significantly confound or explain widely distributed cortical activations reliably induced by the naturalistic stimulation. Scene-transition events induce reliable responses in visual cortex Next, we examined to what degree brain activity was driven by scene transitions in the movie, which reflected the transient change in visual input (Fig 6A). The scene transitions were found to contribute to cortical responses mostly at areas along the ventral visual stream (Fig 6C), where the fMRI signal was significantly correlated with a modeled response based solely on a train of discrete scene-transition events weighted by the total motion between consecutive movie frames at each transition time (Fig 6B). In addition, these scene transitions also confounded the intra-subject reproducibility in cortical fMRI activity during the movie. Regressing out the contributions from scene transitions to the fMRI signal significantly reduced (p < 0.001, FDR < 0.03), but did not eliminate, the reproducibility of cortical activity at distributed extrastriate cortical areas, including those along the ventral pathway (Fig 6D). 10.1371/journal.pone.0161797.g006Fig 6 Dependence of cortical fMRI activity and its reproducibility on scene transitions during natural vision. (A) A scene transition (video shot boundary) occurred as visual input changed abruptly due to film editing. (B) A train of scene-transition events were defined by the timing and degree of the transition evaluated by the total motion change between two adjacent movie frames. (C) Cross correlations were calculated between the fMRI signals and the regressor derived from the scene-transition events (p < 0.0017, DOF = 17, FDR < 0.03). Representative time courses from one subject are shown to demonstrate the similarity between the scene-transition predicted response and the fMRI signal from a voxel in the ventral visual pathway (D) Comparison between the intra-subject reproducibility without (left) vs. with (right) regressing out the scene-transition induced response. Their difference (bottom) was evaluated and shown to depict the effects of scene transitions on the reproducibility of cortical activity during natural vision. The color shows the correlation coefficient or the difference in correlation coefficient, following the t-test (p < 0.001, FDR < 0.03). Therefore, the scene transitions in commercial movies partly drive the extrastriate cortical activity, confounding the assessment and interpretation of the reproducibility in fMRI activity during naturalistic visual stimulation. Discussion In this study, we used concurrent fMRI and eye tracking to assess the degree and extent to which reproducible brain activity could be affected by various experimental and behavioral factors during free viewing of sustained, dynamic, and complex movie stimuli. Results demonstrate that the varying gaze behavior affected the reproducibility of fMRI activity in the peripheral V1, and that the scene transitions in the movie affected the reproducibility in the areas along the ventral visual pathway. For these areas, mathematically removing the confounding contributions of these factors from fMRI signals led to significantly less (but not diminished) reproducibility in cortical activity. In addition, we found that the highly reliable and widely extended activations evoked by a movie was mainly due to the naturalistic content of the movie. Removing the perceptually meaningful content while keeping the low-level visual properties largely reduced the degree and extent of the evoked cortical response. Taken together, measuring and mapping the reproducibility of fMRI activity constitutes a robust functional imaging technique to uncover neural processes responsible for visual processing and perception in a dynamic, complex, and realistic context. However, it is worth taking into account (more importantly) scene transitions and (less importantly) individuals’ gaze behavior as confounding factors. Reliable visually evoked activations arise from naturalistic content In the seminal work by Hasson et al., an important finding was the functional selectivity of cortical activity during natural movie stimuli [5]. For example, face or building related regions exhibited dynamic cortical fluctuations with peak activity selectively occurring when various forms of face or building appeared in a movie (see Fig 3 in Hasson et al., 2004). This finding pointed to a possibly exciting opportunity to infer natural and abstract visual content from dynamic cortical activity of the brain in action [6]. However, natural vision paradigms with movie stimuli are different from conventional fMRI paradigms in that fluctuations and patterns of cortical activity are not analyzed by contrasting the target vs. control conditions. The lack of a control condition requires more systematic investigations on various sources that modulate cortical activity, in order to disentangle the sources of interest for visual perception (e.g. house, building) as well as those unrelated or loosely related to perception (e.g. scene transitions, eye movements). Here, we asked whether trivial movie stimuli in the absence of perceptually meaningful content would also induce reliable cortical activity to confound the interpretation of activity fluctuations in terms of whether, what, and how natural information is processed and perceived by human subjects. Previous studies have reported that individual cortical neurons could fire reliably with precise spike timing given irregularly fluctuating or transient inputs, but not with constant or steady state stimuli [27,28]. These findings suggest the intrinsic reliability and precision of neural coding may depend on the statistical irregularity (or complexity) of the input information [27]. Extending this coding characteristic from single neurons to distributed cortical networks, a valid alternative hypothesis is that reliable cortical activity may be primarily attributed to the irregular visual patterns of the movie stimulus, as opposed to its naturalistic nature. Although natural vision necessarily entails irregular and dynamic visual patterns, spatiotemporally complex visual inputs do not always convey natural content. Therefore, it is unclear whether and to what extent the naturalistic nature of the visual stimulus is necessary for the highly reliable and widely distributed neural responses during sustained, complex and dynamic visual stimuli. To address this question, we separated the irregular vs. naturalistic property of a complex movie stimulus by scrambling the naturalistic movie stimulus to dissociate the low-level irregular features from the high-level naturalistic features (Fig 1). Results indicate that the evoked cortical activations, in terms of intra- and inter-subject reproducibility of fMRI activity, were largely eliminated by the absence of the natural content when subjects freely watched the scrambled movie, with barely significant residual activations mostly confined to V1 (Fig 3). This finding suggests that the reliable cortical activation with natural vision mainly results from the natural content in the movie, rather than its spatiotemporal irregularity and complexity. In real life, the natural content is embedded into spatial and temporal structures of visual elements. Disruption in either or both of these structures compromises, to a varying degree, the natural information. In previous studies, researchers scrambled movie frames in time to disrupt the temporal structure within a varying period [13]. When the temporal structure was disrupted, early visual areas still exhibited highly reliable responses, whereas the response reliability at higher visual areas was reduced to a varying degree depending on the length of the time-scrambled window [13]. This important finding has led to the discovery of a hierarchical organization of the temporal receptive field, similar to that of the spatial receptive field [13]. In line with (but different from) these studies, we scrambled the visual elements not only in time but also in space by using phase-shuffling. Such a scrambled movie disrupted both spatial and temporal structures of a natural movie stimulus, such that movie frames did not appear natural either individually or collectively. This complete absence of naturalistic information was different from time-scrambled movies, for which natural content was still preserved in individual movie frames, or connected in a longer (non-scrambled) time scale. In the present study, the lack of reliable cortical response to spatiotemporally scrambled movie stimuli extends the previous findings, and further indicates the importance of the natural content in driving highly reproducible and widely extended brain activity under movie stimulation. It might be expected that the scrambled movie did not evoke reliable cortical responses at higher visual areas, as these areas have been conventionally considered to be responsive to high-level visual features. However, the lack of reliable responses to the scrambled movie even at early visual areas came as a surprise. Neurons in lower visual areas (e.g. V1, V2) have small receptive fields and are considered to respond to low-level features of stimuli [29], whereas neurons in higher cortical areas are responsible for increasingly complex and abstract features [30,31]. Neuronal connections to early visual areas arise in part from retinal inputs through bottom-up or feed-forward pathways, and also in part from higher visual areas through top-down or feedback pathways. Both bottom-up and top-down connections interplay through a complex network, with feedback inputs shape or modulate the processing of feed-forward visual information, e.g. attention-modulated visual processing [32]. Therefore, we speculate that the phase-shuffling of the movie eliminates the reliable response at higher visual areas, and results in the loss of reliable feedback control over early visual areas. Such feedback inputs likely serve as important drivers or modulators of reliable slowly fluctuating cortical responses observed with fMRI [33,34]. This interpretation is speculative, and awaits future studies to address in various spatial (network vs. synaptic levels) and temporal (seconds vs. milliseconds) scales with brain imaging and neural recording measurements. The dense inter-connections between lower and higher visual areas makes it complicated to argue for any specific source of reliable activity during natural vision. The reliability of cortical responses at the lower and higher visual areas cannot be simply dissociated. The response reliability at higher visual areas requires their input areas to also exhibit reliable responses. Although our results highlight the essential role of natural content in driving reliable cortical fMRI activity, it does not imply that high-level visual features in natural movie stimuli account entirely for the intra- and inter-subject reproducible fMRI responses at any visual area. The scenario of hierarchical processing is necessary to support progressive and selective processing of various visual features underlying natural vision. While low-level visual features can be isolated from high-level features, it is not possible vice versa. More systematic understanding of the distributing visual processing should perhaps benefit from network analysis, as opposed to an over-simplified dichotomy of lower vs. higher level processing, or region-specific assignment of visual functions. Scene transitions contribute to fMRI activity and its reproducibility With natural movie paradigms, previous studies emphasized the functional selectivity of reproducible activity at individual visual areas [5,6]. As supporting evidence for this functional selectivity, peak activity was retrospectively linked to the occurrence of specific objects or actions in the movie stimuli. Results from this study call for cautions in such analyses or interpretation. Scene transitions in almost all commercial movies occur as discrete events temporally superimposed with the continuous and complex movie stimulation. Previous studies also show that brain activity is, at least in part, time-locked to discrete transition events in movies [14], and such events cause a greater degree and extent of inter-subject correlation (ISC) in movie-stimulus evoked activity than those with single-shot, unstructured movies containing no scene transitions [3]. Therefore, it is likely of concern that scene transition events may evoke event-related responses in activity time series, to confound or mislead the functional interpretation of regional activity. Using the regression based analysis, we found that scene transitions contributed to fMRI activity at extended cortical areas, most notably along the ventral pathway, e.g. V4 and ventral occipital (VO) areas, and less notably along the dorsal pathway, e.g. V3A/B, V7, intra-parietal sulcus (IPS) (Fig 5C). Moreover, the transition related activity partly accounted for the observed reproducibility of fMRI activity at these areas during natural movie stimulation (Fig 5D). Since scene transitions contribute to fMRI activity and may confound the reproducibility of activity along the ventral visual stream, which is widely thought to be responsible for object recognition, one should perhaps be cautious interpreting activity time series at ventral visual areas. For example, the lateral-occipital complex (LOC) plays an important role in human object and face recognition and has been shown to respond more strongly to pictures of objects than to their scrambled images [35,36]. The ventral occipital-temporal cortex (VOT) consists of the fusiform face area (FFA) and the collateral sulcus (CoS) that are thought to be responsible for face and building recognition, respectively. The peak activity at these areas may indicate the occurrence of their encoded objects, or likely result from scene transitions. These two alternative possibilities should be considered for the interpretation of previous and future natural vision experiments. These two possibilities are likely mixed but not mutually exclusive. Russ et al. showed that scene cuts (or transitions) did not account for significant variance in the fMRI signal [2]. Huth et al., showed that semantic information was encoded in the fMRI voxel time series at the ventral visual areas, even after regressing out the motion energy, which likely included scene transitions [37]. Therefore, it is plausible to retrieve rich semantic information from fMRI signals during movie watching, despite the confounding effects from scene transitions. Effects of gaze behavior on fMRI activity and reproducibility In response to an identical movie stimulus, the gaze behavior may vary within and across subjects. As the retinotopic mapping is with respect to the center of the visual field defined by the gaze position, we expected the varying gaze behavior to affect the fMRI response at all or most retinotopic areas, as well as those areas involved in eye movement. In line with our expectation, the gaze behavior, described as the time-varying saccade amplitude, affected widely-spread brain regions including retinotopic areas as peripheral V1 and LGN, and specific regions in the oculomotor network including superior colliculus, parietal eye fields, frontal eye fields, supplementary eye fields and dorsal-lateral prefrontal cortex (Fig 5A), which are all related to controlling eye-movement and/or attention, as previously reported [38]. However, to our surprise, neither mathematically removing our gaze variation from fMRI time series nor experimentally controlling the subject’s eye movements (with the fixation) showed dramatic and wide-spread reduction in intra-subject correlation (Fig 4B and 4C). The gaze effects only influenced mainly in peripheral V1 with the mathematical regression and influenced a limited set of regions including peripheral part of V1, V3 and IPS with the fixation control. The limited gaze-related effects were perhaps partially attributable to the sub-optimal model used to describe the relationship between gaze movement and the fMRI signal. As this relationship was not fully understood and could likely be better described by other linear or non-linear models, the gaze-related effects may be under-estimated. Since we did not explore all possible models, we cannot rule out the possibility that eye movement might have stronger effects than were observed and reported in this paper. Despite this limitation in modeling, the contrast between free viewing and fixation conditions also revealed limited effects of the varying gaze behavior. ==== Refs References 1 Hadjikhani N , Liu AK , Dale AM , Cavanagh P , Tootell RB . Retinotopy and color sensitivity in human visual cortical area V8 . Nat Neurosci . 1998 7 ;1 (3 ):235 –41 . 10195149 2 Russ BE , Leopold DA . NeuroImage Functional MRI mapping of dynamic visual features during natural viewing in the macaque . Neuroimage . Elsevier B.V.; 2015 ;109 :84 –94 . 10.1016/j.neuroimage.2015.01.012 25579448 3 Hasson U , Landesman O , Knappmeyer B , Vallines I , Rubin N , Heeger DJ . Neurocinematics: The Neuroscience of Film . Projections . 2008 ;2 (1 ):1 –26 . 4 Bartels A , Zeki S . Functional Brain Mapping during Free Viewing of Natural Scenes . Hum Brain Mapp . 2004 ;21 (2 ):75 –85 . 14755595 5 Hasson U , Nir Y , Levy I , Fuhrmann G , Malach R . Intersubject Synchronization of Cortical Activity During Natural Vision . Science (80-) . 2004 ;303 (MARCH ):1634 –40 . 6 Hasson U , Malach R , Heeger DJ . Reliability of cortical activity during natural stimulation . Trends Cogn Sci . 2010 1 ;14 (1 ):40 –8 . 10.1016/j.tics.2009.10.011 20004608 7 Jääskeläinen IP , Koskentalo K , Balk MH , Autti T , Kauramäki J , Pomren C , et al Inter-subject synchronization of prefrontal cortex hemodynamic activity during natural viewing . Open Neuroimag J . 2008 1 1 ;2 (1 ):14 –9 .19018313 8 Chang W-T , Jääskeläinen IP , Belliveau JW , Huang S , Hung A-Y , Rossi S , et al Combined MEG and EEG show reliable patterns of electromagnetic brain activity during natural viewing . Neuroimage . 2015 7 1 ;114 :49 –56 . 10.1016/j.neuroimage.2015.03.066 25842290 9 Lankinen K , Saari J , Hari R , Koskinen M . Intersubject consistency of cortical MEG signals during movie viewing . Neuroimage . Elsevier Inc.; 2014 ;92 :217 –24 . 10.1016/j.neuroimage.2014.02.004 24531052 10 Honey CJ , Thesen T , Donner TH , Silbert LJ , Carlson CE , Devinsky O , et al Slow cortical dynamics and the accumulation of information over long timescales . Neuron . 2012 10 18 ;76 (2 ):423 –34 . 10.1016/j.neuron.2012.08.011 23083743 11 Meshulam M , Ramot M , Harel M , Kipervasser S , Andelman F , Neufeld MY , et al Selectivity of audiovisual ECoG responses revealed under naturalistic stimuli in the human cortex . J Neurophysiol . 2013 5 1 ;109 (9 ):2272 –81 . 10.1152/jn.00474.2012 23407355 12 Belitski A , Gretton A , Magri C , Murayama Y , Montemurro MA , Logothetis NK , et al Low-frequency local field potentials and spikes in primary visual cortex convey independent visual information . J Neurosci . 2008 5 28 ;28 (22 ):5696 –709 . 10.1523/JNEUROSCI.0009-08.2008 18509031 13 Hasson U , Yang E , Vallines I , Heeger DJ , Rubin N . A Hierarchy of Temporal Receptive Windows in Human Cortex . J Neurosci . 2008 ;28 (10 ):2539 –50 . 10.1523/JNEUROSCI.5487-07.2008 18322098 14 Zacks JM , Braver TS , Sheridan MA , Donaldson DI , Snyder AZ , Ollinger JM , et al Human brain activity time-locked to perceptual event boundaries . Nat Neurosci . 2001 6 ;4 (6 ):651 –5 . 11369948 15 Brainard DH . The psychophysics toolbox . Spat Vis . 1997 ;10 :433 –6 . 9176952 16 Pelli DG . The VideoToolbox software for visual psychophysics: transforming numbers into movies . Spat Vis. Brill ; 1997 1 1 ;10 (4 ):437 –42 . 17 Smith SM , Jenkinson M , Woolrich MW , Beckmann CF , Behrens TEJ , Johansen-Berg H , et al Advances in functional and structural MR image analysis and implementation as FSL . Neuroimage . 2004 1 ;23 Suppl 1 :S208 –19 . 15501092 18 Cox RW . AFNI: Software for Analysis and Visualization of Functional Magnetic Resonance Neuroimages . Comput Biomed Res . 1996 6 ;29 (3 ):162 –73 . 8812068 19 Benjamini Y , Hochberg Y . Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing . Source J R Stat Soc Ser B . 1995 ;57 (1 ):289 –300 . 20 Kauppi J-P , Pajula J , Tohka J . A versatile software package for inter-subject correlation based analyses of fMRI . Front Neuroinform . 2014 ;8 :2 10.3389/fninf.2014.00002 24550818 21 Lee C , Rohrer WH , Sparks DL . Population coding of saccadic eye movements by neurons in the superior colliculus . Nature . Nature Publishing Group; 1988 3 24 ;332 (6162 ):357 –60 . 3352733 22 Klier EM, Wang H, Crawford JD. The superior colliculus encodes gaze. 2001;627–32. 23 Lu Jianhua , L LM . A simple and efficient search algorithm for block-matching motion estimation . IEEE Trans Circuits Syst Video Technol . 1997 ;7 (2 ):429 –33 . 24 Marcus DS , Harwell J , Olsen T , Hodge M , Glasser MF , Prior F . Informatics and data mining tools and strategies for the human connectome project . Front Neuroinform . 2011 ;5 :4 10.3389/fninf.2011.00004 21743807 25 Pierrot-Deseilligny C . Decisional role of the dorsolateral prefrontal cortex in ocular motor behaviour . Brain. Oxford University Press ; 2003 4 8 ;126 (6 ):1460 –73 . 26 Grosbras M-H , Laird AR , Paus T . Cortical regions involved in eye movements, shifts of attention, and gaze perception . Hum Brain Mapp . 2005 5 ;25 (1 ):140 –54 . 15846814 27 Mainen Z , Sejnowski T . Reliability of spike timing in neocortical neurons . Science (80-) . American Association for the Advancement of Science; 1995 6 9 ;268 (5216 ):1503 –6 . 28 Mechler F , Victor JD , Purpura KP , Shapley R . Robust Temporal Coding of Contrast by V1 Neurons for Transient But Not for Steady-State Stimuli . J Neurosci . 1998 8 15 ;18 (16 ):6583 –98 . 9698345 29 Hubel DH , Wiesel TN . Receptive fields and functional architecture of monkey striate cortex . J Physiol . 1968 3 1 ;195 (1 ):215 –43 . 4966457 30 Maunsell JH , Newsome WT . Visual processing in monkey extrastriate cortex . Annu Rev Neurosci . 1987 ;10 :363 –401 . 3105414 31 Vogels R , Orban GA . Extrageniculostriate Mechanisms Underlying Visually-Guided Orientation Behavior . Progress in Brain Research . Elsevier; 1996 195 –211 p. 32 Kanwisher N , Wojciulik E . Visual attention: insights from brain imaging . Nat Rev Neurosci . 2000 11 ;1 (2 ):91 –100 . 11252779 33 Sherman SM , Guillery RW . On the actions that one nerve cell can have on another: Distinguishing “drivers” from “modulators.” Proc Natl Acad Sci . 1998 6 9 ;95 (12 ):7121 –6 . 9618549 34 Abbott LF , Chance FS . Drivers and modulators from push-pull and balanced synaptic input . Prog Brain Res . 2005 1 ;149 :147 –55 . 16226582 35 Grill-Spector K , Kourtzi Z , Kanwisher N . The lateral occipital complex and its role in object recognition . Vision Res . 2001 5 ;41 (10–11 ):1409 –22 . 11322983 36 Lerner Y . A Hierarchical Axis of Object Processing Stages in the Human Visual Cortex . Cereb Cortex . 2001 4 1 ;11 (4 ):287 –97 . 11278192 37 Huth AG , Nishimoto S , Vu AT , Gallant JL . A Continuous Semantic Space Describes the Representation of Thousands of Object and Action Categories across the Human Brain . Neuron . 2012 ;76 (6 ):1210 –24 . 10.1016/j.neuron.2012.10.014 23259955 38 Corbetta M , Akbudak E , Conturo TE , Snyder AZ , Ollinger JM , Drury HA , et al A common network of functional areas for attention and eye movements . Neuron . 1998 ;21 (4 ):761 –73 . 9808463

PMC005xxxxxx/PMC5001719.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756372610.1371/journal.pone.0161923PONE-D-15-48305Research ArticleSocial SciencesSociologyHuman FamiliesBiology and Life SciencesBiochemistryLipidsCholesterolMedicine and Health SciencesPublic and Occupational HealthBehavioral and Social Aspects of HealthMedicine and Health SciencesVascular MedicineBlood PressureSocial SciencesSociologySocial StratificationMedicine and Health SciencesPharmacologyPharmacokineticsDrug MetabolismMedicine and Health SciencesHealth CareSocioeconomic Aspects of HealthMedicine and Health SciencesPublic and Occupational HealthSocioeconomic Aspects of HealthPeople and PlacesPopulation GroupingsAge GroupsAdultsWork-Family Life Courses and Metabolic Markers in the MRC National Survey of Health and Development Work-Family Life Courses and Metabolic Riskhttp://orcid.org/0000-0002-3510-0795Lacey Rebecca E. 1Kumari Meena 2Sacker Amanda 1Stafford Mai 3Kuh Diana 3McMunn Anne 1*1 Department of Epidemiology and Public Health, University College London, London, United Kingdom2 Institute for Social and Economic Research, University of Essex, Colchester, United Kingdom3 Medical Research Council Unit for Lifelong Health and Ageing, University College London, London, United KingdomKirchmair Rudolf EditorMedical University Innsbruck, AUSTRIACompeting Interests: The authors have declared that no competing interests exist. Conceptualization: AM RL AS MK MS DK. Data curation: DK RL MS. Formal analysis: RL AS AM. Funding acquisition: AM. Investigation: AM RL AS MK MS DK. Methodology: RL AS AM. Project administration: AM. Resources: AM AS DK. Software: RL. Supervision: AM. Validation: AM RL AS MK MS DK. Visualization: RL. Writing – original draft: RL AM AS MK MS DK. Writing – review & editing: AM RL AS MK MS DK. * E-mail: a.mcmunn@ucl.ac.uk26 8 2016 2016 11 8 e01619239 11 2015 15 8 2016 © 2016 Lacey et al2016Lacey et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.The aim was to investigate whether the combined work-family life courses of British men and women were associated with differences in metabolic markers—waist circumference, blood pressure, high density lipoprotein cholesterol, triglycerides, and glycated haemoglobin—in mid-life. We used data from the Medical Research Council’s National Survey of Health and Development—the 1946 British birth cohort. Multi-channel sequence analysis was used to create a typology of eight work-family life course types combining information on work, partnerships and parenthood between ages 16–51. Linear regression tested associations between work-family types and metabolic outcomes at age 53 on multiply imputed data (20 imputations) of >2,400 participants. Compared with men with strong ties to employment and early transitions to family life, men who made later transitions to parenthood and maintained strong ties to paid work had smaller waist circumferences (-2.16cm, 95% CI: -3.73, -0.59), lower triglycerides (9.78% lower, 95% CI: 0.81, 17.94) and lower blood pressure (systolic: -4.03mmHg, 95% CI: -6.93, -1.13; diastolic: -2.34mmHg, 95% CI: -4.15, -0.53). Married men and women who didn’t have children had increased high density lipoprotein cholesterol (7.23% higher, 95% CI: 0.68, 14.21) and lower waist circumferences (-4.67cm, 95% CI: -8.37, -0.97), respectively. For men later transitions to parenthood combined with strong ties to paid work were linked to reduced metabolic risk in mid-life. Fewer differences between work-family types and metabolic markers were seen for women. http://dx.doi.org/10.13039/501100000781European Research CouncilERC-2011-StG_20101124McMunn Anne http://dx.doi.org/10.13039/501100000269Economic and Social Research CouncilES/J019119/1Sacker Amanda http://dx.doi.org/10.13039/501100000269Economic and Social Research CouncilES/J019119/1Kumari Meena http://dx.doi.org/10.13039/501100000265Medical Research CouncilMC_UU_12019/1Kuh Diana http://dx.doi.org/10.13039/501100000265Medical Research CouncilMC_UU_12019/4Kuh Diana http://dx.doi.org/10.13039/501100000265Medical Research CouncilMC_UU_12019/5Stafford Mai This work was supported by the European Research Council starter (https://erc.europa.eu/) grant (PI: Anne McMunn, grant number ERC-2011-StG_20101124). RL's time on this study was also supported by the above grant. AS and MK's time on this project was partially supported by the UK Economic and Social Research Council's (http://www.esrc.ac.uk/) International Centre for Life Course Studies in Society and Health (grant number ES/J019119/1). The Medical Research Council National Survey of Health and Development is supported by the UK Medical Research Council (MRC) (http://www.mrc.ac.uk/) through core funding to the MRC Unit for Lifelong Health and Ageing at UCL (MC_UU_12019/1). DK and MS's time was supported by the UK Medical Research Council (MC_UU_12019/4, MC_UU_12019/5). The Medical Research Council provided funding for the collection of data involved in this study. The European Research Council and Economic and Social Research Council played no role in the study design, data collection and analysis, decision to publish or preparation of the manuscript. Data AvailabilityData are available on request to the NSHD Data Sharing Committee. NSHD data sharing policies and processes meet the requirements and expectations of MRC policy on sharing of data from population and patient cohorts: (http://www.mrc.ac.uk/research/research-policy-ethics/data-sharing/policy). Data requests should be submitted to mrclha.swiftinfo@ucl.ac.uk; further details can be found at (http://www.nshd.mrc.ac.uk/data.aspx). These policies and processes are in place to ensure that the use of data from this 69 year old national birth cohort study is within the bounds of consent given previously by study members, complies with MRC guidance on ethics and research governance, and meets rigorous MRC data security standards.Data Availability Data are available on request to the NSHD Data Sharing Committee. NSHD data sharing policies and processes meet the requirements and expectations of MRC policy on sharing of data from population and patient cohorts: (http://www.mrc.ac.uk/research/research-policy-ethics/data-sharing/policy). Data requests should be submitted to mrclha.swiftinfo@ucl.ac.uk; further details can be found at (http://www.nshd.mrc.ac.uk/data.aspx). These policies and processes are in place to ensure that the use of data from this 69 year old national birth cohort study is within the bounds of consent given previously by study members, complies with MRC guidance on ethics and research governance, and meets rigorous MRC data security standards. ==== Body Introduction When examining the importance of the interdependence of work and family life for health, studies have found that the combination of paid work with family responsibilities is associated with better health[1–6]. However this is not always the case[7,8]. Previous research in this area has largely focused upon women, for whom participation in paid work is particularly affected by caring for children[9,10]. The health benefits of employment have been well documented[11,12]. Also, married people tend to live longer, and report better health, than those who are not married[13–17], a difference that is starker for men. The timing of key life course events, such as the transition to parenthood, are likely to be important for health, with previous work showing that early parenthood is associated with increased cardiovascular risk[18] and mortality[19]. This may be particularly disadvantageous for health when combined with weak ties to paid work and partnerships[20]. Existing research into the importance of the combination of work and family life for health has been limited through the exclusion of men from analyses, the use of data on social roles from only one or two time points, and the reliance on subjective measures of health. Also the potential of early life factors, such as selection by health and socioeconomic factors, to influence work-family life courses and subsequent health has been little investigated. Work-family life courses are thought to influence later health through biological and behavioural stress processes[21]. For instance, the multiple demands of combining paid work with parenting have been linked to poorer health in early studies[8]. However, more recent work has shown that weak ties to paid work and marriage, as well as early transitions to parenthood, are more often linked to increased physiological stress[11,14,20]. Part of this stress response might operate directly through altered physiological functioning (e.g. hypothalamo-pituitary-adrenal axis dysregulation and sympathetic-adrenal-medullary activation). More specifically, cortisol has also been shown to bind to glucocorticoid receptors on adipocytes in visceral fat, which can lead to triglyceride build up in visceral adipose tissue and hence increased central adiposity[22]. In turn visceral adiposity is pro-inflammatory, releasing cytokines such as interleukin-6 and tumour necrosis factor-α, resulting in changes in glucose and lipid metabolism and in the development of insulin resistance[23]. Mediation through differences in socioeconomic position[24], or an uptake, maintenance or increase in risky health behaviours[25–27], such as smoking, problem alcohol consumption and low physical activity levels, are also likely to play a role in explaining associations between work-family life courses and metabolic markers. This paper focuses upon metabolic markers, which are known to be important indicators of insulin resistance, operating upstream from type II diabetes and cardiovascular disease[28–30]. We used a British birth cohort to characterise combined work and family life courses of both men and women using multichannel sequence analysis. These work-family life courses were then related to metabolic markers (waist circumference, blood pressure, high density lipoprotein (HDL) cholesterol, triglycerides, glycated haemoglobin (HbA1c)) in mid-life thought to mediate the association between stress and later health. The hypothesis was that strong ties to paid work and to partnership in combination with later transitions to parenthood would be associated with reduced metabolic risk (e.g. smaller waist circumference). The importance of early life circumstances in setting people onto more disadvantaged work-family life courses was also assessed. Finally, the potential mediating role of adult socioeconomic position and health behaviours were evaluated. Methods Data This study used the UK Medical Research Council’s National Survey of Health and Development (NSHD), also known as the 1946 British birth cohort. The study received Multi-Centre Research Ethics Committee approval and written informed consent was provided by participants[31]. From an initial maternity survey of babies born during a single week in 1946, a stratified sample of 5,362 of all babies born to fathers in non-manual and agricultural employment and a quarter of births to fathers in manual employment was formed[32]. Our analyses were weighted to account for the stratified composition of the NSHD sample. Participants have been surveyed more than 23 times and information collected on economic, social, developmental and biological aspects[31]. At 53 years 3,035 (56.6%) participants were still part of the study (469 had died, 668 had permanently refused, 580 were living abroad, 330 were lost to follow-up and 280 temporarily refused the age 53 survey)[32]. Measures Metabolic markers Data on waist circumference, blood pressure (systolic (SBP) and diastolic (DBP)), and blood samples for the assessment of triglycerides, HDL cholesterol, and HbA1c were available at age 53. Blood samples were collected from non-fasted participants at age 53. Details of the biochemical procedures of triglyceride, HDL cholesterol and HbA1c measurement have been published previously[33]. Participants were asked about medication status at time of blood sampling. Those taking medications affecting blood pressure (β-blockers, Calcium channel blockers, diuretics and drugs affecting the renin-angiotensin system) had their SBP values increased by 10mmHg and their DBP values increased by 5mmHg as recommended[34]. A similar procedure could not be followed for medications affecting HbA1c, triglycerides, and HDL cholesterol. Information on anti-diabetes and lipid-lowering medications was used to conduct sensitivity analyses in associations between work-family types and HbA1c, triglycerides, and HDL cholesterol, respectively. The sensitivity analyses for HbA1c involved running the analyses for all participants and then comparing these results to those when particpants taking anti-diabetic medications were removed. This was repeated for analyses involving triglycerides and HDL cholesterol using data on lipid-lowering medications. Work-family life courses Information on work, partnerships and parenthood was available at each adulthood survey (ages 19, 20, 22, 23, 25, 26, 31, 36, 43, and 53). Work, partnership and parenthood status variables were derived annually between ages 16 and 51 years (ending two years prior to metabolic risk evaluation at age 53). Work status was defined as full-time employment, part-time employment (≤30 hours/week), full-time homemaking, or other not employed (unemployed, sick, in education or other reason). Partnership status was defined as married, cohabiting, or not living with a partner. Parental status was categorised as no children in the household or youngest child >16 years, youngest child in the household <5 years, or youngest child in the household 5–16 years. These three life course domains were cross-classified to create 35 combined work-family state variables (one for each year between 16 and 51 years), each with 36 possible combinations of work, partnership and parenthood (4 work states x 3 partnership states x 3 parenthood states). Sequence analysis was used to condense this detailed life course information into a work-family typology. This method measured the distance of each cohort member’s individual work-family sequence to a set of eight pre-defined model biographies. These model biographies were specified based upon previous knowledge of this cohort, and with a view to including as much variation across genders as possible whilst still maintaining adequate power (Table 1). Distances from each individual’s work-family sequence to the eight model biographies was calculated using the Dynamic Hamming approach[35], which is particularly appropriate when the timing of transitions is considered to be important. Participants were then categorised based upon their closest model biography, thereby creating a single work-family type variable with eight categories. Further information on how the work-family types were derived can be found in S1 Appendix and also in McMunn et al[36] and Lacey et al[20]. Estimates for work-family types containing fewer than 2% of men or women are not shown in subsequent results as these are unlikely to be reliable (Table 1). 10.1371/journal.pone.0161923.t001Table 1 Distribution of Work-Family Life Course Types and Associated Model Biographies in the NSHD. Work-family type Men %a (n = 1252)b Women %a (n = 1251)b Model biography sequence ‘Work, early family’ 47.7 15.1 Continuous full-time employment; married and children from early 20s ‘Work, marriage, non-parent’ 7.9 9.0 Continuous full-time employment; married from early 20s; no children ‘Work, no family’ 11.5 6.1 Continuous full-time employment; no partner or children ‘Work, later family’ 30.6 3.5 Continuous full-time employment; cohabiting mid-20s, married from late 20s; children from early 30s ‘Later family, work break’ 1.0c 11.6 Employed full-time until late 20s, homemaking from early 30s; married from mid 20s; children from early 30s ‘Early family, work break’ 0.6c 14.6 Employed full-time until early 20s, homemaking from early-late 20s, employed part-time from early 30s; marriage and children from early 20s ‘Part-time work, early family’ 0.7c 29.9 Employed full-time until early 20s, part time employed from early 20s; marriage and children from early 20s ‘No paid work, early family’ 0.02c 10.3 Employed part-time until early 20s, homemaking from early 20s; marriage and children from early 20s a Results presented as percentages as are imputed data b Descriptives given for those with at least one observed outcome (n = 2,503) c Work-family groups containing fewer than 2% of participants are not presented in subsequent analyses as estimates are unlikely to be reliable Covariates Indicators of early life health and socioeconomic position (SEP) were included in this study to account for potential selections into different work-family types. At age 15 the cohort member’s parent was asked whether they had any concerns regarding the child’s health. Also information on internalising and externalising behaviours was derived from precursors of Rutter’s behavioural scales at ages 13 and 15, reported by the cohort member’s teacher. Factor analysis was used to derive a measure of internalising and externalising disorders by categorising scores based on established percentile cut-points[37]. For internalising scores: 0–50% (absent), 51–87% (mild) and ≥88% (severe). For externalising scores: 0–75% (absent), 76–93% (mild) and ≥94% (severe). Father’s social class (UK Registrar General’s Social Class schema) was used to indicate childhood SEP at age 4. Social class was categorised as professional and managerial (I), intermediate (II), skilled non-manual (IIINM), skilled manual (IIIM), semi-skilled (IV) or unskilled manual (V). Where this information was not available at age 4 information was taken from age 11 (n = 125) or age 15 (n = 48). Educational attainment was indicated by the highest qualification achieved by age 26 and categorised as no qualifications, secondary school education (Ordinary-level or Certificate of Secondary Education) or vocational training, advanced secondary education (Advanced-level or Burnham A2), degree or higher qualification. Adult mediators included in this study were health behaviours, social class and body mass index (BMI) at age 53. Health behaviours considered were smoking status (never/ex/current), whether the cohort member participates in physical activity and problem drinking as indicated by the CAGE score (score of ≥2). The social class (Registrar General Social Class schema) of the head of household was used to indicate adult SEP. Height and weight were measured by the study nurses and BMI calculated as kg/m2. Statistical analysis Missing data Missing data are a particular problem for longitudinal studies, potentially resulting in bias, reduced sample sizes and loss of statistical power[38]. In this study missing information on work, partnerships and parenthood was imputed using a recommended method to overcome problems of collinearity and inaccurate estimation of missing sequence data[39]. Twenty imputed datasets were created. Multiple imputation by chained equations was then conducted to impute the covariates for those with complete work-family information following imputation (n = 2,513). The approach of imputation then deletion[40] was employed whereby all covariates were imputed for all cases before excluding those with missing data on each metabolic outcome. Descriptive analyses are presented for those with at least one observed outcome (n = 2,503). Regression analyses Ordinary least squares regression was used to test associations between work-family types and metabolic markers. Firstly the crude association was tested (model 1). Secondly we controlled for early life confounders–childhood SEP, educational attainment and child health (model 2). Finally we included potential adult mediators (health behaviours, adult SEP and BMI) in model 3. Models in which waist circumference was the outcome did not include BMI to avoid ‘over-adjusting’ for insulin resistance. As HbA1c, triglycerides and HDL cholesterol were log-transformed, results are presented as percentage difference to aid interpretation. All analyses were conducted using Stata version 13[41]. Results The distribution of work-family types in the NSHD is shown in Table 1. The majority (97.7%) of men were in work-family types characterised by continuous full-time employment; almost half (47.7%) of men combined this with early transitions to family life, whilst 30.6% made a later transition. Women’s work-family types were more diverse than men’s and women were more likely to occupy work-family types with weaker ties to paid work, for example long-term part-time employment (29.9%) and long-term full-time homemaking (10.3%). Further characteristics of the study sample are shown in S1 Table and S2 Table for men and women, respectively. Work-family types and metabolic markers in men Table 2 shows the associations between work-family types and metabolic markers in men. Compared to men in the ‘Work, early family’ type, men who made later transitions to family life (‘Work, later family’) had smaller waist circumferences, which remained after controlling for early life factors (-2.16cm, 95% CI: -3.73, -0.59). The estimate did not change upon considering adult mediators of interest (adult SEP and heath behaviours). This was also the case for SBP for the same work-family type. The ‘Work, later family’ type also had lower DBP (-2.34mmHg, 95% CI: -4.15, -0.53) and triglycerides (9.78% lower, 95% CI: 0.81, 17.94), and in both cases this association was largely explained by the lower than average BMIs of this group. Men who were married but who were not parents (‘Work, marriage, non-parent’) had higher HDL cholesterol levels. This association was explained largely by differences in BMI. No differences in HbA1c levels by work-family type were seen for men. The results for HDL cholesterol and triglycerides were robust to exclusion of those taking lipid-lowering medication (n = 43 and n = 37, respectively). Results for HbA1c did not change upon removal of those taking anti-diabetes medications (n = 24). 10.1371/journal.pone.0161923.t002Table 2 Associations Between Work-Family Life Courses and Metabolic Markers at Age 53 for Men. Model 1 –crude association Model 2 –model 1 + early life factors Model 3 –model 2 + adult mediators Regression coefficienta 95% CI Regression coefficienta 95% CI Regression coefficienta 95% CI Waist circumference (n = 1244)             Work, early family Ref Ref Ref Work, marriage, non-parent -2.11 -4.51, 0.29 -1.56 -4.01, 0.89 -1.26 -3.73, 1.21 Work, no family -1.56 -4.23, 1.10 -1.10 -3.74, 1.53 -0.40 -3.02, 2.23 Work, later family -2.66 -4.23, -1.09 -2.16 -3.73, -0.59 -2.16 -3.71, -0.62 R-squared (%) 1.2 4.7 8.9 Systolic blood pressure (n = 1243)             Work, early family Ref Ref Ref Work, marriage, non-parent 2.73 -2.48, 7.94 3.31 -1.90, 8.53 4.65 -0.46, 9.75 Work, no family -3.51 -7.64, 0.62 -2.68 -6.84, 1.48 -1.09 -5.10, 2.91 Work, later family -4.68 -7.55, -1.81 -4.03 -6.93, -1.13 -3.19 -6.00, -0.38 R-squared 1.8 5.6 12.7 Diastolic blood pressure (n = 1243) Work, early family Ref Ref Ref Work, marriage, non-parent 0.40 -2.57, 3.36 0.78 -2.23, 3.80 1.75 -1.04, 4.54 Work, no family -1.93 -4.71, 0.86 -1.50 -4.25, 1.24 -0.34 -3.01, 2.32 Work, later family -2.63 -4.44, -0.81 -2.34 -4.15, -0.53 -1.75 -3.52, 0.02 R-squared 1.5 5.3 12.9 % differenceb 95% CI % differenceb 95% CI % differenceb 95% CI Triglycerides (n = 1033) Work, early family Ref Ref Ref Work, marriage, non-parent -13.78 -25.72, 0.06 -12.37 -24.86, 2.19 -7.76 -19.85, 6.15 Work, no family -12.65 -24.35, 0.88 -10.81 -22.42, 2.55 -2.91 -14.52, 10.28 Work, later family -10.47 -18.32, -1.86 -9.78 -17.94, -0.81 -6.57 -14.67, 2.29 R-squared 1.4 4.5 16.2 HDL cholesterol (n = 1001) Work, early family Ref Ref Ref Work, marriage, non-parent 7.02 0.48, 13.99 7.23 0.68, 14.21 5.59 -1.02, 12.65 Work, no family -5.09 -11.14, 1.37 -3.65 -10.89, 2.02 -5.86 -12.19, 0.93 Work, later family 3.54 -1.09, 8.39 3.06 -1.67, 8.03 3.07 -1.40, 7.74 R-squared 1.4 3.2 14.5 HbA1c (n = 1110) Work, early family Ref Ref Ref Work, marriage, non-parent -0.73 -3.06, 1.65 0.02 -2.45, 2.55 0.56 -1.86, 3.05 Work, no family 2.01 -1.00, 5.07 2.28 -0.71, 5.37 2.77 -0.32, 5.95 Work, later family -1.19 -2.79, 0.45 -0.55 -2.25, 1.16 -0.27 -1.94, 1.43 R-squared 0.9 4.3 5.9 aUnstandardised regression coefficients b Results for triglycerides, HDL cholesterol and HbA1c are presented as percentage difference as these outcomes were log-transformed Model 1 –crude association Model 2 additionally includes childhood social class, physical health, internalising and externalising disorders, and educational attainment Model 3 additionally includes household social class, smoking status, physical activity, problem drinking, and BMI (except in regressions with waist circumference) Work-family types and metabolic markers in women Associations between work-family life course types and metabolic markers for women are shown in Table 3. In general, fewer associations were seen for women than for men. Regarding waist circumference, women who were married but who didn’t have children (‘Work, marriage, non-parent’) had smaller waist circumferences (-4.67cm, 95% CI: -8.37, -0.97) than the reference group who had children (‘Work, early family’). This difference was not explained by the adult mediators of interest in this study. No other associations between work-family types and metabolic markers were seen for women. Results did not change during sensitivity analyses in which participants taking medications were removed (see results for men–same procedure followed, removing 15 and 12 women taking lipid-lowering medications in triglyceride and HDL cholesterol analyses, respectively. Also removing 21 women taking anti-diabetic medications in HbA1c analyses). 10.1371/journal.pone.0161923.t003Table 3 Associations Between Work-Family Life Courses and Metabolic Markers at Age 53 for Women. Model 1 –crude association Model 2 –model 1 + early life factors Model 3 –model 2 + adult mediators   Regression coefficienta 95% CI Regression coefficienta 95% CI Regression coefficienta 95% CI Waist circumference (n = 1244)             Work, early family Ref Ref Ref Work, marriage, non-parent -4.27 -8.02, -0.51 -4.67 -8.37, -0.97 -4.74 -8.50, -0.98 Work, no family -1.90 -6.42, 2.63 -1.33 -5.82, 3.16 -0.46 -4.81, 3.90 Work, later family -0.45 -6.20, 5.30 -0.35 -5.92, 5.21 0.28 -5.05, 5.60 Later family, work break -0.47 -3.87, 2.93 -0.67 -3.98, 2.64 -0.31 -3.61, 3.00 Early family, work break -0.89 -4.08, 2.30 -1.41 -4.54, 1.72 -1.26 -4.29, 1.77 Part-time work, early family -1.76 -4.65, 1.12 -2.66 -5.52, 0.20 -2.14 -4.65, 0.66 No paid work, early family -0.80 -4.99, 3.39 -1.73 -5.85, 2.40 -1.87 -5.95, 2.21 R-squared 0.9 5.2 10.1 Systolic blood pressure (n = 1213)             Work, early family Ref Ref Ref Work, marriage, non-parent 2.43 -4.14, 8.99 1.54 -5.04, 8.12 3.07 -3.32, 9.47 Work, no family -1.12 -7.91, 5.68 -1.08 -7.97, 5.81 -0.14 -7.11, 6.82 Work, later family 2.82 -6.27, 11.91 2.70 -6.65, 12.04 1.54 -7.41, 10.50 Later family, work break -1.20 -6.35, 3.95 -1.62 -6.94, 3.69 -2.04 -7.49, 3.41 Early family, work break 2.90 -2.59, 8.38 2.26 -3.29, 7.81 2.39 -3.13, 7.91 Part-time work, early family 1.38 -2.89, 5.66 0.79 -3.62, 5.19 1.25 -3.21, 5.71 No paid work, early family 2.57 -3.24, 8.37 1.65 -4.21, 7.52 2.02 -3.67, 7.72 R-squared 0.2 1.3 8.1 Diastolic blood pressure (n = 1219) Work, early family Ref Ref Ref Work, marriage, non-parent 0.18 -3.51, 3.88 -0.24 -3.92, 3.45 0.77 -2.87, 4.41 Work, no family 0.29 -3.61, 4.18 0.09 -3.90, 4.07 0.76 -3.36, 4.88 Work, later family 0.56 -3.73, 4.84 0.32 -4.12, 4.77 -0.33 -4.68, 4.02 Later family, work break -0.44 -3.28, 2.41 -0.66 -3.53, 2.21 -0.73 -3.64, 2.18 Early family, work break -0.10 -3.07, 2.88 -0.26 -3.27, 2.74 -0.03 -3.05, 2.99 Part-time work, early family 0.10 -2.54, 2.74 -0.15 -2.83, 2.54 0.16 -2.49, 2.81 No paid work, early family 1.36 -2.14, 4.86 1.15 -2.42, 4.71 1.54 -1.95, 5.03 R-squared 0.2 1.2 8.4 % differenceb 95% CI % differenceb 95% CI % differenceb 95% CI Triglycerides (n = 1448) Work, early family Ref Ref Ref Work, marriage, non-parent 13.43 -4.12, 34.21 13.99 -3.44, 34.57 24.09 6.68, 44.35 Work, no family 2.63 -14.09, 22.60 5.76 -11.44, 26.29 12.43 -4.24, 32.00 Work, later family 10.40 -14.86, 43.15 11.70 -14.18, 45.38 9.99 -13.12, 39.35 Later family, work break -2.39 -14.12, 10.96 -1.78 -13.71, 11.79 1.19 -10.41, 14.30 Early family, work break 10.61 -3.34, 26.58 10.66 -3.37, 26.73 14.71 1.62, 29.47 Part-time work, early family 6.11 -5.68, 19.37 5.06 -6.66, 18.25 8.84 -2.18, 21.11 No paid work, early family 7.73 -9.72, 28.55 7.62 -9.52, 28.02 9.29 -5.91, 26.95 R-squared 0.7 2.6 19.3 HDL cholesterol (n = 1419) Work, early family Ref Ref Ref Work, marriage, non-parent 3.77 -4.65, 12.92 3.63 -4.71, 12.71 -0.14 -7.27, 7.53 Work, no family 2.40 -7.19, 12.99 1.13 -8.27, 11.50 -0.50 -9.22, 9.06 Work, later family -6.63 -17.36, 5.50 -8.46 -19.54, 4.16 -8.86 -18.49, 1.91 Later family, work break 3.17 -4.35, 11.28 3.12 -4.56, 11.43 0.45 -6.78, 8.23 Early family, work break 0.60 -6.46, 8.19 0.48 -6.49, 7.97 -2.00 -8.29, 4.72 Part-time work, early family -0.21 -6.63, 6.65 0.50 -5.87, 7.31 -1.07 -6.82, 5.02 No paid work, early family 0.16 -8.26, 9.36 0.87 -7.39, 9.86 -1.40 -8.57, 6.35 R-squared 0.8 4.7 19.6 HbA1c (n = 1500) Work, early family Ref Ref Ref Work, marriage, non-parent -0.05 -3.58, 3.61 -0.24 -3.76, 3.41 0.75 -2.86, 4.50 Work, no family 1.73 -1.45, 5.01 1.89 -1.33, 5.21 2.48 -0.75, 5.82 Work, later family -0.37 -3.14, 2.48 -0.61 -3.42, 2.28 -0.93 -3.82, 2.05 Later family, work break -0.67 -3.07, 1.78 -0.77 -3.18, 1.71 -0.62 -3.10, 1.93 Early family, work break -1.01 -3.28, 1.32 -1.07 -3.36, 1.28 -0.56 -2.74, 2.68 Part-time work, early family 0.80 -1.36, 3.01 0.51 -1.64, 2.72 0.83 -1.31, 3.03 No paid work, early family 3.09 -1.74, 8.16 2.69 -2.07, 7.67 2.76 -1.55, 7.27 R-squared 0.5 2.9 10.6 aUnstandardised regression coefficients b Results for triglycerides, HDL cholesterol and HbA1c are presented as percentage difference as these outcomes were log-transformed Model 1—gender adjusted Model 2 additionally includes gender, childhood social class, internalising and externalising disorders, and educational attainment Model 3 additionally includes household social class, smoking status, physical activity, problem drinking, and BMI (except in regression with waist circumference) Discussion Using a British birth cohort and an innovative method of characterising the work and family lives of British men and women, we found that later parenthood in combination with continuous full-time employment and marriage is associated with a more favourable metabolic risk profile (smaller waist circumference, lower blood pressure, and lower triglycerides) for men. The health advantage of later parenthood has previously been shown in this cohort[18] and in other studies[19,42,43]. However the present study adds to these findings by showing that it is later parenthood in combination with full-time employment which appears to be most beneficial for health. The link between early parenthood and higher blood pressure has been shown before in this cohort and was thought to be explained by increased stress resulting in prolonged sympathetic nervous system activation[18]. The more advantageous metabolic risk profiles of men entering parenthood later, at least in relation to SBP and waist circumference, were not fully explained by adult SEP, BMI or health behaviours in our study. It is therefore possible that there is a direct physiological response which does not operate through these factors. For instance, an increase in glucocorticoids, through hypothalamo-pituitary-adrenal axis dysregulation–one of the likely stress mechanisms linking social stressors, such as early family formation, to later health–may result in increased central adiposity through the differentiation and proliferation of adipocytes in visceral and abdominal adipose tissue. This has been shown in both animal and human studies[44–47]. It is possible that more accurately captured health behaviours measured over longer periods of time may play a role. Alternatively there may be a role of other mediators or residual confounding that we have not been able to consider in this study. For example, it is possible that other lifestyle factors are involved, such as diet, which we have not accounted for in this study but which are to some extent socially controlled within families[48]. Interestingly, women who were in this same work-family type (‘Work, later family’) did not have a better metabolic profile in mid-life than those who made an earlier transition to family life (‘Work, early family’–the reference group). This work-family group a smaller group of women (3.5% of women) and it is therefore possible that there was insufficient statistical power to detect a significant difference between this group and the reference. A power calculation was performed suggesting that 14.2% (approximately 177 women) would be required in this category to detect a statistical difference at the 80% power level. However the estimates appear to indicate little difference in metabolic markers. Also the findings for women are partly consistent with previous work by Hardy et al on the same cohort[18], which showed that women who had children earlier did not have significantly different DBP, waist:hip ratios, triglycerides, or HbA1c levels than women who had children later. However they did find statistically significant differences in relation to SBP which we did not find in the present study. It is possible that any differences are diluted for women when additionally considering work and partnerships. Men in the ‘Work, marriage, non-parent’ group had higher HDL cholesterol and women lower waist circumferences, indicative of lower metabolic risk. The association for HDL cholesterol amongst men appeared to be mediated by differences in BMI, however the smaller waist circumferences of women who did not have children was not fully explained by our mediators of interest. It is possible that there is a residual pregnancy effect; that women who do not have children retain a smaller waist circumference than their peers who do. Interestingly, in a previous study using the same cohort, we found that this work-family type had lower levels of life satisfaction at ages 60–64[6], suggesting a discrepancy between objective markers of health and subjective wellbeing. The work-family life courses in this cohort were not found to be associated with differences in HbA1c for men or women. This is consistent with previous work on this cohort,[18] in which no differences in HbA1c were seen by age of parenthood. Methodological considerations This study has a number of strengths and limitations. In the present study we were unable to take account of more detailed processes, such as work and relationship quality. It has been shown previously that these factors may be more important than role occupation[49,50]. Secondly, non-fasting blood samples were taken from participants. HDL cholesterol and HbA1c do not require fasting to be accurate and reliable[51]; however, triglycerides are sensitive to fasting status. Despite this, non-fasting triglycerides have previously been highlighted as markers of insulin resistance and risk factors for myocardial infarction, ischaemic strokes and cardiovascular mortality[52]. This study has many strengths. For instance, this is one of the first studies to use multichannel sequence analysis to simultaneously consider the work, partnerships and parenthood histories of both men and women, recognising the interdependence of these important life course domains. The prospective longitudinal design enabled the inclusion of early life factors to account for potential selection into different work-family life courses. Results from this study are likely to be generalizable to British men and women of a similar age. Missing data were accounted for using multiple imputation, including an approach appropriate to categorical time series data. Finally, in contrast to many previous studies, work and family histories were linked to objective markers of health. In conclusion, this study suggests that later parenthood combined with strong ties to paid work may result in a more favourable metabolic profile in mid-life for men. However, further research is needed into the detailed causal processes, such as qualitative aspects of work and family life (e.g. work stress and employment conditions), which might further explain these associations. Our findings also allude to the timing of parenthood as driving many of the associations seen here, and further research is needed to assess why this might be. Policies which promote the health of young parents and enable strong work ties may be fruitful in improving health. Supporting Information S1 Appendix Further information on sequence analysis. (DOCX) Click here for additional data file. S1 Table Descriptive statistics of analysis variables by work-family type for NSHD men. (DOCX) Click here for additional data file. S2 Table Descriptive statistics of analysis variables by work-family type for NSHD women. (DOCX) Click here for additional data file. We would like to thank the cohort members and study team involved in the MRC National Survey of Health and Development. ==== Refs References 1 McMunn A , Bartley M , Hardy R , Kuh D . Life course social roles and women’s health in mid-life: causation or selection? J Epidemiol Community Health . 2006 ;60 : 484 –489 . 16698977 2 McMunn A , Bartley M , Kuh D . Women’s health in mid-life: Life course social roles and agency as quality . Soc Sci Med . 2006 ;63 : 1561 –1572 . 16698159 3 Janzen B , Muhajarine N . Social role occupancy, gender, income adequacy, life stage and health: A longitudinal study of employed Canadian men and women . Soc Sci Med . 2003 ;57 : 1491 –1503 . 12927478 4 Nordenmark M . Multiple Social Roles and Well-Being: A Longitudinal Test of the Role Stress Theory and the Role Expansion Theory . Acta Sociologica . 2004 pp. 115 –126 . 5 Sabbath EL , Guevara IM , Glymour MM , Berkman LF . Use of Life Course Work-Family Profiles to Predict Mortality Risk Among US Women . Am J Public Health . 2015 ;105 : e96 –e102 . 10.2105/AJPH.2014.302471 25713976 6 Lacey RE , Stafford M , Sacker A , McMunn A . Work-family life courses and subjective wellbeing in the MRC National Survey of Health and Development (the 1946 British birth cohort) . J Popul Ageing . 2015 ; 10.1007/s12062-015-9126-y 7 Johansson G , Huang Q , Lindfors P . A life-span perspective on women’s careers, health, and well-being . Soc Sci Med . 2007 ;65 : 685 –697 . 17493728 8 Hewitt B , Baxter J , Western M . Family, work and health: the impact of marriage, parenthood and employment on self-reported health of Australian men and women . J Sociol . 2006 ;42 : 61 –78 . 9 Pailhe A , Robette N , Solaz A . Work and family over the life course: a typology of French long-lasting couples using optimal matching . Longit Life Course Stud . 2013 ;4 : 196 –217 . 10 Schober P . The Parenthood Effect on Gender Inequality: Explaining the Change in Paid and Domestic Work When British Couples Become Parents . Eur Sociol Rev . 2013 ;29 : 74 –85 . 11 Janicki-Deverts D , Cohen S , Matthews KA , Cullen MR . History of unemployment predicts future elevations in C-reactive protein among male participants in the Coronary Artery Risk Development in Young Adults (CARDIA) Study . Ann Behav Med . 2008 ;36 : 176 –85 . 10.1007/s12160-008-9056-5 18784972 12 McKee-Ryan F , Song Z , Wanberg C , Kinicki A . Psychological and physical well-being during unemployment: a meta-analytic study Frankenhaeuser M , Lundberg U , Chesney M , editors. J Appl Psychol . Boston, MA : Springer US ; 2005 ;90 : 53 –76 . 15641890 13 Grundy EMD , Tomassini C . Marital history, health and mortality among older men and women in England and Wales . BMC Public Health . 2010 ;10 : 554 10.1186/1471-2458-10-554 20843303 14 McFarland M , Hayward M , Brown D . I’ve got you under my skin: marital biography and biological risk . J Marriage Fam . 2013 ;75 : 363 –380 . 26078480 15 Strohschein L , McDonough P , Monette G , Shao Q . Marital transitions and mental health: Are there gender differences in the short-term effects of marital status change? Soc Sci Med . 2005 ;61 : 2293 –2303 . 16099576 16 Ebrahim S , Wannamethee G , McCallum A , Walker M , Shaper AG . Marital status, change in marital status, and mortality in middle-aged British men . Am J Epidemiol . 1995 ;142 : 834 –842 . 7572960 17 Jaffe DH , Manor O , Eisenbach Z , Neumark YD . The Protective Effect of Marriage on Mortality in a Dynamic Society . Ann Epidemiol . 2007 ;17 : 540 –547 . 17434751 18 Hardy R , Lawlor DA , Black S , Mishra GD , Kuh D . Age at birth of first child and coronary heart disease risk factors at age 53 years in men and women: British birth cohort study . J Epidemiol Community Health . 2009 ;63 : 99 –105 . 10.1136/jech.2008.076943 18782806 19 Mirowsky J . Age at first birth, health, and mortality . J Health Soc Behav . 2005 ;46 : 32 –50 . 15869119 20 Lacey RE , Sacker A , Kumari M , Worts D , McDonough P , Booker CL , et al Work-family life courses and markers of stress and inflammation in mid-life: evidence from the National Child Development Study . Int J Epidemiol . 2015 ; 10.1093/ije/dyv205 21 Kostiainen E , Martelin T , Kestila L , Martikainen P , Koskinen S . Employee, Partner, and Mother: Woman’s Three Roles and Their Implications for Health . J Fam Issues . 2009 ;30 : 1122 –1150 . 22 Björntorp P . Do stress reactions cause abdominal obesity and comorbidities? Obes Rev . 2001 ;2 : 73 –86 . 12119665 23 Kyrou I , Chrousos GP , Tsigos C . Stress, visceral obesity, and metabolic complications . Ann N Y Acad Sci . 2006 ;1083 : 77 –110 . 10.1196/annals.1367.008 17148735 24 Lahelma E , Arber S , Kivelä K , Roos E . Multiple roles and health among British and Finnish women: The influence of socioeconomic circumstances . Social Science and Medicine . 2002 pp. 727 –740 . 11999489 25 Read S , Grundy E , Wolf DA . Fertility history, health, and health changes in later life: a panel study of British women and men born 1923–49 . Popul Stud (NY) . 2011 ;65 : 201 –215 . 26 Nomaguchi KM , Bianchi SM . Exercise Time: Gender Differences in the Effects of Marriage, Parenthood, and Employment . J Marriage Fam . 2004 ;66 : 413 –430 . 27 Little M , Handley E , Leuthe E , Chassin L . The impact of parenthood on alcohol consumption trajectories: Variations as a function of timing of parenthood, familial alcoholism, and gender . Dev Psychopathol . Cambridge University Press; 2009 ;21 : 661 10.1017/S0954579409000352 19338703 28 Goldbourt U , Yaari S , Medalie JH . Isolated Low HDL Cholesterol As a Risk Factor for Coronary Heart Disease Mortality: A 21-Year Follow-up of 8000 Men . Arterioscler Thromb Vasc Biol . 1997 ;17 : 107 –113 . 9012644 29 Vazquez G , Duval S , Jacobs DR , Silventoinen K . Comparison of body mass index, waist circumference, and waist/hip ratio in predicting incident diabetes: a meta-analysis . Epidemiol Rev . 2007 ;29 : 115 –28 . 17494056 30 Balkau B , Deanfield JE , Després J-P , Bassand J-P , Fox KAA , Smith SC , et al International Day for the Evaluation of Abdominal Obesity (IDEA): a study of waist circumference, cardiovascular disease, and diabetes mellitus in 168,000 primary care patients in 63 countries . Circulation . 2007 ;116 : 1942 –51 . 17965405 31 Kuh D , Pierce M , Adams J , Deanfield J , Ekelund U , Friberg P , et al Cohort profile: updating the cohort profile for the MRC National Survey of Health and Development: a new clinic-based data collection for ageing research . Int J Epidemiol . 2011 ;40 : e1 –9 . 10.1093/ije/dyq231 21345808 32 Wadsworth M , Kuh D , Richards M , Hardy R . Cohort Profile: The 1946 National Birth Cohort (MRC National Survey of Health and Development) . Int J Epidemiol . 2006 ;35 : 49 –54 . 16204333 33 Skidmore PML , Hardy RJ , Kuh DJ , Langenberg C , Wadsworth MEJ . Birth Weight and Lipids in a National Birth Cohort Study . Arterioscler Thromb Vasc Biol . 2004 ;24 : 588 –594 . 14715646 34 Cui JS , Hopper JL , Harrap SB . Antihypertensive treatments obscure familial contributions to blood pressure variation . Hypertension . 2003 ;41 : 207 –10 . 12574083 35 Lesnard L . Setting Cost in Optimal Matching to Uncover Contemporaneous Socio-Temporal Patterns . Sociol Methods Res . 2010 ;38 : 389 –419 . 36 McMunn A , Lacey R , Worts D , McDonough P , Stafford M , Booker CL , et al De-standardization and gender convergence in work-family life courses in Great Britain: a multi-channel sequence analysis . Adv Life Course Res . 2015 ;26 : 60 –75 . 37 Richards M , Abbott R , Collis G , Hackett P , Hotopf M , Kuh D , et al Childhood mental health and life chances in post-war Britain London : Sainsbury Centre for Mental Health ; 2009 . 38 Abraham WT , Russell D . Missing data: a review of current methods and applications in epidemiological research . Curr Opin Psychiatry . 2004 ;17 : 315 –321 . 39 Halpin B . Multiple imputation for life-course sequence data [Internet]. Limerick : University of Limerick ; 2012 Available: http://www.academia.edu/2601453/Multiple_Imputation_for_Life-Course_Sequence_Data 40 Von Hippel P . Regression with missing Ys: An improved strategy for analyzing multiply imputed data . Sociol Methodol . 2007 ;37 : 83 –117 . 41 StataCorp. Stata version 13.1 Texas : StataCorp LP ; 2013 . 42 Grundy E , Read S . Pathways from fertility history to later life health: results from analyses of the English Longitudinal Study of Ageing . Demogr Res . 2015 ;32 : 107 –146 . 43 Henretta JC . Early Childbearing, Marital Status, and Women’s Health and Mortality after Age 50 . J Health Soc Behav . 2007 ;48 : 254 –266 . 17982867 44 Rebuffe-Scrive M , Walsh U , McEwen B , Rodin J . Effect of chronic stress and exogenous glucocorticoids on regional fat distribution and metabolism . Physiol Behav . 1992 ;52 : 583 –590 . 1409924 45 Anagnostis P , Athyros VG , Tziomalos K , Karagiannis A , Mikhailidis DP . Clinical review: The pathogenetic role of cortisol in the metabolic syndrome: a hypothesis . J Clin Endocrinol Metab . Endocrine Society; 2009 ;94 : 2692 –701 . 10.1210/jc.2009-0370 19470627 46 Epel E , McEwen B , Seeman T , Matthews K , Castellazzo G , Brownell K , et al Stress and body shape: stress-induced cortisol secretion is consistently greater among women with central fat . Psychosom Med . 2000 ;62 : 623 –632 . 11020091 47 Marniemi J , Kronholm E , Aunola S , Toikka T , Mattlar C , Koskenvuo M , et al Visceral fat and psychosocial stress in identical twins discordant for obesity . J Intern Med . 2002 ;251 : 35 –43 . 11851863 48 Umberson D , Crosnoe R , Reczek C . Social Relationships and Health Behavior Across Life Course . Annu Rev Sociol . 2010 ;36 : 139 –157 . 21921974 49 Plaisier I , Beekman ATF , de Bruijn JGM , de Graaf R , ten Have M , Smit JH , et al The effect of social roles on mental health: A matter of quantity or quality? J Affect Disord . 2008 ;111 : 261 –270 . 10.1016/j.jad.2008.03.007 18448169 50 Butterworth P , Leach L , Strazdins L , Olesen S , Rodgers B , Broom D . The psychosocial quality of work determines whether employment has benefits for mental health: results from a longitudinal national household panel survey . Occup Environ Med . 2011 ;68 : 806 –12 . 10.1136/oem.2010.059030 21406384 51 Emberson JR , Whincup PH , Walker M , Thomas M , Alberti KGMM. Biochemical measures in a population-based study: effect of fasting duration and time of day . Ann Clin Biochem . SAGE Publications; 2002 ;39 : 493 –501 . 12227856 52 Bansal S , Buring JE , Rifai N , Mora S , Sacks FM , Ridker PM . Fasting compared with nonfasting triglycerides and risk of cardiovascular events in women . JAMA . American Medical Association; 2007 ;298 : 309 –16 . 17635891

PMC005xxxxxx/PMC5001720.txt

==== Front PLoS Negl Trop DisPLoS Negl Trop DisplosplosntdsPLoS Neglected Tropical Diseases1935-27271935-2735Public Library of Science San Francisco, CA USA 2756423210.1371/journal.pntd.0004968PNTD-D-16-00640Research ArticleBiology and life sciencesOrganismsVirusesRNA virusesFlavivirusesZika VirusBiology and Life SciencesMicrobiologyMedical MicrobiologyMicrobial PathogensViral PathogensFlavivirusesZika VirusMedicine and Health SciencesPathology and Laboratory MedicinePathogensMicrobial PathogensViral PathogensFlavivirusesZika VirusBiology and Life SciencesOrganismsVirusesViral PathogensFlavivirusesZika VirusBiology and Life SciencesEcologyEcological NichesEcology and Environmental SciencesEcologyEcological NichesMedicine and Health SciencesEpidemiologyDisease VectorsInsect VectorsMosquitoesBiology and Life SciencesOrganismsAnimalsInvertebratesArthropodaInsectsMosquitoesPeople and placesGeographical locationsNorth AmericaUnited StatesEarth SciencesAtmospheric ScienceClimatologyClimate ChangePeople and placesGeographical locationsNorth AmericaBiology and Life SciencesEvolutionary BiologyEvolutionary SystematicsPhylogeneticsBiology and Life SciencesTaxonomyEvolutionary SystematicsPhylogeneticsComputer and Information SciencesData ManagementTaxonomyEvolutionary SystematicsPhylogeneticsEarth sciencesAtmospheric scienceClimatologyEl Niño-Southern OscillationEarth sciencesMarine and aquatic sciencesOceanographyEl Niño-Southern OscillationAn Ecological Assessment of the Pandemic Threat of Zika Virus The Ecological Niche of Zika VirusCarlson Colin J. 1*Dougherty Eric R. 1Getz Wayne 121 Department of Environmental Science, Policy, and Management, University of California, Berkeley, Berkeley, California, United States of America2 School of Mathematical Sciences, University of KwaZulu-Natal, Durban, South AfricaJohansson Michael A EditorCenters for Disease Control and Prevention, UNITED STATESThe author have declared that no competing interests exist. Conceptualization: CJC. Data curation: CJC. Formal analysis: ERD CJC. Methodology: CJC ERD. Software: CJC ERD. Validation: ERD CJC. Visualization: ERD CJC. Writing - original draft: CJC ERD WG. Writing - review & editing: CJC ERD WG. * E-mail: cjcarlson@berkeley.edu26 8 2016 8 2016 10 8 e00049687 4 2016 9 8 2016 © 2016 Carlson et al2016Carlson et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.The current outbreak of Zika virus poses a severe threat to human health. While the range of the virus has been cataloged growing slowly over the last 50 years, the recent explosive expansion in the Americas indicates that the full potential distribution of Zika remains uncertain. Moreover, many studies rely on its similarity to dengue fever, a phylogenetically closely related disease of unknown ecological comparability. Here we compile a comprehensive spatially-explicit occurrence dataset from Zika viral surveillance and serological surveys based in its native range, and construct ecological niche models to test basic hypotheses about its spread and potential establishment. The hypothesis that the outbreak of cases in Mexico and North America are anomalous and outside the native ecological niche of the disease, and may be linked to either genetic shifts between strains, or El Nino or similar climatic events, remains plausible at this time. Comparison of the Zika niche against the known distribution of dengue fever suggests that Zika is more constrained by the seasonality of precipitation and diurnal temperature fluctuations, likely confining autochthonous non-sexual transmission to the tropics without significant evolutionary change. Projecting the range of the diseases in conjunction with three major vector species (Aedes africanus, Ae. aegypti, and Ae. albopictus) that transmit the pathogens, under climate change, suggests that Zika has potential for northward expansion; but, based on current knowledge, our models indicate Zika is unlikely to fill the full range its vectors occupy, and public fear of a vector-borne Zika epidemic in the mainland United States is potentially informed by biased or limited scientific knowledge. With recent sexual transmission of the virus globally, we caution that our results only apply to the vector-borne transmission route of the pathogen, and while the threat of a mosquito-carried Zika pandemic may be overstated in the media, other transmission modes of the virus may emerge and facilitate naturalization worldwide. Author Summary A combination of media attention and the declaration of a World Health Organization state of emergency have made the pandemic expansion of Zika virus a topic of great public concern. Understanding the threat North America faces from the still-expanding viral range requires an understanding of the historical range and ecology of the disease, a topic currently difficult to study due to incomplete occurrence data. We compile the most comprehensive geospatial dataset of Zika occurrences in its native range, beginning with its discovery in 1947, and build bioclimatic models that set an outer bound on where the virus is likely to persist. Our results suggest Zika is likely far more constrained than the closely-related dengue fever, on which many projections have been based. While Zika poses a serious threat in current outbreak regions and is clearly a high-priority neglected tropical disease, our models suggest that even under an extreme climate change scenario for 2050, the disease is unlikely to become cosmopolitan in most temperate regions as a vector-borne disease, a discrepant finding from the results of non-ensemble modeling methods. Despite that, sexual transmission remains a serious public health concern, and a route by which Zika could become a severe public health emergency in temperate zones, including in the United States. The authors received no specific funding for this work. Data AvailabilityOur data are currently accessioned on a preprint on the bioRxiv server, available at http://dx.doi.org/10.1101/040386Data Availability Our data are currently accessioned on a preprint on the bioRxiv server, available at http://dx.doi.org/10.1101/040386 ==== Body Introduction Following a twenty-fold upsurge in microcephalic newborns in Brazil linked to Zika virus (ZIKV), the World Health Organization has declared an international health emergency. [1] Despite being profiled for the first time in 1947. [2] Zika remained poorly characterized at a global scale until the last six months. Thus, the present pandemic expansion in the Americas poses a threat of currently unknown magnitude. Closely related to dengue fever, Zika conventionally presents as a mild infection, with 80% of cases estimated to be asymptomatic. [3] The cryptic nature of infection has resulted in sporadic documentation of the disease and rarely includes spatially explicit information beyond the regional scale. [1, 4–6] This greatly limits the confidence with which statistical inferences can be made about the expansion of the virus. With an estimated 440,000–1,300,000 cases in Brazil in 2015, [3] and continuing emergence of new cases in Central America and, most recently, the United States, assessing the full pandemic potential of the virus is an urgent task with major ramifications for global health policy. Current evidence portrays the global spread of ZIKV as a basic diffusion process facilitated by human and mosquito movement, a hypothesis supported by the frequency of infected traveler case studies in the Zika literature. [7–10] Tracing phylogenetic and epidemiological data has revealed the expansion of ZIKV has occurred in a stepwise process through the South Pacific, moving the disease from Southeast Asia into French Polynesia and the Philippines, and subsequently to Easter Island. [1, 4–6] Based on phylogenetic reconstruction, ZIKV is assumed to have dispersed into South America as recently as three years ago from the last of those locations, [11] and the virus is presumed to be at a biogeographic disequilibrium in the Americas. With cases in the ongoing outbreak in Colombia, El Salvador, Guatemala, Paraguay, and Venezuela, and by November of last year, as far north as Mexico and Puerto Rico, the full potential distribution of the disease remains unknown. Moreover, several alternative explanations for the disease’s expansion remain overlooked; most notably, the role of climate change in Zika’s expansion has not yet been thoroughly investigated. [12] We present three competing hypotheses that describe the path of expansion that Zika could take, based on evaluations of the ecological niche of the virus within and outside of its vectors. First, if Zika has no additional climatic constraints relative to those of its vectors, future range expansions should match mosquito ranges. Second, if Zika has a transmission niche that is constrained by climatic factors within the ranges of its mosquito vectors, its range may be much more limited—with, as we show below, possible confinement to the tropics—and cases in North America could be driven by human dispersal or extreme episodic weather events. Finally, it is possible that the expansion of Zika into North America may be a steady range expansion beyond the known niche in its native range, facilitated by climatic shifts or by genetic shifts in the virus or vectors. To test these hypotheses, we present a spatially explicit database of Zika occurrences from the literature and an ensemble of ecological niche models [13] using that data to map the potential distribution of the virus. Methods Occurrence Data Occurrence data for Zika virus was compiled from the literature from studies dating as far back as the original discovery of the virus in Zika Forest, Uganda in 1947. While the asymptomatic nature of the virus limits the total availability of data, lack of evidence for spatial patterns in symptoms in the native range suggest this is an unlikely cause of spatial bias (and instead, merely limits total dataset size). Special attention was paid to correctly attributing cases of travelers to the true source of infection. Locality data was extracted from a combination of clinical cases and seropositivity surveys in humans and mosquitoes, and georeferenced using a combination of Google Maps for hospitals and the Tulane University GEOLocate web platform for the remainder, [14] which allows for the attribution of an uncertainty radius to points only identified to a regional level. Four points were georeferenced in the New World but excluded from niche models because a limited sample as small as four points was likely to significantly bias predictions (compared to the necessary number of pseudoabsences in the same region). Thus, sixty points from the Old World were used in the final models presented in our paper after eliminating data from the current outbreak in the Americas. All points included in our dataset had an outer-bound of at most 65 km of uncertainty, with most substantially less. Constraining datasets based on an uncertainty threshold will become more statistically feasible in future studies once more survey data become available. In the present study, we deemed that the additional information gained from each point outweighed the potential impact of the uncertainty on model performance (S1 Table). We note that for similar reasons, we did not subsample our dataset for spatial thinning in our main models, as software packages like spThin allow, [15] due to information-accuracy tradeoffs; and the strong final performance of models (and the correspondence of our predictions for dengue and Aedes species to published “gold standard” niche models) speaks to the appropriateness of the underlying data and variables. Sensitivity analyses in the literature unequivocally suggest that accuracy of the modeling methods we employ plateaus at or near 50 points, justifying the use of a dataset of this size. [16–18] Occurrence data for the other species included in our study were compiled from the literature. For Aedes africanus, we used a dataset of 99 points downloaded from the Global Biodiversity Informatics Facility (www.gbif.org). GBIF’s coverage of Aedes aegypti and Aedes albopictus was deemed to be lacking, so occurrences for those species were taken from the previously published work of Kraemer et al. [19–20] Finally, Messina et al.’s database was used for dengue, [21] as it has been previously published and used with great success to generate a global distribution model. [22] Both of these datasets were reduced down to point-only data (i.e., polygons of occurrence were excluded), leaving 5,216 points for dengue and 13,992 and 17,280 points for Ae. aegypti and Ae. albopictus respectively. A number of other Zika vectors are known from previous reports, including at least a dozen Aedes species, as well as Anopheles coustani, Culex perfuscus, and Mansonia uniformis. [23–24] While we do not include these vectors in this study in order to keep focus on the most likely globally-cosmopolitan Aedes vectors, we note these species could be important in regional patterns of establishment. These species lack the globally comprehensive datasets that dominant arbovirus-vectoring Aedes species have, and require future attention by similarly-dedicated researchers. Ecological Niche Modeling Due to the potentially transient nature of the New World distribution of Zika virus, our model uses presence and 1000 randomly selected pseudo-absence points from the Eurasian, African, and Australian regions where the virus is established. We used the WorldClim data set BIOCLIM at 2.5 arcminute resolution, an aggregated dataset across values from 1950 to 2000, to provide all but one of our climate variables. [25] The BIOCLIM features 19 variables (BIO1-BIO19) that summarize trends and extremes in temperature and precipitation at a global scale. Given the relevance of the normalized difference vegetation index (NDVI) in previous studies of dengue and as a predictor of vector mosquito distributions, [26] we downloaded monthly average NDVI layers for each month in 2014 from the NASA Earth Observations TERRA/MODIS data portal, [27] at a resolution of 0.25 degrees to maintain compatibility with the BIOCLIM layers (0.25 degrees is equivalent to 15 arcminutes). The twelve monthly layers were averaged to provide a single mean NDVI layer. Due to the absence of NDVI data at the necessary resolution associated with many of the historical records (especially prior to 1992), the use of a recent mean NDVI layer was deemed the most pragmatic method of including vegetation in our models. We also make the simplifying assumption that areas of prior presence correspond to areas of current presence, an assumption that allows the use of current NDVI and is relatively standard for the niche modeling literature. Species distribution models were executed using the BIOMOD2 package in R 3.1.1, which produces ensemble species distribution models using ten different methods: general linear models (GLM), general boosted models or boosted regression trees (GBM), general additive models (GAM), classification tree analysis (CTA), artificial neural networks (ANN), surface range envelope (SRE), flexible discriminant analysis (FDA), multiple adaptive regression splines (MARS), random forests (RF), and maximum entropy (MAXENT). [28] The BIOMOD algorithm runs a series of distribution models using training data, each of which is subsequently weighted and stacked across methods based on relative predictive performance with test data. As Thuiller et al. note, if a single modeling method is consistently most accurate, use of that method should be favored over ensemble approaches, [28] but in our study model performance varied, making ensemble approaches informed by degree-of-belief in a given model the most powerful option available. With recent publication of two Zika niche modeling papers using MAXENT and boosted regression trees, respectively, [29–30] differences between these two modeling methods may be responsible for differences in predictions–an issue that makes ensemble models particularly robust to idiosyncrasies of any individual methods. Models were run individually for Zika (ZIKV), dengue (DENV), Ae. aegypti, Ae. albopictus, and Ae. africanus. For Zika, models trained on Old World environmental data (from Europe, Africa, Asia and Australia) were used to establish the potential distribution of the virus in the Americas under climatic conditions captured by WorldClim data, which are an aggregate of data between 1950 and 2000 (appropriately matching the date range of historical Zika occurrence data), and represent an expected range of variability that does not incorporate anomalous events like 2015 El Niño Southern Oscillation. Extrapolation between continents is a procedure with the potential for error: if novel environments exist in the New World with incomparable covariance structure between climate variables, predictive accuracy is likely to decline. While using only Old World data could potentially bias our models towards a subset of the niche, this can be readily tested for, by comparing models that include or exclude South American occurrence data. To address colinearity in the environmental variable set, we produced a correlation matrix for our 20 variables, and identified each pair with a correlation coefficient > 0.8. For each species, we ran a single ensemble model with all ten methods and averaged the variable importance for our 20 predictors across the methods (S2–S6 Tables). In each pair we identified the variable with the greater contribution, and we produced species-specific reduced variable sets used in the final published models by eliminating any covariates that universally performed more poorly than their pair-mate. Based on this criterion, we excluded the following variables for each species to reduce colinearity: ZIKV: BIO8, BIO9, BIO14, BIO18 DENV: BIO3, BIO5, BIO12, BIO17 Ae. aegypti: BIO6, BIO8, BIO12, BIO17 Ae. africanus: BIO5, BIO6, BIO12, BIO17 Ae. albopictus: BIO8, BIO9, BIO16, BIO17 The AUC of every model run with reduced variable sets is presented in S7 Table. We found no significant correlation between NDVI and any individual BIOCLIM variable, so NDVI was included in every model of current distributions. We ran five iterations of each reduced variable set model and eliminated any prediction methods from the ensemble with an AUC of lower than 0.95, so that the final model had only included the best predicting models. This greatly limited the models available for ZIKV and DENV, so a cutoff of 0.9 was applied in those cases, to keep the ensemble approach constant across datasets. The final models were run with the following methods with ten iterations using an 80/20 training-test split in the final presentation: ZIKV: GLM, GBM, GAM, CTA, FDA, MARS, RF DENV: GLM, GBM, GAM, FDA, MARS, RF, MAXENT Ae. aegypti: GLM, GBM, GAM, CTA, ANN, FDA, MARS, RF Ae. africanus: GLM, GBM, GAM, CTA, ANN, FDA, MARS, RF Ae. albopictus: GLM, GBM, GAM, CTA, FDA, MARS, MAXENT, RF The importance of variables of the reduced model set for each are presented in S8–S13 Tables, and the final ensemble models are projected from the BIOMOD output in S1–S5 Figs. Model Validation To assess the transferability of our Zika model across environmental space, we conducted a geographic cross validation (GCV) between African and Asian datasets (an analysis we did not repeat for Aedes species or dengue, given the far greater sample size and geographic coverage of those species, and the publication of more intensive niche modeling efforts by experts for those systems). While under normal circumstances, a model would be trained on New World data and projected onto the Old World to cross-validate results, the lack of data prior to the current outbreak makes such a direct comparison infeasible. However, given the evidence for separate Asian and African strains, a cross-validation between the two was supported, and models trained on those two continents were projected globally to test the performance of the model across geographic regions, and evaluate how sensitive our projections in the Americas are to the environmental covariates sampled. The clustering of points in western India narrows the environmental range sampled by presences, potentially limiting the apparent transferability of the Asian sub-model. In contrast, the African sub-model performs well in new regions, and corresponds well to the global model. Climate Change Projections The potential contribution of climate change to Zika’s current expansion, and the outer bounds of transmission under future expansion, are largely unaddressed. While these have not been the subject of any concerted speculation, Shapshak et al. [31] point out that the majority of arboviruses are potentially implicated in the climate change-driven expansion of global disease burden, with a shared set of drivers that quite probably extends to Zika as well. Consequently this analysis serves two purposes; to address the potential expansion and thereby assist public health planning, and to test whether even a liberal post-climate-change interpretation of range margins matches the predictions of Messina et al. [29] and Samy et al. [30] that we consider limited in specificity and potentially over-predictive. To project the distribution of the species under a worst-case scenario for climate change, we reran each model with the previously chosen method and variable sets but excluded NDVI, as future values could not be simulated effectively. BIOCLIM forecasts were taken from WorldClim using the Hadley Centre Global Environmental Model v. 2 Earth System climate forecast (HadGEM2-ES) predictions for representative climate pathway 8.5 (RCP85), which, within that model, represents a worst-case scenario for carbon emissions and climate warming. [32] All five species’ models were retrained on current climate data and projected onto forecasts for the year 2050. While we could have also included milder climate change forecasts and scenarios in our analysis, public concern over the future spread of Zika make the worst case scenario the most relevant question of interest for public health research (and intermediate scenarios would fall between current ranges and the worst case scenario we project). Niche Comparison To compare the niche of dengue and Zika and thereby address whether dengue models can be appropriately used to forecast the Zika pandemic, we used the R package ecospat, which uses principal component analysis to define the position of species’ ecological niche relative to background environmental variation. [33–34] The ecospat analysis was run using the full 64 point database and the full extent of global environmental data, because, while the niche of Zika in the Americas is uncertain, dengue is well established, and the analysis was most appropriately done with global coverage. Niche similarity tests were run with 500 iterations and using the entire set of 20 environmental variables (BIOCLIM + NDVI). Model Comparison with Global Data Coverage Our study is centered on the assumption that incorrect predictions at the country level can have drastic consequences for the misinterpretation of science. As a final precautionary analysis, we supplemented the data published in the Messina et al. study [30] to our own for a final re-analysis. Broennimann & Guisan [35] recommend the pooling of data from native and invasive ranges for ecological niche modeling during the course of a biological invasion, an approach we adopt in this final analysis. The Messina data is heavily clustered in Brazil, with a high degree of aggregation, and especially compared against our less-aggregated, smaller dataset this made the combination of datasets potentially inaccurate. To address this problem, the 390 pooled points were reduced down to 242 points using the package spThin, [15] with a 40km buffer between points (the width of an average grid cell for our environmental data). Models were rerun using the same variable and model set as for the primary Zika model and the results of the analysis are included in the supplementary information as S6 Fig and, with a threshold applied based on the true skill statistic, S7 Fig. The final model performs poorer than our main ensemble (weighted model: AUC = 0.970), and while it more appropriately predicts presences in southern Brazil, it does a far poorer job in the rest of the world, once again most likely due to the relative balance of points even after thinning the dataset. Results Our final Zika model combines seven methods with a variable set chosen from bioclimatic variables and a vegetation index to minimize predictor covariance. The ensemble model performs very well (AUC = 0.993; Fig 1), to a degree that resembles overfitting but is in fact driven by the strength of the ensemble modeling approach (which preferentially weights the best models across iterations, minimizing the error associated with any given high-performing iteration). The model strongly matches most occurrences including the hotspots of Brazilian microcephaly. It also predicts additional regions where Zika is so far unrecorded, but where further inquiry may be desired (in particular, Southern Sudan and the northern coast of Australia). Our model indicates that certain occurrences, like the 1954 report from Egypt and almost all North American cases, are likely outside the stable transmission niche (i.e., persistent over time) of the virus (sensu [36]). Moreover, we note that visual presentation of cases–or, of ecological niche models–at the country level may make the range of the virus appear far larger than our models suggest (see Fig 1). 10.1371/journal.pntd.0004968.g001Fig 1 The global distribution of case reports of Zika virus (1947 to February 2016) broken down by country (yellow shading) and an ensemble niche model built from occurrence data (red shading). Our model correspond well to shaded countries, with only minor discrepancies (Paraguay, the Central African Republic; a single case in Egypt in the 1950s), We emphasize that displaying cases at country resolution overstates the distribution of the virus, especially in the Americas (for example, Alaska, a point of significant concern given Messina et al.’s presentation of their niche model in terms of “highly suitable” countries with broad geographic expanse like the United States, China, and Argentina. Given the public health crisis posed by Zika, and the potential costs associated with underpredicting the extent of the current outbreak, we pay special attention to evaluating the sensitivity of our models to variations in our preliminary dataset. Historical geographical data on cases in the Americas are lacking, given the recent introduction of the virus, and the routes and drivers of transmission involved in that outbreak are uncertain, preventing meaningful cross-validation of models of the current outbreak with our Old World model. However, it is worth noting that recent phylogenetic work suggests a deep phylogenetic division between African and Asian strains, the latter of which as a monophyletic group include the entire radiation through French Polynesia into current outbreak areas; [11, 37] to address the potential evidence that African and Asian strains of the virus may be ecologically distinct, we present models trained on each continent and projected globally as a basic sensitivity analysis (Fig 2). 10.1371/journal.pntd.0004968.g002Fig 2 Geographical cross validation of (a) the sub-model built from occurrences on the African continent (n = 27) as projected upon the global climate space and (b) the sub-model built from occurrences on the Asian continent (n = 33) projected at the global scale. The two models cross-validate weakly compared to the performance of the global model; driven by both the 50% reduction in sample size and the higher degree of aggregation of Asian occurrences, the two projected distributions are dramatically different. Despite the over-prediction of the Asian model in Africa and the possible overfitting of the African model, we emphasize that neither extreme scenario predicts any substantially greater range in North America than our main ensemble model. Moreover, our Asian model underpredicts but does predict two major hotspots of occurrence in Brazil, the Ceara/Rio Grande do Norte region and Roraima, both of which spatially correspond to hotspots of Zika according to the recent Faria et al. publication in Science, [11] adding further support to the model. Finally, despite low transferability between continents, both sub-models are well matched by our aggregated model in their native range, further supporting the accuracy and predictive power of our global projection. Recently published work by Bogoch et al. [38] uses an ecological niche model for dengue as a proxy for the potential full distribution of ZIKV in the Americas, presenting findings in terms of potential seasonal vs. full-year transmission zones. While that approach has been effectively validated for dengue transmission in mosquitoes, using a model of one disease to represent the potential distribution of another emerging pathogen is only a placeholder, and is particularly concerning given the lack of evidence in our models that ZIKV and dengue have a similar niche breadth. [39] Comparing our niche models for dengue and ZIKV reveals that the two niches are significantly different (Schoener’s D = 0.176; p < 0.01; Fig 3). While the two occupy a similar region of global climate space, Zika is more strictly tropical than dengue, occupying regions with higher diurnal temperature fluctuations and seasonality of precipitation (Fig 3A). 10.1371/journal.pntd.0004968.g003Fig 3 The ecological niche of Zika and dengue in principal component space (a). Solid and dashed lines are 100% and 50% boundaries for all environmental data, respectively. Despite apparent overlap in environmental niche space, the dissimilarity between the black shading in each principal component graph indicates statistically significant differences between the niches, evident in the projections of our niche models for dengue (b) and Zika (c). Projecting niche models to the year 2050 suggests that expansion of Zika’s niche outside the tropics is an unlikely scenario, independent of vector availability (Fig 4). However, significant westward expansion in South America and eastward expansion in Africa implies that Zika may continue to emerge in the tropics. Moreover, our future projections for dengue (which strongly agree with previously published ones [40]) show an expansion out of the tropics that is not shared with Zika (Fig 4). These results call into question the applicability of dengue niche models used to project a significant future range for Zika in North America. [38] 10.1371/journal.pntd.0004968.g004Fig 4 The estimated global distribution of Zika (red) and dengue (blue) based on current (a, b) and 2050 climate projections (c, d), compared against the current (light grey) and future distribution (dark grey) of all three mosquito vectors Aedes aegypti, Ae. africanus and Ae. albopictus (a-d). Finally, we add a last layer of validation in the form of an analysis aggregating our and Messina et al.’s data, and include the results of an updated ensemble model in Fig 5 (as well as S6 and S7). Even with spatial thinning, that updated model is still heavily biased in favor of the South American occurrence data, which it predicts excellently, compared to a weaker fit in Africa and Asia. That accompanying loss of specificity is partly responsible for a lower AUC than our main model (AUC = 0.970) and the low TSS-based threshold (271, from 0 to 1000) that produces the substantially-greater predicted range shown in S7 Fig. The model does predict the current outbreak more effectively than ours, in particular better encompassing the southern half of Brazil where a surprising number of cases are clustered. But those southward expansions are accompanied by far less expansion above the equator in the Americas, and once again with the exception of the southernmost tip of Florida, there is no substantial predicted range in the United States, even along the Gulf Coast. If model discrepancies are attributed to evolutionary change and not to differences in model methods and specificity, those evolutionary changes seem to have done little to expand the North American niche of the virus (S8 Fig). 10.1371/journal.pntd.0004968.g005Fig 5 An updated ecological niche model incorporating aggregated global data, with Messina et al.’s full dataset (red) and ours (blue) against the updated weighted ensemble model. Discussion Ecological niche modeling has become one of the most generalized and useful parts of the streamlined response process for emerging infections. Recently published ecological niche models for Zika using MAXENT [30] and boosted regression trees [29] have resulted in somewhat conflicting results. Samy et al., using data exclusively from the range of the current outbreak, project autochthonous transmission in the southeastern United States, and potentially throughout the U.S. following regional outbreaks introduced by travelers. Their analysis incorporates socioeconomic factors into prediction, a valuable extra dimension we did not incorporate into our analysis; but the prediction of regions throughout the United States and most of the European continent as suitable based on only these criteria (i.e. despite lacking available vectors) seems uninformative except for the prediction of sexual outbreaks. Samy et al., however, conclude: “In Western Europe, ZIKV transmission risk is enhanced by travel times and connectivity to known transmission areas; as such, isolated autochthonous cases may occur at least seasonally when competent vector species are present.” [30] Messina et al. have a similar finding, based on a primarily ecological approach applied to 323 occurrences mostly from the New World; they map out most countries in the world as highly suitable, including the United States, with the conclusion that 2.17 billion people live in countries within Zika’s potential expanse. [29] These studies, being contemporaneous, do not refer to each other, and their conflicting results could render Zika forecasts unclear to the media and policymakers. Interpreting conflicts between these models and those published here requires acknowledging three fundamental problems. First, differences in virulence between American and Asian strains of the virus may have changed the range limits. The niche of the vector-borne disease is manifest in its transmission and prevalence in mosquitoes (as well as humans and reservoirs), and increases in virulence could change the threshold of habitat suitability manifest in range limits. Without comparative work using updated data in Samy et al. and Messina et al.’s papers, equal support exists for our differences being attributable to methodological discrepancies or to a difference between Asian and American strains. But in the preliminary analysis we present in the supplementary information, incorporating data from the New World does not substantially expand projections in the United States (though a greater region of Brazil is predicted); and we believe a combination of evolutionary shifts and methodological differences is likely the most parsimonious explanation for differing results. Second, we acknowledge the untested possibility that Zika has been expanding in its range since discovery in the 1940s (though, the virus was soon recorded in Borneo and Vietnam in the 1950s [23]), which would also decrease both the accuracy of our models in that region, and their power in the New World compared to the models published in the other two studies. Testing that possibility using our data broken down by time periods would be strongly statistically biased by the non-random element of viral discovery in different tropical countries, a factor for which it would be nearly impossible to control. Phylogenetic evidence has placed the introduction in the Americas within the last decade [11], but the age of divergence between Zika and closely related viruses like Japanese and St. Louis Encephalitis Viruses is less certain. Improving phylogenetic evidence based on updated Old World genomes in the coming years is a far more appropriate methodology for testing different biogeographic theories within that region. Third and finally, we acknowledge the possibility that dispersal limitations have changed between the Old and New World, in such a way that the present expansion of Zika is not the emergence of novel niche space but the manifestation of hidden plasticity. This possibility is troubling from a public health perspective: if Zika’s niche is simply more expansive than current data/models capture, its geographic expansion could progress much further than we predict. This problem is fundamental to all predictive models applied to biological invasions, but Broenniman & Guisan [35] suggest that combining data from the native and invasive range maximizes the utility of ENMs in these scenarios. In our combined model we find evidence for subtle differences, especially in South America, but our findings remain sound with respect to the boundaries of transmission in North America. In any niche modeling study, there is always the possibility for error by omission; but we find no evidence that this has occurred in our study. The dynamics of arboviruses at the range margins of their vectors are complex. In the case of dengue, the distribution of the virus in the United States (and elsewhere in temperate regions) remains more constrained than the range of its vectors. Our paper tests and rejects the hypothesis that predictions of Zika will occupy the entire niche of Aedes populations in North America, disagreeing with the two recently published niche model studies. Our models imply a similar constraint on Zika transmission to that of dengue if not a more pronounced one, and owing to the complexities surrounding transmission dynamics at the edges of suitable ranges, [41] the potential existence of Zika in even the southernmost parts of Florida [42] may not sustain autochthonous Zika transmission indefinitely. Making more specific predictions within Florida can be done through ecological niche models, but is likely more appropriately achieved through conventional epidemiological models that explicitly model vector abundance, biting rates and phenology. Our models find an ecological nonequivalence of Zika and dengue, and suggest that the niche of the virus in both Africa and Asia is far narrower than what other models project based on current outbreak data or based on knowledge of dengue’s spread. We reject our first hypothesis, but based on the occurrence of Zika cases outside our predicted suitable range for the virus, we cannot eliminate our second hypothesis that the 2016 Zika outbreak may be in ephemeral, rather than stable, parts of the Zika transmission niche due to episodic climatic conditions. Specifically, El Niño Southern Oscillation (ENSO) events drive outbreaks of dengue in the Americas and in Southeast Asia, [43] and Paz et al. [12] have conjectured that the 2015 ENSO event could have contributed to the severity of the ZIKV outbreak in North and Central America (in response to Bogoch et al. [38]). While wind-dispersed mosquitoes carrying infections can be responsible for the introduction of diseases to new regions, [44] reported cases in the United States have all been contracted sexually or while traveling abroad to regions with endemic outbreaks, further supporting the tropical constraint hypothesis. However, in the second hypothesis scenario, the rapid expansion during the current outbreak beyond the boundaries of the stable transmission niche is unlikely to be followed by naturalization of the pathogen in the United States in the future, except perhaps in the southernmost tip of Florida. While ecological niche models relate occurrence to climate, drivers of disease may operate at the temporal scale of weather, and we suggest further analyses of a different methodology are necessary to confirm or reject the potential contribution of El Nino or anomalous storms to Zika’s expansion. In the case of our third hypothesis, if alternative modeling efforts based on data from the Americas are evidence that the niche of the American strain of the virus has broadened, it is possible that mutations allowing increased virulence or changing transmission dynamics have occurred (and that weather events have not driven the severity of the current outbreak). From the results of our supplementary analysis using aggregated global data, we continue to treat the third hypothesis as a hypothesis for which there may be weak evidence. But we suggest it cannot be rejected or accepted confidently unless alternative hypotheses are eliminated and more evidence is collected–in particular, empirical data demonstrating or failing to find differences in transmission dynamics or virulence between the native Asian virus and its invasive descendant (rather than global comparisons and cross-validations of different ecological niche models). Our models nevertheless suggest it could be premature to expect Zika naturalization as a widespread eventuality in North America, as other models have forecasted. Without more definitive information on the basic biology of Zika, however, the confidence with which niche models can forecast pandemics is limited. In particular, we also draw attention to recent evidence suggesting Zika persistence may depend on wildlife reservoirs in addition to human hosts and mosquitoes. Primates have been suggested as the primary candidate clade because the Zika flavivirus was first isolated in a rhesus macaque in the Zika Forest in Uganda. But as rhesus macaques do not occur on the African continent, and were captive there for inoculation experiments, the primate reservoir hypothesis remains unsupported. A 2015 case of an Australian presumed to have contracted Zika from a monkey bite while traveling in Indonesia, however, indicates that primates may transmit the virus directly. [9] Additionally, antibodies against Zika have been observed in several rodent and livestock species in Pakistan, [45] as well as several large mammal species, including orangutans, zebras, and elephants. [46] The potential for any North American wildlife species to play host to Zika is, at the present time, entirely unknown, and the emergence of novel amplification hosts (which may allow the virus to proliferate above the host density threshold in vectors in regions otherwise unsuitable for sustained transmission) could potentially expand the suitable range margins of Zika infection on a global scale. From the results of our model we find strong evidence for the hypothesis that the global threat of a specifically vector-borne Zika pandemic, though devastating, may be most acute in the tropics; and we find that the evidence of future North American transmission in the literature is not unequivocal. However, we concur with the scientific majority that sexual transmission of Zika infections may still facilitate a significant outbreak in the United States and other previously unsuitable regions, particularly under evolutionary processes that select for the most directly transmissible strains of pathogens. [47] A case of sexual transmission in Texas has been suspected in the 2016 outbreak, and two previous reports of likely sexual transmission of ZIKV occurred in 2011 and 2015. [5, 48] Even if the Zika cases in the United States represent a rare spillover outside of the mosquito-borne viral niche, sexual transmission could create a new, unbounded niche in which the virus could spread. We draw attention to the potential parallels with simian and human immunodeficiency virus (SIV/HIV), for which a sexually transmitted pandemic has overshadowed the zoonotic origin of the disease. [49] With Zika’s asymptomatic presentation and the overall confusion surrounding its basic biology and transmission modes, we caution that its potential for severe sexually-transmitted outbreaks cannot be overlooked in the coming months. To address the broader community of modelers and ecologists involved in the Zika intervention, we conclude with a final cautionary note. The consequences of under-predicting an outbreak’s potential distribution are obvious and our results are phrased cautiously as a result. But there are also economic and social consequences to over-predicting the potential distribution, especially in the United States. The response to Zika is necessarily political and consequently involves the division of resources between domestic preparedness and international relief; while new tools are being developed to help allocate funds efficiently based on epidemiological principles (we particularly highlight the work of Alfaro-Murillo et al. [50]), global overestimation of the virus’s trajectory could vastly reduce the power of those methods. Models like those of Messina et al. and Samy et al. that predict substantial Zika expansion in the United States, and in the case of the former suggest Zika could threaten up to 2.17 billion people, contribute (independent of accuracy) to fear of an American pandemic. This prediction necessarily diverts funding away from relief efforts in Brazil and other affected countries in Latin America, increasing the probability of traveler infections feeding sexual outbreaks in the U.S.; and further reduces the credibility and impact of the American foreign response to Zika by mobilizing potentially-unnecessary domestic responses. At the time of writing, the Zika Vector Control Act passed by the U.S. House of Representatives weakens permit requirements for spraying pesticides near bodies of water without reallocating any funding for Zika interventions; and preventative efforts in New York City alone will cost $21 million to trap mosquitoes and hire epidemiological experts, with other cities outside our predicted range investing in preparation and vector control to similar degrees. Voices of scientific authority contributing to fear in the United States can substantially impact the political response to Zika, and it serves future modeling efforts to be as accurate, cautious, and objective as possible in the information and statistics that underpin media and policy conversations. But even more importantly, scientific teams with different approaches and data must work collaboratively to interpret the discrepancies between their results and to build an unbiased scientific consensus that is accessible to the public. Supporting Information S1 Table Global occurrence database for Zika virus. A dataset containing the country, locality string used for geo‐referencing, latitude and longitude (in decimal degrees), uncertainty radius, comments, and the reference from which the data were obtained, followed by an exhaustive reference list. (PDF) Click here for additional data file. S2 Table Zika full variable set preliminary model variable importance. Variable contributions are based on one preliminary run with 20 variables and 10 candidate models. (PDF) Click here for additional data file. S3 Table Dengue full variable set preliminary model variable importance. Variable contributions are based on one preliminary run with 20 variables and 10 candidate models. (PDF) Click here for additional data file. S4 Table Aedes aegypti full variable set preliminary model variable importance. Variable contributions are based on one preliminary run with 20 variables and 10 candidate models. (PDF) Click here for additional data file. S5 Table Aedes africanus full variable set preliminary model variable importance. Variable contributions are based on one preliminary run with 20 variables and 10 candidate models. (PDF) Click here for additional data file. S6 Table Aedes albopictus full variable set preliminary model variable importance. Variable contributions are based on one preliminary run with 20 variables and 10 candidate models. (PDF) Click here for additional data file. S7 Table AUC of ten models for five species (with reduced variable sets). Bolded models were shown in the final models. Updated Zika model incorporating New World outbreak data included as “ZIKV+”. (PDF) Click here for additional data file. S8 Table Zika final model variable importances. The final ensemble model includes seven modeling methods using sixteen variables, each run for 10 iterations. (PDF) Click here for additional data file. S9 Table Dengue final model variable importances. The final ensemble model includes eight modeling methods using sixteen variables, each run for 10 iterations. (PDF) Click here for additional data file. S10 Table Aedes aegypti final model variable importances. The final ensemble model includes eight modeling methods using sixteen variables, each run for 10 iterations. (PDF) Click here for additional data file. S11 Table Aedes africanus final model variable importances. The final ensemble model includes eight modeling methods using sixteen variables, each run for 10 iterations. (PDF) Click here for additional data file. S12 Table Aedes albopictus final model variable importances. The final ensemble model includes eight modeling methods using sixteen variables, each run for 10 iterations. (PDF) Click here for additional data file. S13 Table Variable importance in supplementary ZIKV+ model. (PDF) Click here for additional data file. S1 Fig Final ensemble model for Zika virus. (TIF) Click here for additional data file. S2 Fig Final ensemble model for dengue fever. (TIF) Click here for additional data file. S3 Fig Final ensemble model for Aedes aegypti. (TIF) Click here for additional data file. S4 Fig Final ensemble model for Aedes africanus. (TIF) Click here for additional data file. S5 Fig Final ensemble model for Aedes albopictus. (TIF) Click here for additional data file. S6 Fig Expanded niche model with global data coverage. (TIF) Click here for additional data file. S7 Fig Expanded niche model with threshold. (TIF) Click here for additional data file. S8 Fig Niche Overlap Analysis between Dengue and Global Zika Database. In the equivalency test, we find significant evidence for differences (Schoener’s D = 0.295; p = 0.004). (TIF) Click here for additional data file. C.J.C. thanks Fausto Bustos for feedback on initial ideas presented in the paper, and Kevin Burgio for extensive methodological support, training, and mentorship. We also thank four anonymous reviewers for their aid in strengthening the manuscript. A preprint of this manuscript was made available at http://biorxiv.org/content/early/2016/02/19/040386 on February 19, 2016. ==== Refs References 1 Attar N. Zika virus circulates in new regions . Nature Rev Microbiol . 2016 ; 14 : 62 . 2 Dick GWA , Kitchen SF , Haddow AJ . Zika virus (I). Isolations and serological specificity . Trans R Soc Trop Med Hyg . 1952 ; 46 : 509 –520 . 12995440 3 Hennessey M . Zika virus spreads to new areas—region of the Americas, May 2015-January 2016 . CDC Morbidity and Mortality Report 2016 ; 65 : 55 –58 . 4 Musso D , Nilles EJ , Cao-Lormeau VM . Rapid spread of emerging Zika virus in the Pacific Area . Clin Microbiol Infec . 2014 ; 20 : O595 –O596 .24909208 5 Musso D , Cao-Lormeau VM , Gubler DJ . Zika virus: following the path of dengue and chikungunya? Lancet 2015 ; 386 : 243 –244 . 10.1016/S0140-6736(15)61273-9 26194519 6 Gatherer D . Zika virus: a previously slow pandemic spreads rapidly through the Americas? J Gen Virol . 2015 ; 97 : 269 –273 . 10.1099/jgv.0.000381 26684466 7 Summers DJ , Acosta RW , Acosta AM . Zika virus in an American recreational traveler . J Travel Med . 2015 ; 22 : 338 –340 . 10.1111/jtm.12208 25996909 8 Fonseca K . Case report: first case of Zika virus infection in a returning Canadian traveler . Am J Trop Med Hyg . 2014 ; 91 : 1035 –1038 . 10.4269/ajtmh.14-0151 25294619 9 Kwong JC , Druce JD , Leder K . Case report: Zika virus infection acquired during brief travel to Indonesia . Am J Trop Med Hyg . 2013 ; 89 : 516 –517 . 10.4269/ajtmh.13-0029 23878182 10 Zammarchi L . Zika virus infection in a traveler returning to Europe from Brazil, March 2015 . Euro Surveill . 2015 ; 20 : pii-21153. 11 Faria NR , Azevedo RSS , Kraemer MUG , Souza R , Cunha MS , et al Zika virus in the Americas: Early epidemiological and genetic findings . Science 2016 ; 352 : 345 –349 . 10.1126/science.aaf5036 27013429 12 Paz S , Semenza JC . El Niño and climate change–contributing factors in the dispersal of Zika virus in the Americas? Lancet 2016 ; 387 : 745 . 13 Peterson AT . Ecological niche modeling and spatial patterns of disease transmission . Emerg Infect Dis . 2006 ; 12 : 1822 –1826 . 17326931 14 GEOLocate. http://www.museum.tulane.edu/geolocate/web/WebGeoref.aspx. Accessed 2/7/2016. 15 Aiello-Lammens ME , Boria RA , Radosavljevic A , Vilela B , Anderson RP . spThin: an R package for spatial thinning of species occurrence records for use in ecological niche models . Ecography 2015 ; 38 : 541 –545 . 16 Stockwell DRB , Peterson AT . Effects of sample size on accuracy of species distribution models . Ecol Mod . 2002 ; 148 : 1 –13 . 17 Hernandez PA , Graham CH , Master LL , Albert DL . The effect of sample size and species characteristics on performance of different species distribution modeling methods . Ecography 2006 ; 29 : 773 –785 . 18 Wisz MS , Hijmans RJ , Li J , Peterson AT , Graham CH , et al Effects of sample size on the performance of species distribution models . Div. Dist. 2008 ; 14 : 763 –773 . 19 Kraemer MUG , Sinka ME , Duda KA , Mylne AQN , Shearer FM , et al The global distribution of the arbovirus vectors Aedes aegypti and Ae. albopictus . eLife 2015 ; 4 : e08347 10.7554/eLife.08347 26126267 20 Kraemer MUG , Sinka ME , Duda KA , Mylne AQN , Shearer FM , et al The global compendium of Aedes aegypti and Ae. albopictus occurrence . Sci Data . 2015 ; 2 : 150035 10.1038/sdata.2015.35 26175912 21 Messina JP , Brady OJ , Pigott DM , Brownstein JS , Hoen AG , et al A global compendium of human dengue virus occurrence . Sci Data . 2015 ; 1 : 140004 . 22 Bhatt S , Gething PW , Brady OJ , Messina JP , Farlow AW , et al The global distribution and burden of dengue . Nature 2013 ; 496 : 504 –507 . 10.1038/nature12060 23563266 23 Haddow AD , Schuh AJ , Yasuda CY , Kasper MR , Heang V , et al Genetic characterization of Zika virus strains: geographic expansion of the Asian lineage . PLoS Negl Trop Dis . 2012 ; 6 : e1477 10.1371/journal.pntd.0001477 22389730 24 Weaver SC , Costa F , Garcia-Blanco MA , Ko AI , Ribeiro GS , et al Zika virus: history, emergence, biology, and prospects for control . Antiviral Research 2016 ; 130 : 69 –80 . 10.1016/j.antiviral.2016.03.010 26996139 25 Hijmans RJ , Cameron SE , Parra JL , Jones PG , Jarvis A . Very high resolution interpolated climate surfaces for global land areas . Int J Climatol . 2005 ; 25 : 1965 –1978 . 26 Peterson AT , Martínez-Campos C , Nakazawa Y , Martínez-Meyer E . Time-specific ecological niche modeling predicts spatial dynamics of vector insects and human dengue cases . T Roy Soc Trop Med H . 2005 ; 99 : 647 –655 . 27 NASA Earth Observations TERRA/MODIS NDVI. http://neo.sci.gsfc.nasa.gov/view.php?datasetId=MOD13A2_M_NDVI. Accessed 2/7/2016. 28 Thuiller W , Lafourcade B , Engler R , Araujo MB . BIOMOD: a platform for ensemble forecasting of species distributions . Ecography 2009 ; 32 : 369 –373 . 29 Messina JP , Kraemer MUG , Brady OJ , Pigott DM , Shearer FM , et al. Mapping global environmental suitability for Zika virus . eLife 2016 ; 5 : e15272 10.7554/eLife.15272 27090089 30 Samy AM , Thomas SM , El Wahed AA , Cohoon KP , Peterson AT . Mapping the global geographic potential of Zika virus spread [Submitted]. Mem Inst Oswaldo Cruz . E-pub: 15 April 2016. 10.1590/0074-02760160149. 31 Shapshak P, Somboonwit C, Foley BT, Alrabaa SF, Wills T, et al. Chapter 18: Zika virus. In: P. Shapshak et al., editors. Global Virology I–Identifying and Investigating Viral Diseases, 10.1007/978-1-4939-2410-3_18. 32 Jones C , Robertson E , Arora V , Friedlingstein P , Shevliakova E , et al Twenty-first-century compatible CO2 emissions and airborne fraction simulated by CMIP5 Earth System models under four representative concentration pathways . J Climatol . 2013 ; 26 : 4398 –4413 . 33 Broenniman O, Di Cola V, Petitpierre B, Breiner F, D’Amen M, et al. Ecospat: spatial ecology: miscellaneous methods. R package version 1.0. http://www.unil.ch/ecospat/home/menuinst/tools-data/tools.html. (2014) 34 Broenniman O , Fitzpatrick MC , Pearman PB , Petitpierre B , Pellissier L , et al Measuring ecological niche overlap from occurrence and spatial environmental data . Global Ecol Biogeogr . 2012 ; 21 : 481 –497 . 35 Broenniman O , Guisan A . Predicting current and future biological invasions: both native and invaded ranges matter . Biology Letters 2008 ; 4 : 585 –589 . 10.1098/rsbl.2008.0254 18664415 36 Pigott DM , Golding N , Mylne A , Huang Z , Henry AJ , et al Mapping the zoonotic niche of Ebola virus disease in Africa . eLife 2014 ; 3 : e04395 10.7554/eLife.04395 25201877 37 Lanciotti RS , Lambert AJ , Holodniy M , Saavedra S , Signor LCC . Phylogeny of Zika virus in the western hemisphere, 2015 . Emerg Inf Dis . 2016 ; 22 : 933 –935 . 38 Bogoch II , Brady OJ , Kraemer MUG , German M , Creatore MI , et al Anticipating the international spread of Zika virus from Brazil . Lancet 2016 ; 387 : 335 –336 . 10.1016/S0140-6736(16)00080-5 26777915 39 Christofferson RC . Zika virus emergence and expansion: Lessons learned from dengue and chikungunya may not provide all the answers . Am J Trop Med Hyg . in press. 10.4269/ajtmh.15-0866 40 Messina JP , Brady OJ , Pigott DM , Golding N , Kraemer MUG , et al The many projected futures of dengue . Nature Rev Microbiol . 2015 ; 13 : 230 –239 .25730702 41 Eisen L , Moore CG . Aedes (Stegomyia) aegypti in the continental United States: a vector at the cool margin of its geographic range . J Med Entomol . 2013 ; 50 : 467 –478 . 23802440 42 Brady OJ , Gething PW , Bhatt S , Messina JP , Brownstein JS , et al Refining the global spatial limits of dengue virus transmission by evidence-based consensus . PLoS Negl Trop Dis 2012 ; 6 : e1760 10.1371/journal.pntd.0001760 22880140 43 Johansson MA , Cummings DAT , Glass GE . Multiyear climate variability and dengue–El Niño southern oscillation, weather, and dengue incidence in Puerto Rico, Mexico, and Thailand: a longitudinal data analysis . PLoS Medicine 2009 ; 6 : e1000168 10.1371/journal.pmed.1000168 19918363 44 Ritchie SA , Rochester W . Wind-blown mosquitoes and introduction of Japanese encephalitis into Australia . Emerg Infect Dis . 2001 ; 7 : 900 11747709 45 Darwish MA , Hoogstraal H , Roberts TJ , Ahmed IP , Omar F . A sero-epidemiological survey for certain arboviruses (Togaviridae) in Pakistan . T Roy Soc Trop Med Hyg . 1983 ; 77 : 442 –445 . 46 Ioos S , Mallet H-P , Goffart IL , Gauthier V , Cardoso T , et al Current Zika virus epidemiology and recent epidemics . Med Maladies Infect . 2014 ; 44 : 302 –307 . 47 Neumann G , Noda T , Kawaoka Y . Emergence and pandemic potential of swine-origin H1N1 influenza virus . Nature 2009 ; 459 : 931 –939 (2009). 10.1038/nature08157 19525932 48 Foy BD , Kobylinski KC , Foy JLC , Blitvich BJ , da Rosa AT , et al Probable non-vector-borne transmission of Zika virus, Colorado, USA . Emerg Infect Dis . 2011 ; 17 : 880 –882 . 10.3201/eid1705.101939 21529401 49 Gao F , Bailes E , Robertson DL , Chen Y , Rodenburg CM , et al Origin of HIV-1 in the chimpanzee Pan troglodytes troglodytes . Nature 1999 ; 397 : 436 –441 . 9989410 50 Alfaro-Murillo JA , Paripa AS , Fitzpatrick MC , Tamagnan JA , Medlock J , et al A cost-effectiveness tool for informing policies on Zika virus control . PLoS Negl Trop Dis . 2016 ; 10 : e0004743 10.1371/journal.pntd.0004743 27205899

PMC005xxxxxx/PMC5001721.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756455210.1371/journal.pone.0162020PONE-D-16-19448Research ArticleBiology and Life SciencesGeneticsGene ExpressionProtein TranslationBiology and Life SciencesBiochemistryRibosomesBiology and Life SciencesCell BiologyCellular Structures and OrganellesRibosomesBiology and Life SciencesOrganismsBacteriaActinobacteriaMycobacterium TuberculosisBiology and life sciencesBiochemistryNucleic acidsRNANon-coding RNATransfer RNAMedicine and Health SciencesPharmacologyDrugsAntimicrobialsAntibioticsBiology and Life SciencesMicrobiologyMicrobial ControlAntimicrobialsAntibioticsResearch and Analysis MethodsChemical SynthesisBiosynthetic TechniquesProtein SynthesisBiology and Life SciencesBiochemistryProteinsProtein SynthesisResearch and Analysis MethodsSeparation ProcessesCentrifugationBiology and Life SciencesGeneticsGene ExpressionGene RegulationElongation FactorsBiology and Life SciencesBiochemistryProteinsRegulatory ProteinsElongation FactorsReconstitution of Protein Translation of Mycobacterium Reveals Functional Conservation and Divergence with the Gram-Negative Bacterium Escherichia coli Biochemical Reconstitution of Mycobacterial Protein TranslationSrivastava Aashish 1¤Asahara Haruichi 1Zhang Meng 2Zhang Weijia 2Liu Haiying 2Cui Sheng 2Jin Qi 2Chong Shaorong 1*1 New England Biolabs, Inc., 240 County Road, Ipswich, MA, 01938, United States of America2 Key Laboratory of Systems Biology of Pathogens, Institute of Pathogen Biology, Chinese Academy of Medical Sciences & Peking Union Medical College, Beijing, 100176, ChinaChatterji Dipankar EditorIndian Institute of Science, INDIACompeting Interests: Our affiliation with New England Biolabs does not alter our adherence to PLOS ONE policies on sharing data and materials. Conceptualization: S. Chong S. Cui HL QJ. Formal analysis: AS HA S. Cui S. Chong. Funding acquisition: S. Chong. Investigation: AS HA MZ WZ S. Chong. Methodology: AS HA MZ WZ. Resources: S. Chong HL S. Cui QJ. Supervision: S. Chong S. Cui QJ. Writing – original draft: S. Chong AS S. Cui. Writing – review & editing: AS HA S. Cui S. Chong. ¤ Current address: Department of Biochemistry & Biophysics, Texas A&M University, College Station, TX, 77843, United States of America * E-mail: chong@neb.com26 8 2016 2016 11 8 e016202013 5 2016 16 8 2016 © 2016 Srivastava et al2016Srivastava et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Protein translation is essential for all bacteria pathogens. It has also been a major focus of structural and functional studies and an important target of antibiotics. Here we report our attempts to biochemically reconstitute mycobacterial protein translation in vitro from purified components. This mycobacterial translation system consists of individually purified recombinant translation factors from Mycobacterium tuberculosis (M. tuberculosis), purified tRNAs and ribosomes from Mycobacterium smegmatis (M. smegmatis), and an aminoacyl-tRNA synthetase (AARS) mixture from the cell-extract of M. smegmatis. We demonstrate that such mycobacterial translation system was efficient in in vitro protein synthesis, and enabled functional comparisons of translational components between the gram-positive Mycobacterium and the gram-negative E. coli. Although mycobacterial translation factors and ribosomes were highly compatible with their E. coli counterparts, M. smegmatis tRNAs were not properly charged by the E. coli AARSs to allow efficient translation of a reporter. In contrast, both E. coli and M. smegmatis tRNAs exhibited similar activity with the semi-purified M. smegmatis AARSs mixture for in vitro translation. We further demonstrated the use of both mycobacterial and E. coli translation systems as comparative in vitro assays for small-molecule antibiotics that target protein translation. While mycobacterial and E. coli translation were both inhibited at the same IC50 by the antibiotic spectinomycin, mycobacterial translation was preferentially inhibited by the antibiotic tetracycline, suggesting that there may be structural differences at the antibiotic binding sites between the ribosomes of Mycobacterium and E. coli. Our results illustrate an alternative approach for antibiotic discovery and functional studies of protein translation in mycobacteria and possibly other bacterial pathogens. http://dx.doi.org/10.13039/100000865Bill and Melinda Gates Foundation1057809Srivastava Aashish Institute of Pathogen Biology, Chinese Academy of Medical SciencesCui Sheng http://dx.doi.org/10.13039/100004774New England BiolabsChong Shaorong National Science and Technology Major Project of China2013ZX10004601Zhang Meng This study was supported by the Melinda & Bill Gates Foundation, GCE OPP1057809, www.gatesfoundation.org, SCh; National Science and Technology Major Project of China [2013ZX10004601], http://www.most.gov.cn/eng/programmes1, MZ; and Institute of Pathogen Biology, Chinese Academy of Medical Sciences. Drs. Haruichi Asahara and Shaorong Chong are employed by New England Biolabs. This funder provided support in the form of salaries for these authors [HA and S. Chong], but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction The World Health Organization estimates about one-third of the world’s population is infected with Mycobacterium tuberculosis (M. tuberculosis), the bacillus that causes pulmonary tuberculosis (TB). In 2013, >8.6 million people developed TB and 1.3 million people died from the disease [1]. Current multi-drug treatment takes lengthy 6–9 months and poor patient compliance often leads to multi-drug resistant TB [2]. There is an urgent need to develop rapid diagnostic tools and new classes of drugs to combat TB [3]. This calls for deeper understanding of the M. tuberculosis biology and innovative approaches for antibiotic discovery. Protein synthesis is essential for the pathogenesis of M. tuberculosis and an important target of antibiotics [4]. There are significant differences in the size and charge of the mycobacterial ribosomal proteins from those of E. coli, and a recent cryo-EM structure of the 70S ribosome of Mycobacterium smegmatis (M. smegmatis), a non-pathogenic laboratory model of M. tuberculosis, has revealed a number of structural differences when compared to the 70S ribosome of E. coli [5]. These studies suggest potential differences in ribosome functions between Mycobacterium and E. coli and the possibility of designing antibiotics that target only the gram-positive M. tuberculosis. Additionally, a first-line tuberculosis drug, Pyrazinamide, has been recently found to target the ribosomal protein S1 in M. tuberculosis, inhibiting the trans-translation process [6]. Mistranslation of RNA polymerase proteins by the mycobacterial translational machinery has been implicated in the phenotypic resistance to the antibiotic rifampicin [7]. In spite of these recent progresses, the structural and functional studies of M. tuberculosis protein translation is still lacking, hampering the efforts to design new drugs targeting this core biological pathway. We have previously demonstrated a bottom-up approach that biochemically reconstitutes the translation and transcription of E. coli and the translation of Thermus thermophilus for "in vitro genetic" and comparative function studies [8–10]. In this work, we use the same approach to reconstitute mycobacterial protein translation. We expressed recombinant translation factors of M. tuberculosis in E. coli and individually purified these M. tuberculosis proteins. To obtain other mycobacterial translational components, we grew M. smegmatis cells and purified ribosomes, tRNAs and amino acyl-tRNA synthetase (AARS) mixture from the cell extract. The energy regeneration enzymes from the reconstituted E. coli translation system were used to provide ATP and GTP for mycobacterial protein translation. T7 RNA polymerase was used to couple transcription to translation and allow direct use of DNA templates. We demonstrate that such reconstituted mycobacterial translation system not only was efficient in in vitro synthesis of full-length proteins but also allowed functional studies of mycobacterial translation. Since we had both mycobacterial and E. coli translation systems, we compared the translational components between a gram-positive and a gram-negative bacterium by swapping the components between two systems. We also used both systems in parallel to assay species-specific antibiotics that target protein translation. We perceive that our approach has a number of advantages over other methods and represents a significant progress over previous studies. For instance, genetic methods for investigating TB biology require specialized facilities to grow live and infectious M. tuberculosis cells, which can pose health hazards to researchers. The cell-based assays for screening small-molecule inhibitors are not necessarily target-specific, generally dependent on cell-growth, and adversely affected by bacterial efflux pumps [11]. A previous attempt to biochemically reconstitute M. smegmatis translation has not involved M. smegmatis tRNAs and AARSs and thus cannot synthesize a full-length reporter protein for convenient in vitro assays [12]. Results and Discussion Purification of recombinant M. tuberculosis translation factors expressed in E. coli The overall workflow for the reconstitution of mycobacterial protein translation is illustrated in Fig 1. The genes for the M. tuberculosis translation factors (Tables 1 and 2) were de novo synthesized and codon-optimized for expression in E. coli, and the recombinant proteins were histidine-tagged for rapid purification on nickel-affinity chromatography. The initiation factors (IF1, IF2), elongation factor G (G, from the fusA1 gene), release factors (RF1, RF2) and ribosome-recycling factor (RRF) were over-expressed as highly soluble proteins and purified to near homogeneity after the nickel column as indicated by the SDS-PAGE analysis (S1A Fig). The over-expressed initiation factor 3 (IF3) was insoluble, found exclusively in inclusion bodies (IB), and was purified under the denaturation condition followed by in vitro renaturation (S1B Fig). The elongation factor Ts (Ts) was expressed as partially soluble protein and a large portion of the expressed protein remained in the pellet (S1C Fig). The soluble Ts was subsequently purified from the nickel column. The elongation factor Tu (Tu) when over-expressed alone was completely insoluble, and therefore was purified from inclusion bodies (IB) and renatured in vitro (S1D Fig). Alternatively, we co-expressed both Tu and Ts genes under a single promoter and purified the Tu/Ts complex to near homogeneity under the native condition after several chromatographic steps (S1E Fig). This Tu/Ts complex was used for the reconstitution of the complete mycobacterial translation system, while the separately purified Tu and Ts were used for the functional conservation experiments in which they were substituted for the E. coli Tu and Ts. 10.1371/journal.pone.0162020.g001Fig 1 A workflow for the biochemical reconstitution of mycobacterial protein translation. 10.1371/journal.pone.0162020.t001Table 1 Sequence homology of the components involved in protein translation in Mycobacterium and Escherichia coli. Functions Protein names Mycobacterium gene names E. coli gene name % identity Initiation Initiation factor 1 (IF1) infA infA 68 Initiation factor 2 (IF2) infB infB 33 Initiation factor 3 (IF3) infC infC 47 Elongation Elongation factor Tu (EF-Tu) tuf tufA 75 Elongation factor Ts (EF-Ts) tsf tsf 44 Elongation factor G (EF-G) fusA1 fusA 32 fusA2 - - Elongation factor P (P) efp efp 42 Termination and recycling Release factor 1 (RF1) prfA prfA 45 Release factor 2 (RF2) prfB prfB 43 Release factor 3 (RF3) - prfC - Ribosome-recycling factor (RRF) frr frr 41 Aminoacylation of tRNAs AlaRS alaS alaS 41 ArgRS argS argS 24 AsnRSa gatCAB asnS - AspRS aspS aspS 48 CysRS cysS cysS 43 GlnRSa gatCAB glnS - GluRS gltX gltX 36 GlyRS glyS glyQ, glyS - HisRS hisS hisS 43 IleRS ileS ileS 26 LeuRS leuS leuS 37 LysRS lysS lysS 39 MetRS metS metS 26 PheRS pheS, pheT pheS, pheT 47, 33 ProRS proS proS 42 SerRS serS serS 38 ThrRS thrS thrS 41 TrpRS trpS trpS 53 TyrRS tyrS tyrS 46 ValRS valS valS 44 tRNAs - 45 tRNAs 86 tRNAs - Ribosomes - - - - aIn Mycobacterium, AspRS charges both tRNAasp and tRNAasn to yield Asp-tRNAasp and Asp-tRNAasn, respectively. The mis-charged Asp-tRNAasn is immediately converted by M. smegmatis amidotransferase (a complex of three-gene product gatCAB) to Asn-tRNAasn. The similar process also occurs to make Glu-tRNAglu and Gln-tRNAgln. 10.1371/journal.pone.0162020.t002Table 2 A complete list of components in the mycobacterial protein synthesis system. Name Final concentration in in vitro translation reactions M. tuberculosis translation factors IF1 2.7 μM IF2 0.52 μM IF3 0.60 μM RF1 0.29 μM RF2 0.37 μM RRF 2.75 μM EF-Tua 3.0 μM EF-Tsa 3.0 μM EF-G 0.45 μM M. smegmatis AARS mix 0.96 mg/ml M. smegmatis tRNAs 3 mg/ml M. smegmatis ribosomes 2.4 μM E. coli energy regeneration enzymes MK 0.26 μM CK 0.22 μM NDK 0.07 μM PPA 0.04 μM E. coli methionyl tRNA formyltransferase (MTF) 0.57 μM T7 RNA polymerase 0.10 μM Small molecules and buffer ATP 2 mM GTP 2 mM CTP 1 mM UTP 1 mM Creatine phosphate 20 mM 20 amino acids 0.3 mM N10-formyl-tetrahydrofolate 0.02 mM Spermidine 2 mM DTT 7.2 mM Mg(OAc)2 10 mM K-Glutamate 100 mM HEPES-KOH pH7.5 50 mM a The purified EF-Tu/EF-Ts complex was used in the mycobacterial translation system. The complex contains an equal molar concentration of EF-Tu and EF-Ts based on the estimation from the SDS-PAGE analysis. Purification of ribosomes, tRNAs and aminoacyl-tRNA synthetases from M. smegmatis Since M. smegmatis is a non-infectious close relative of M. tuberculosis, we grow M. smegmatis cells as the source for ribosomes, tRNAs and aminoacyl-tRNA synthethases (AARSs) (Fig 1). Based on high sequence homologies between the translational components of these two closely related species, we expected M. smegmatis ribosomes and tRNAs to be functionally compatible with M. tuberculosis translation factors. M. smegmatis ribosomes were purified from the cell extract via ultracentrifugation and the ribosomal proteins were analyzed on a SDS-PAGE gel (S2A Fig). Compared to their E. coli counterparts, M. smegmatis ribosomal proteins exhibited somewhat different band patterns and notably lacked the S1 protein, suggesting that S1 was dissociated from the intact M. smegmatis ribosome during purification. The function of purified M. smegmatis ribosomes was examined by substituting E. coli ribosomes in a reconstituted E. coli system to synthesize a reporter. The reconstituted E. coli system did not generate any reporter activity in the absence of ribosomes, but produced ~36% of the reporter activity with M. smegmatis ribosomes as compared to E. coli ribosomes (S2B Fig). The data indicate that purified M. smegmatis ribosomes (in spite of the lack of S1) were active in in vitro protein synthesis. The cell extract of M. smegmatis cells was used for isolation of total tRNAs and enrichment of AARSs (Fig 1). Total tRNAs were purified by phenol extraction and isopropanol fractionation and shown to be comparable to the purified total E. coli tRNAs (S3A Fig). We expected all M. smegmatis AARSs to be soluble in the cell extract and they could be enriched by ion exchange chromatography. We collected and pooled the peak fractions from the elution of the DEAE-ion exchange column (S3B Fig), and used the combined fractions as the source for M. smegmatis AARSs. As suggested by the SDS-PAGE gel, this AARSs mixture likely contains most of the cytosolic proteins of M. smegmatis cells (S3B Fig). However, the AARS mixture does not support any significant in vitro synthesis of a reporter luciferase in the absence of translation factors (TF) or in the presence of translation factors (TF) and/or ribosomes without purified M. smeg tRNAs (S4 Fig, second to fifth columns, and S9 Fig). A significant luciferase activity was observed only in the presence of translation factors (TF), AARSs,purified tRNAs, and ribosomes at the same time (S4 Fig, first column). The final protein concentration of the AARS mixture in the mycobacterial translation system was 0.96 mg/ml (Table 2), which on its own apparently contained sufficient aminoacyl tRNA synthetases, but not other factors to support in vitro synthesis. This is consistent with early studies of in vitro protein synthesis with the E. coli cell extract which was normally used in excess of 10 mg/ml [13]. Reconstitution of mycobacterial protein translation from purified components for efficient in vitro protein synthesis For in vitro protein synthesis from a DNA template, we used T7 RNA polymerase to couple the transcription from a T7 promoter to in vitro translation. We reconstituted mycobacterial protein translation by mixing purified translation factors, ribosomes, tRNAs and the AARS mixture following similar protocols as described previously for the reconstituted E. coli and Thermus thermophilus translation systems [9, 14]. The final concentrations of the components in the mycobacterial translation system are listed in Table 2 and are largely similar to those of the reconstituted E. coli translation system with some modifications. For instance, we used a higher concentration of mycobacterial tRNAs (3 mg/ml) compared to the E. coli translation system (2 mg/ml). The mycobacterial AARSs was a mixture of native enzymes enriched from the M. smegmatis cell extract in contrast to the E. coli translation system in which AARSs consist of 20 individually purified recombinant proteins. We used the purified Tu/Ts complex instead of separate proteins to achieve the highest in vitro synthesis activity in the mycobacterial translation system. The reconstituted mycobacterial translation system was tested for its ability to synthesize a reporter from a DNA template. We found that the protein synthesis yield was ~19% of that of the reconstituted E. coli system (Fig 2, third column). The lower synthesis yield of the mycobacterial translation system could be due to the inherent slow translation by mycobacterial ribosomes, as M. smegmatis grows almost 10 times slower than E. coli. Another possibility was the non-optimal concentrations of M. smegmatis AARSs for more efficient in vitro translation. The M. smegmatis AARSs were purified as a mixture from the cell extract. This precludes the possibility of adjusting the concentration of each M. smegmatis AARS for maximal in vitro protein synthesis as it was the case for the AARSs in the E. coli translation system. We also cannot exclude the possibility that certain components that contribute to efficient protein translation were missing, less active or present at low concentrations in the mycobacterial translation system. To fully reconstitute mycobacterial protein translation, it will be necessary to individually purify all mycobacterial tRNA synthetases as recombinant proteins, including aspartyl/glutamyl tRNAAsn/Gln amidotransferases (products of gatCAB genes) involved in the generation of Asn-tRNAasn and Gln-tRNAgln (Table 1) [15, 16]. 10.1371/journal.pone.0162020.g002Fig 2 Comparison of protein synthesis yields of mycobacterial and E. coli translation systems and functional conservation of the mycobacterial and E. coli ribosomes. M. smegmatis: Mycobacterium smegmatis; M. tuberculosis: Mycobacterium tuberculosis. The data are shown as means from at least two independent reactions; error bars show s.d. Functional comparison of the translational components between Mycobacterium and E. coli The availability of both mycobacterial and E. coli translation systems allowed us to perform the functional comparison of major translational components between gram-positive and gram-negative bacteria. First, we compared M. smegmatis ribosomes with E. coli ribosomes in the reconstituted translation systems for in vitro synthesis of a reporter. When substituting for E. coli ribosomes, M. smegmatis ribosomes in the E. coli translation system resulted in 34% of the reporter activity compared to the complete E. coli translation system, suggesting that M. smegmatis ribosomes are generally compatible with the translation components of E. coli (Fig 2, compare the fourth column to the first column). These data are consistent with similar studies that used Bacillus and M. smegmatis ribosomes in a reconstituted E. coli translation system [6, 17]. In comparison, when substituting for M. smegmatis ribosomes, E. coli ribosomes in the mycobacterial translation system resulted in only 15% of the synthesis activity compared to that of the complete E. coli translation system (Fig 2, compare the second column to the first column). At first glance, it seemed that E. coli ribosomes were not very compatible with the mycobacterial translation components. However, compared to the complete mycobacterial translation system, E. coli ribosomes in the mycobacterial translation system resulted in ~76% activity (Fig 2, compare the second column to the third column), suggesting that E. coli ribosomes were actually compatible with mycobacterial translation components. Next we investigated the compatibility of the translation factors, AARSs and tRNAs between Mycobacterium and E. coli in the context of the mycobacterial translation (with M. smegmatis ribosomes). A comparison of the amino acid sequence homology between mycobacterial and E. coli translational components reveals varying degrees of the sequence identity (Table 1). Thus it was interesting to investigate how the sequence identity corresponds to the functional compatibility. First, we substituted mycobacterial translation factors (TFs) with their E. coli counterparts in the mycobacterial translation system and analyzed the effect of each substitution on the activity of the translated reporter (S5–S7 Figs). Use of the E. coli initiation factors (IFs) in the mycobacterial translation system decreased in vitro synthesis of the reporter by a factor of 50% compared to the mycobacterial initiation factors (S5 Fig). On the other hand, E. coli elongation factors (Tu and Ts) increased in vitro synthesis of the reporter compared to their mycobacterial counterparts (S6 Fig, compare first and second columns). However, Tu and Ts from mixed species reduced the amount of the synthesized reporter (S6 Fig, compare third and fourth columns), suggesting Tu from one species was not very compatible with Ts from another species. Similar hybrid experiments were performed with the ribosome release factor (RRF) and elongation factor G (EF-G). However, in this case, no significant differences were observed whether these factors were from the same species or mixed species (S7 Fig). Our findings in general suggest that E. coli translation factors can substitute for their mycobacterial counterparts in the mycobacterial system without significantly affecting in vitro synthesis of the reporter. This functional conservation of translation factors between Mycobacterium and E. coli is consistent with a previous study that used a similar in vitro biochemical approach [12]. Though major translational components from Mycobacterium and E. coli were purified under similar conditions and used in the same concentrations in these comparative studies, we cannot exclude the possibility that one or a few mycobacterial components were less active and led to the differences we observed in the results described above. The most striking differences were observed when E. coli AARSs and/or tRNAs replaced their mycobacterial counterparts in the mycobacterial translation system (Fig 3). M. smegmatis AARSs seemed to be able to charge E. coli tRNAs at a similar efficiency as M. smegmatis tRNAs, as judged by the similar reporter activities (albeit ~30% more for E. coli tRNAs) (Fig 3, first and second columns). In contrast, E. coli AARSs resulted in almost a complete loss in the reporter activity with M. smegmatis tRNAs (Fig 3, third column). Such drastic decrease in the reporter activity did not seem to be caused by the use of E. coli translation factors (TFs) (Fig 3, third column) since E. coli TFs were generally compatible with the rest of the mycobacterial translation components (S5–S7 Figs). These data suggest that E. coli AARSs appeared to be incapable of properly recognizing M. smegmatis tRNAs, possibly resulting in uncharged or mis-charged tRNAs and the loss of the reporter activity. This incompatibility could be due to the sequence (e.g., N73) and modification differences between M. smegmatis and E. coli tRNAs in general, and non-discriminating M. smegmatis aminoacyl tRNA synthetases and their tRNAs in particular. For instance, M. smegmatis tRNAasp and tRNAasn were both charged by the same M. smegmatis non-discriminating aspartyl-tRNA sythetases (ND-AspRS) to yield Asp-tRNAasp and Asp-tRNAasn, respectively. The mis-charged Asp-tRNAasn is immediately converted by M. smegmatis amidotransferase to Asn-tRNAasn. The similar process also occurs in M. smegmatis to make Glu-tRNAglu and Gln-tRNAgln. In contrast, E. coli AARSs consist of 20 individually purified synthetases, none of which is a non-discriminating synthetase. E. coli tRNAasn and tRNAgln are charged by the specific (discriminating) aspariginyl-tRNA and glutaminyl-tRNA synthetases, respectively. The lack of amidotransferases in E. coli and the possibility of E. coli AARSs not being able to properly charge M. smegmatis tRNAs due to the sequence differences may account for the above observation (Fig 3, column 3). 10.1371/journal.pone.0162020.g003Fig 3 Functional comparison of tRNA aminoacylation between Mycobacterium and E. coli translation systems. Note that M. smegmatis aminoacyl-tRNA systhetases (AARSs) are a mixture of proteins from the soluble fractions of M. smegmatis cell extract, whereas E. coli AARSs consist of 20 individually purified recombinant enzymes. The data are shown as means from at least two independent reactions; error bars show s.d. Though a number of M. smegmatis AARSs have very low amino acid sequence identities with those of E. coli AARSs (Table 1, e.g., Ile- and Met-tRNA synthetases at 26%), as far as the synthesis of the active reporter is concern, M. smegmatis AARSs seemed to be able to properly charge E. coli tRNAs (Fig 3, second column). The reason for this observation was not clear. M. smegmatis AARSs in these experiments lacked Asn- and Gln-tRNA synthetases, and contained essentially most soluble proteins from the M. smegmatis extract. M. smegmatis relies on a two-step mechanism and amidotransferases to generate M. smegmatis Asn-tRNAAsn and Gln-tRNAGln. Whether this process occurred with E. coli tRNAs in the mycobacterial translation system remains to be tested. Future experiments will individually purify recombinant M. smegmatis AARSs including amidotransferases, which would allow us to address the above questions and at the same time fully reconstitute the mycobacterial protein translation. Use of both mycobacterial and E. coli translation systems as a comparative in vitro assay platform for testing antibiotics that specifically target M. tuberculosis protein translation Structural studies of antibiotics-bound ribosomes have provides insights on the mechanisms of translation inhibition by antibiotics. However, these studies are often limited to the model organisms such as E. coli and Thermus thermophilus [18, 19]. To date, there is no high-resolution crystal structure of the ribosome from any pathogenic bacterium, hindering the rational design of antibiotics that specifically target the pathogenic bacterium. The structural differences of the mycobacterial ribosome from the E. coli ribosome revealed by the cryo EM studies [5] suggest a possibility of screening small molecules that preferentially target the mycobacterial translation. Since it consists of only ribosomes and other translational components, a totally reconstituted bacterial protein synthesis system could be an ideal platform to assay and screen small-molecules that target the translation process. With the availability of both mycobacterial and E. coli translation systems, small molecules that target mycobacterial translation more specifically than that of E. coli or vice versa could be readily identified. To demonstrate such feasibility, we tested two known antibiotics, spectinomycin and tetracycline. We examined the inhibitory effects of the antibiotics by measuring the activity of the reporter synthesized in either mycobacterial or E. coli translation system in the presence of different concentrations of the antibiotics. We found that spectinomycin inhibited both mycobacterial and E. coli translation with a similar in vitro efficacy (IC50 at ~1 μg/ml) (Fig 4). In contrast, tetracycline preferentially inhibited mycobacterial translation with the estimated IC50 at ~2 μg/ml, while the IC50 for E. coli translation was ~40 μg/ml (Fig 4). Based on co-crystal structures with the ribosomes from E. coli or Thermus thermophilus, both antibiotics inhibit protein translation by binding to the 30S ribosomal subunit [20–22]. Spectinomycin interacts with the head domain of the 30S subunit and block translocation of mRNA and tRNAs on the ribosome [21]. On the other hand, tetracycline interferes with aminoacyl-tRNA entrance by binding to the A site of the ribosome [22]. Our data point to a possibility that there may be structural differences at the ribosomal A site between M. smegmatis and E. coli that allow tetracycline to inhibit the M. smegmatis ribosomes at much lower concentrations. However, we cannot rule out the possibility of other mechanisms of action. For instance, E. coli ribosomes may contain some species-specific ribosome-associated proteins that interfere with the tetracycline binding. Tetracycline may preferentially interact with other translational components in the mycobacterial translation system and inhibit protein translation through yet-to-identified mechanisms. We additionally tested the effects of chloramphenicol, which prevents translation elongation by inhibiting peptidyl transferase activity of ribosomes (S8 Fig). Both mycobacterial and E. coli translation were inhibited by chloramphenicol with relatively similar in vitro efficacies (IC50 at ~5 and 1 μg/ml, respectively) (S8 Fig). 10.1371/journal.pone.0162020.g004Fig 4 Use of the mycobacterial and E. coli translation systems as comparative in vitro inhibition assays for antibiotics Spectinomycin (left panel) and tetracycline (right panel). The activities of the luciferase reporter synthesized in either mycobacterial (black circles) or E. coli (white circles) are determined in the presence of various concentrations of the antibiotics. The data are shown as means from at least two independent reactions; error bars show s.d. We have previously demonstrated the use of the E. coli translation system for screening small molecule inhibitors in 1536-well plates [10], mycobacterial translation system in this study should be amenable to microplate-based high throughput assays. The total number of assays will be determined by the amount of mycobacterial translation system one can produce at reasonable costs. Most of the components of mycobacterial translation systems can be produced in fairly large amounts with small-scale cell cultures. The most limiting component is M. smegmatis tRNAs, which we purified ~31 mg from 25 g cells, corresponding to 400 reactions (25 μl per reaction). Therefore, mycobacterial translation system at the current scale may be suitable for screening small molecules from targeted libraries. Conclusions This work illustrates a bottom-up biochemical approach for functional studies of the protein translation machinery of mycobacteria and possible other bacterial pathogens. We have established the feasibility that protein translation systems derived from different bacteria may be used to identify species-specific antibiotics. Our goal has been to construct a translation system consisting of the translational components entirely from an infectious M. tuberculosis strain. However, the current mycobacterial translation system is still a hybrid system in which AARS, tRNAs and ribosomes are purified from Mycobacterium smegmatis, an easy-to-grow close relative of M. tuberculosis. To achieve a complete biochemical reconstitution of protein translation as well as transcription regulation of an infectious M. tuberculosis strain, the future works would involve purification of recombinant M. tuberculosis AARSs, recombinant amidotransferases, RNA polymerase holoenzyme and transcription factors, and isolation of ribosomes and tRNAs from M. tuberculosis cells. Such a complete M. tuberculosis cell-free system may be used as an "in vitro genetic tool" [8, 10] for functional studies of M. tuberculosis-specific genes and pathways and for screening M. tuberculosis-specific antibiotics. Materials and Methods Reagents Unless specified otherwise, the chemicals were purchased from Sigma-Aldrich (St. Louis, MO), the reagents including the reconstituted E. coli translation system (derived from PURExpress™) were from New England Biolabs (NEB) (Ipswich, MA), and the primers were ordered from Integrated DNA Technologies (Coralville, IA). Cloning, expression and purification of recombinant M. tuberculosis translation factors from E. coli The genes encoding IF1, IF2, IF3, RF1, RF2, RRF, EF-Tu, EF-Ts, EF-G were synthesized by GeneScript (Piscataway, NJ) with codons optimized for expression in E. coli. All genes were cloned into the expression vector, pCOATexp, derived from pTYB1 vector (New England Biolabs), to allow expression of recombinant proteins with a C-terminal histidine-tag. The expression vectors were transformed into E. coli strains ER3095 (New England Biolabs). For each histidine-tagged protein, cells transformed with the expression vector were grown at 37°C to OD600 of 0.6 in 2–6 L Luria-Bertani (LB) broth. Isopropyl-β-D-thiogalactoside (IPTG) was then added to a final concentration of 0.1 mM and the cells were grown for an additional 4–5 hr at 37°C. Cells were harvested by centrifugation and lysed by sonication in lysis buffer (50 mM Tris-HCl, pH7.5, 300 mM KCl, 10 mM MgCl2, 20 mM imidazole, and 1 mM β-mercaptoethanol). Cell debris was removed by centrifugation at 16,000 g for 1 hr at 4°C and the supernatant was applied to a 10 ml HisTrap FF column (GE healthcare, Piscataway, NJ). After washing the column with 100 ml lysis buffer, the his-tagged protein was eluted with a linear gradient of 20 mM to 250 mM imidazole in lysis buffer. Fractions containing the his-tagged protein were combined and dialyzed against the storage buffer (25 mM Tris-HCl, pH7.5, 100 mM K-glutamate, 10 mM Mg(OAc)2, 30% glycerol, and 1 mM β-mercaptoethanol) and stored frozen in small aliquots at -80°C. In the case of IF3 and EF-Tu, both recombinant proteins were insoluble and mostly found in the fractions of the inclusion bodies (IB) (S1B and S1D Fig). To solubilize the inclusion bodies, cell pellet after centrifugation was washed twice with a washing buffer (20 mM Tris, pH7.5, 10 mM EDTA and 1% Triton X-100) and then dissolved in a denaturation buffer (20 mM Tris, pH8.5, 4M urea). After centrifugation at 16,000 g for 30 min at 4°C to remove undissolved materials, the supernatant was applied to a 10 ml HisTrap FF column (GE healthcare, Piscataway, NJ) equilibrated with the renaturation buffer 1 (20 mM Tris, pH8.5, 2M urea). After washing the column with the renaturation buffer 1, IF3 or EF-Tu was eluted with a linear gradient of 10 mM to 200 mM imidazole in the renaturation buffer 1. Fractions containing the IF3 or EF-Tu protein were combined and dialyzed against the renaturation buffer 2 (20 mM Tris, pH8.5, 1M urea) and then the renaturation buffer 3 (20 mM Tris, pH8.5). The solubilized proteins were stored frozen in small aliquots at -80°C M. tuberculosis EF-Tu was also purified as a complex with M. tuberculosis EF-Ts. In this case, the genes for M. tuberculosis EF-Ts and EF-Tu were cloned into the MCS1 and MCS2 of pETDuet-1 respectively. A peptide MGSSHHHHHHSQDPNS was engineered to the N-terminus of EF-Ts to add a 6xHis-tag and a linker region to facilitate the affinity purification of the EF-Tu/Ts complex. The plasmid was transformed into Rosetta DE3 competent cells. The cells were grown at 37°C to OD600 = 1.0 and then induced by adding IPTG to a final concentration of 0.5mM. The induction was continued at 25°C overnight with shaking. Cells were harvested by centrifugation at 3,000 rpm for 20 mins. The cell pellet was re-suspended in the Lysis buffer containing 50 mM of Tris pH 8.0, 150 mM of NaCl and 10mM of Imidazole. Cells were disrupted by ultrasonication on ice, followed by centrifugation at 15,000rpm for 30mins at 4°C to remove the cell debris. The supernatant was loaded to a 3ml Ni-NTA(Qiagen) pre-equilibrated with the Lysis buffer. The resin was washed with 20 column volumes Lysis buffer and then eluted with the elution buffer containing 50 mM Tris 8.0, 150 mM NaCl and 100 mM Imidazol. The eluate was immediately dialyzed against a buffer containing 50mM of Tris pH 8.0 and 150mM of NaCl. The sample was loaded to a 5-ml anion exchange column Hitrap Q HP (GE healthcare) and eluted with a linear gradient of 150 mM to 1M NaCl. Two peaks from the ion exchange chromatography were observed, both of which contain protein species corresponding to the molecule weights of EF-Tu and EF-Ts. Mass spectrometry analysis was performed to identify that the first peak contained the complex of of E. coli EF-Tu and M. tuberculosis EF-Ts, whereas the second peak contained the complex of M. tuberculosis EF-Tu and M. tuberculosis EF-Ts. Therefore, the second peak was collected and subjected to a size-exclusion column Superose 6 10/300 GL (GE healthcare). The eluate containing the M. tuberculosis EF-Tu/Ts complex was pooled and concentrated to ~30 mg/ml. The protein concentrations were determined by Bradford Assay. For SDS-PAGE analyses, aliquots were taken from the in vitro translation reactions and run on a 10–20% Tris–glycine SDS–PAGE gels (Life Technologies, Carlsbad, CA). Purification of M. smegmatis ribosomes and testing the protein synthesis activity M. smegmatis ribosomes were purified using established protocols with minor modification [9, 23]. All buffers and purification procedures were at 4°C, unless otherwise noted. Specifically, M. smegmatis(ATCC 19420) cells were grown in small-scale fermentation (10 L) in NEB fermentation facility. The culture was cooled rapidly and the cells were harvested by centrifugation. Fresh cell paste (80 g) was washed twice in 200 ml of wash/lysis buffer (20 mM Tris-HCl, pH7.5, 100 mM NH4Cl, 10.5 mM MgCl2, 0.5 mM EDTA, 6 mM β-mercaptoethanol) and then resuspended in 150 ml of the same wash/lysis buffer. The cells were disrupted at 40 kpsi by a cell disruption system (Constant Systems, Low March, Daventry, Northants, United Kingdom) and cell debris was removed by centrifugation at 30,000 g for 1 hr. The supernatant was centrifuged again at 30,000 g for additional 30 min and saved as the M. smegmatis S30 extract. For ribosome purification, 25 ml of the S30 extract was overlayered onto 13 ml of a sucrose cushion (Cushion I: 20 mM Tris-HCl, pH7.5, 10.5 mM MgCl2, 0.5 mM EDTA, 1.1 M Sucrose, 6 mM β-mercaptoethanol, 2mM DTT) in each ultracentrifuge tube, and the ultracentrifugation was conducted in a Beckman SW28 rotor at 28,000 rpm for 22 hr. The pellet containing the ribosome was solubilized in the ribosome storage buffer (25 mM Tris-OAc, pH7.5, 100 mM NH4OAc, 10 mM Mg(OAc)2, 6 mM β-mercaptoethanol, 2mM DTT) by overnight incubation at 4°C. The purified ribosome was concentrated to an appropriate concentration and stored frozen in aliquots at -80°C. The final concentration of the purified ribosome was measured at A260. Aliquots of the purified M. smegmatis and E. coli ribosomes were loaded on a SDS-PAGE gel for comparison and purity check (S2A Fig). The activity of the purified M. smegmatis ribosomes was tested for in vitro synthesis of a reporter in a reconstituted E. coli system in which E. coli ribosomes were removed. The protein synthesis reactions (25μl) were set up by mixing the reconstituted protein synthesis system containing 2.4μM either E. coli or M. smegmatis ribosomes with RNase inhibitor (2 units) and 200 ng reporter DNA templates (pUCAT7Fluc) expressing the firefly luciferase (Fluc) under a T7 promoter. The reaction mixtures were incubated at 37°C for 4 hr and aliquots were taken for luciferase assays. Purification of total M. smegmatis tRNAs and enrichment of M. smegmatis AARSs from the M. smegmatis cell extract Purification of total M. smegmatis tRNAs. M. smegmatis cells were lysed as described above in the ribosome purification protocol. The lysate was clarified by low speed centrifugation (10000 rpm for 20 minutes) in a Sorvall centrifuge. Clarified lysate was divided in equal halves and one half was used for total tRNAs preparation whereas the another half was used for AARSs enrichment (see below). Briefly, The clarified cell lysate was passed through a DEAE column equilibrated with 50 mM Tris-HCl, pH7.5, 0.5 mM EDTA, 6 mM β-mercaptoethanol, 50mM NaCl. Fractions were eluted with a linear gradient of 100 mM to 1500mM NaCl in the aforesaid equilibrium buffer. All the fractions exhibiting UV absorbance (representing either soluble proteins, nucleic acids or a mix of both) were pooled and diluted with the aforesaid equilibrium buffer and mixed with an equal volume of the saturated phenol solution (pH 6.6) and incubated at 37°C for 30 min. After centrifugation at 13,000 g for 15 min, the upper layer was collected and subject to a second phenol extraction and centrifugation step. The nucleic acids (DNA and RNA) in the upper layer were then precipitated by adding 0.1 volume of 3M sodium acetate (pH5.3) and 2.5 volume of cold ethanol, and incubating overnight at -20°C. Following centrifugation at 13,000 g for 30 min, the pellet was dried and then dissolved in 70 ml of 0.3 M sodium acetate (pH7.0). To precipitate DNA and high molecular weight RNA, isopropanol (38 ml) was added slowly in drops with a needle at 4°C. After centrifugation at 13,000 g for 30 min, the supernatant (~108 ml) was collected and mixed with 31 ml of isopropanol. Following centrifugation at 13,000 g for 30 min, pellet of the precipitated tRNAs was dried and dissolved in H2O. The final concentration of total tRNA was measured at A260. Enrichment of M. smegmatis AARSs from the cell lysate. We applied the clarified lysate on a DEAE-cellulose column equilibrated with 50 mM Tris-HCl, pH7.5, 0.5 mM EDTA, 6 mM β-mercaptoethanol followed by elution against a linear gradient of 0.0–0.5M NaCl in the same buffer. The protein fractions were pooled and precipitated with 100% ammonium sulfate. The precipitate was recovered by low speed centrifugation (15,000g for 30 minutes), dissolved in a buffer containing 50 mM Tris-HCl, pH7.5, 0.5 mM EDTA, 6 mM β-mercaptoethanol, 20 mM NaCL and dialyzed twice against the same buffer. The resulting preparation enriched in mycobacterial AARSs was used for the reconstitution of mycobacterial protein translation. Reconstitution of mycobacterial protein translation from purified components Following a protocol similar to that of the reconstituted E. coli translation system, the mycobacterial translation was reconstituted by mixing the purified M. tuberculosis translation factors, M. smegmatis AARSs mixture, M. smegmatis tRNAs and ribosomes, E. coli energy regeneration enzymes, T7 RNA polymerase (for coupling transcription to translation), and small molecules (see Table 2 for the complete list and concentrations). The protein synthesis reactions (typically 25μl each reaction) were set up by mixing the reconstituted protein synthesis system with RNase inhibitor (2 units) and 200 ng reporter DNA templates (pUCAT7Fluc) expressing the firefly luciferase (Fluc) under a T7 promoter. The reaction mixtures were incubated at 37°C for 4 hr and aliquots were taken for the luciferase assays. Functional comparison of the translational components between Mycobacterium and E. coli The experiments swapping the mycobacterial and E. coli translation components (translation factors (TFs, individually or all together), AARSs, tRNAs, ribosomes) were primarily conducted in in vitro translation reactions that synthesized the Fluc reporter (Figs 2 and 3 and S2–S7 Figs). The mycobacterial translation components were either removed or replaced by equal amounts of corresponding E. coli counterparts. In vitro antibiotic inhibition assays in mycobacterial and E. coli translation systems The assays (25μl each assay) were set up by mixing the mycobacterial or E. coli translation system with RNase inhibitor (2 units), 200 ng reporter DNA templates (pUCAT7Fluc) expressing the firefly luciferase (Fluc) and various amounts of spectinomycin or tetracycline. The reaction mixtures were incubated at 37°C for 4 hr and aliquots were taken for the luciferase assays. Luciferase activity assay The activity of the Fluc reporter was assayed using the Luciferase Assay System (Promega, Madison, WI) in a microplate luminometer (Centro LB 640, Berthold Technologies, Oak Ridge, TN) according to manufacturers’ instructions. Protein synthesis reactions were diluted 10-fold in 1x cell culture lysis reagent (Promega, Madison, WI) containing 1 mg/ml BSA. Aliquots (5 μl) were added in triplicate to a microplate for the luciferase assay in the luminometer. Supporting Information S1 Fig SDS-PAGE analysis of the purification of M. tuberculosis translation factors. (A) These soluble M. tuberculosis translation factors were purified under native conditions to near homogeneity after the nickel column; (B) IF3 were insoluble, mostly present in inclusion bodies (IB) and purified under denaturation conditions; M: molecular weight; (C) EF-Ts (Ts) was partially soluble and present in both soluble fraction (S) and inclusion bodies (IB). Ts was purified under native conditions without denaturation. L: lysate; FT: flow through fraction; (D) EF-Tu (Tu) when expressed alone was mostly insoluble and therefore purified under denaturation conditions; The elution fractions from the nickel column are shown here. (E) When co-expressed, EF-Tu and EF-Ts are soluble and form a 1:1 complex, which was purified over several columns. The final elution fractions are shown here. (PPTX) Click here for additional data file. S2 Fig Purification of M.smegmatis ribosomes. (A) SDS-PAGE analysis of M.smegmatis and E. coli ribosomes. (B) Reporter activity assays of M. smegmatis ribosomes in the E. coli translation system in which E. coli ribosomes were replaced by M. smegmatis ribosomes. (PPTX) Click here for additional data file. S3 Fig Purification of total tRNAs and AARSs. (A) Agarose gel analysis of purified M.smegmatis total tRNAs (lane 1) in comparison with purified E. coli total tRNAs (lane 2). (B) DEAE-column elution profile of the cell extract of M.smegmatis and SDS-PAGE analysis of the elution peak fractions, which we expected to contain all M.smegmatis aminoacylation enzymes (AARSs). (PPTX) Click here for additional data file. S4 Fig Reporter activity assays of M.smegmatis AARS mixture and purified M.smegmatis tRNAs in the mycobacterial translation reactions. M. tuberculosis translation factors (TFs) and M.smegmatis ribosomes were used. The luciferase activity below 1000 is considered to be the background. (PPTX) Click here for additional data file. S5 Fig Comparison of E. coli and M.smegmatis initiation factors (IFs) in the mycobacterial translation system for the synthesis of a luciferase reporter. (PPTX) Click here for additional data file. S6 Fig Comparison of E. coli and M.smegmatis elongation factor Tu (Tu) and elongation factor Ts (Ts) in the mycobacterial translation system for the synthesis of a luciferase reporter. (PPTX) Click here for additional data file. S7 Fig Comparison of E. coli and M.smegmatis ribosome-release factor (RRF) and elongation factor G (EF-G) in the mycobacterial translation system for the synthesis of a luciferase reporter. (PPTX) Click here for additional data file. S8 Fig Effect of chloramphenicol on the reporter synthesis in E. coli (open circle) and M. smegmatis (black circle) translation systems. The activities of the synthesized reporter were determined in the presence of various concentrations of the antibiotic. The data are shown as means from two independent reactions; error bars show s.d. (PPTX) Click here for additional data file. S9 Fig M. smegmatis AARS mixture alone or with ribosomes was not sufficient to synthesize a significant amount of the reporter protein in M. smegmatis translation system. (PPTX) Click here for additional data file. We thank Drs. Chris Taron and Bill Jack for suggestions and comments on the manuscript. Drs. Richard Roberts and Donald Comb (New England Biolabs) and Prof. Boqin Chiang (Chinese Academy of Medical Sciences) for encouragement. ==== Refs References 1 World Health Organization. Global tuberculosis report. 2014. 2 Keshavjee S , Farmer PE . Tuberculosis, drug resistance, and the history of modern medicine . N Engl J Med . 2012 ;367 (10 ):931 –6 . 10.1056/NEJMra1205429 .22931261 3 Pai M , Schito M . Tuberculosis diagnostics in 2015: landscape, priorities, needs, and prospects . J Infect Dis . 2015 ;211 Suppl 2 :S21 –8 . 10.1093/infdis/jiu803 25765103 4 Zhang Y , Post-Martens K , Denkin S . New drug candidates and therapeutic targets for tuberculosis therapy . Drug Discov Today . 2006 ;11 (1–2 ):21 –7 . S1359-6446(05)03626-3 [pii] 10.1016/S1359-6446(05)03626-3 .16478687 5 Shasmal M , Sengupta J . Structural diversity in bacterial ribosomes: mycobacterial 70S ribosome structure reveals novel features . PloS one . 2012 ;7 (2 ):e31742 10.1371/journal.pone.0031742 22384065 6 Shi W , Zhang X , Jiang X , Yuan H , Lee JS , Barry CE 3rd, et al Pyrazinamide inhibits trans-translation in Mycobacterium tuberculosis . Science . 2011 ;333 (6049 ):1630 –2 . 10.1126/science.1208813 21835980 7 Javid B , Sorrentino F , Toosky M , Zheng W , Pinkham JT , Jain N , et al Mycobacterial mistranslation is necessary and sufficient for rifampicin phenotypic resistance . Proceedings of the National Academy of Sciences of the United States of America . 2014 ;111 (3 ):1132 –7 . 10.1073/pnas.1317580111 24395793 8 Asahara H , Chong S . In vitro genetic reconstruction of bacterial transcription initiation by coupled synthesis and detection of RNA polymerase holoenzyme . Nucleic acids research . 2010 ;38 (13 ):e141 Epub 2010/05/12. 10.1093/nar/gkq377 20457746 9 Zhou Y , Asahara H , Gaucher EA , Chong S . Reconstitution of translation from Thermus thermophilus reveals a minimal set of components sufficient for protein synthesis at high temperatures and functional conservation of modern and ancient translation components . Nucleic acids research . 2012 ;40 (16 ):7932 –45 . 10.1093/nar/gks568 22723376 10 Zhou Y , Asahara H , Schneider N , Dranchak P , Inglese J , Chong S . Engineering bacterial transcription regulation to create a synthetic in vitro two-hybrid system for protein interaction assays . Journal of the American Chemical Society . 2014 ;136 (40 ):14031 –8 . 10.1021/ja502512g 25188838 11 Ramon-Garcia S , Martin C , De Rossi E , Ainsa JA . Contribution of the Rv2333c efflux pump (the Stp protein) from Mycobacterium tuberculosis to intrinsic antibiotic resistance in Mycobacterium bovis BCG . J Antimicrob Chemother . 2007 ;59 (3 ):544 –7 . 10.1093/jac/dkl510 .17242035 12 Bruell CM , Eichholz C , Kubarenko A , Post V , Katunin VI , Hobbie SN , et al Conservation of bacterial protein synthesis machinery: initiation and elongation in Mycobacterium smegmatis . Biochemistry . 2008 ;47 (34 ):8828 –39 . 10.1021/bi800527k .18672904 13 Pratt JM . Coupled transcription-translation in prokaryotic cell-free systems In Transcription and Translation: A Practical Approach . In: Hames BD , Higgins SJ , editors. Oxford, United Kingdom : IRL Press ; 1984 p. 179 –209 . 14 Shimizu Y , Inoue A , Tomari Y , Suzuki T , Yokogawa T , Nishikawa K , et al Cell-free translation reconstituted with purified components . Nature biotechnology . 2001 ;19 (8 ):751 –5 . Epub 2001/08/02. 10.1038/90802 90802 [pii]. .11479568 15 Garnier T , Eiglmeier K , Camus JC , Medina N , Mansoor H , Pryor M , et al The complete genome sequence of Mycobacterium bovis . Proceedings of the National Academy of Sciences of the United States of America . 2003 ;100 (13 ):7877 –82 . 10.1073/pnas.1130426100 12788972 16 Sheppard K , Soll D . On the evolution of the tRNA-dependent amidotransferases, GatCAB and GatDE . Journal of molecular biology . 2008 ;377 (3 ):831 –44 . 10.1016/j.jmb.2008.01.016 18279892 17 Chiba S , Kanamori T , Ueda T , Akiyama Y , Pogliano K , Ito K . Recruitment of a species-specific translational arrest module to monitor different cellular processes . Proceedings of the National Academy of Sciences of the United States of America . 2011 ;108 (15 ):6073 –8 . 10.1073/pnas.1018343108 21383133 18 Bulkley D , Innis CA , Blaha G , Steitz TA . Revisiting the structures of several antibiotics bound to the bacterial ribosome . Proceedings of the National Academy of Sciences of the United States of America . 2010 ;107 (40 ):17158 –63 . 10.1073/pnas.1008685107 20876130 19 Dunkle JA , Xiong L , Mankin AS , Cate JH . Structures of the Escherichia coli ribosome with antibiotics bound near the peptidyl transferase center explain spectra of drug action . Proceedings of the National Academy of Sciences of the United States of America . 2010 ;107 (40 ):17152 –7 . 10.1073/pnas.1007988107 20876128 20 Carter AP , Clemons WM Jr., Brodersen DE , Morgan-Warren RJ , Hartsch T , Wimberly BT , et al Crystal structure of an initiation factor bound to the 30S ribosomal subunit . Science . 2001 ;291 (5503 ):498 –501 . .11228145 21 Borovinskaya MA , Shoji S , Holton JM , Fredrick K , Cate JH . A steric block in translation caused by the antibiotic spectinomycin . ACS Chem Biol . 2007 ;2 (8 ):545 –52 . 10.1021/cb700100n 17696316 22 Pioletti M , Schlunzen F , Harms J , Zarivach R , Gluhmann M , Avila H , et al Crystal structures of complexes of the small ribosomal subunit with tetracycline, edeine and IF3 . The EMBO journal . 2001 ;20 (8 ):1829 –39 . 10.1093/emboj/20.8.1829 11296217 23 Ohashi H , Shimizu Y , Ying BW , Ueda T . Efficient protein selection based on ribosome display system with purified components . Biochem Biophys Res Commun . 2007 ;352 (1 ):270 –6 . Epub 2006/11/23. S0006-291X(06)02486-7 [pii] 10.1016/j.bbrc.2006.11.017 .17113037

PMC005xxxxxx/PMC5001722.txt

==== Front PLoS PathogPLoS PathogplosplospathPLoS Pathogens1553-73661553-7374Public Library of Science San Francisco, CA USA 2756486510.1371/journal.ppat.1005850PPATHOGENS-D-16-00577Research ArticleResearch and Analysis MethodsPrecipitation TechniquesImmunoprecipitationBiology and Life SciencesPhysiologyImmune PhysiologyAntibodiesMedicine and Health SciencesPhysiologyImmune PhysiologyAntibodiesBiology and Life SciencesImmunologyImmune System ProteinsAntibodiesMedicine and Health SciencesImmunologyImmune System ProteinsAntibodiesBiology and Life SciencesBiochemistryProteinsImmune System ProteinsAntibodiesBiology and Life SciencesMolecular BiologyMolecular Biology TechniquesMolecular Probe TechniquesImmunoblottingResearch and Analysis MethodsMolecular Biology TechniquesMolecular Probe TechniquesImmunoblottingBiology and Life SciencesMicrobiologyVirologyViral StructureVirionsResearch and analysis methodsBiological culturesCell lines293T cellsBiology and life sciencesOrganismsVirusesDNA virusesHerpesvirusesHuman CytomegalovirusBiology and Life SciencesMicrobiologyMedical MicrobiologyMicrobial PathogensViral PathogensHerpesvirusesHuman CytomegalovirusMedicine and Health SciencesPathology and Laboratory MedicinePathogensMicrobial PathogensViral PathogensHerpesvirusesHuman CytomegalovirusBiology and Life SciencesOrganismsVirusesViral PathogensHerpesvirusesHuman CytomegalovirusBiology and Life SciencesMolecular BiologyMolecular Biology TechniquesMolecular Biology Assays and Analysis TechniquesGene Expression and Vector TechniquesProtein ExpressionResearch and Analysis MethodsMolecular Biology TechniquesMolecular Biology Assays and Analysis TechniquesGene Expression and Vector TechniquesProtein ExpressionMedicine and Health SciencesInfectious DiseasesViral DiseasesCytomegalovirus InfectionConsecutive Inhibition of ISG15 Expression and ISGylation by Cytomegalovirus Regulators HCMV Regulation of ISG15 Expression and ISGylationKim Ye Ji 1 Kim Eui Tae 1 Kim Young-Eui 1 Lee Myoung Kyu 1 Kwon Ki Mun 1 Kim Keun Il 2 Stamminger Thomas 3 Ahn Jin-Hyun 1 * 1 Department of Molecular Cell Biology, Samsung Biomedical Research Institute, Sungkyunkwan University School of Medicine, Suwon, Republic of Korea 2 Department of Biological Sciences, Sookmyung Women's University, Seoul, Republic of Korea 3 Institute for Clinical and Molecular Virology, University of Erlangen-Nuremberg, Schlossgarten, Erlangen, Germany Lenschow Deborah J. Editor Washington University School of Medicine, UNITED STATES The authors have declared that no competing interests exist. Conceptualization: YJK ETK JHA. Formal analysis: YJK ETK YEK MKL KMK. Investigation: YJK ETK YEK MKL KMK. Methodology: YJK ETK JHA. Resources: KIK TS. Writing - original draft: YJK ETK JHA. * E-mail: jahn@skku.edu26 8 2016 8 2016 12 8 e100585011 3 2016 8 8 2016 © 2016 Kim et al2016Kim et alThis is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Interferon-stimulated gene 15 (ISG15) encodes an ubiquitin-like protein that covalently conjugates protein. Protein modification by ISG15 (ISGylation) is known to inhibit the replication of many viruses. However, studies on the viral targets and viral strategies to regulate ISGylation-mediated antiviral responses are limited. In this study, we show that human cytomegalovirus (HCMV) replication is inhibited by ISGylation, but the virus has evolved multiple countermeasures. HCMV-induced ISG15 expression was mitigated by IE1, a viral inhibitor of interferon signaling, however, ISGylation was still strongly upregulated during virus infection. RNA interference of UBE1L (E1), UbcH8 (E2), Herc5 (E3), and UBP43 (ISG15 protease) revealed that ISGylation inhibits HCMV growth by downregulating viral gene expression and virion release in a manner that is more prominent at low multiplicity of infection. A viral regulator pUL26 was found to interact with ISG15, UBE1L, and Herc5, and be ISGylated. ISGylation of pUL26 regulated its stability and inhibited its activities to suppress NF-κB signaling and complement the growth of UL26-null mutant virus. Moreover, pUL26 reciprocally suppressed virus-induced ISGylation independent of its own ISGylation. Consistently, ISGylation was more pronounced in infections with the UL26-deleted mutant virus, whose growth was more sensitive to IFNβ treatment than that of the wild-type virus. Therefore, pUL26 is a viral ISG15 target that also counteracts ISGylation. Our results demonstrate that ISGylation inhibits HCMV growth at multiple steps and that HCMV has evolved countermeasures to suppress ISG15 transcription and protein ISGylation, highlighting the importance of the interplay between virus and ISGylation in productive viral infection. Author Summary Type I IFN response is a front-line defense against virus infection. Activation of type I IFN signaling leads to expression of a subset of cellular proteins encoded by interferon-stimulated genes (ISGs). ISG15 encodes an ubiquitin-like protein that is covalently conjugated to protein lysine residues. ISG15 modification (ISGylation) of a protein causes changes of protein function. ISGylation is known to inhibit the replication of many viruses, although pro-viral effects of ISGylation are also reported. Given that ISG15 and the enzymes involved in ISGylation are strongly induced upon virus infection, understanding the interplay between virus and ISGylation is an important issue in virus-host interaction. Nevertheless, viral substrates of ISG15 and viral strategies to regulate ISGylation-mediated antiviral responses are limited to only a few examples. In this study we demonstrate that ISGylation suppresses human cytomegalovirus (HCMV) infection but the virus is armed with countermeasures that consecutively reduce ISG15 transcription and protein ISGylation. Interestingly, a viral ISG15 target is found to inhibit ISGylation. This study highlights that ISGylation is a critical innate immune response against HCMV infection and interfering with ISG15-mediated anti-viral immunity is critical for productive viral infection. This research was supported by the Korean Health Technology R&D Project Grant funded by the Ministry of Health & Welfare (HI14C2114), Republic of Korea to JHA. The work was also supported by the DFG (SFB796) to TS. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Type I interferons (IFNs) are multifunctional cytokines that represent crucial components of the innate immune response to viral infection. Recognition of viral infection by host cells induces the synthesis of type I IFNs. Secreted IFNs interact with IFN receptors on target cells, triggering a signaling cascade that involves Janus kinase (JAK) and signal transducer and activator of transcription (STAT) families. Activated STAT1 and STAT2 heterodimerize and bind to IFN regulatory factor 9 (IRF9) to form a complex called IFN-stimulated gene factor 3 (ISGF3). This complex translocates into the nucleus and induces ISGs with diverse antiviral activities by binding to IFN-stimulated response elements (ISREs) in their promoters (for review [1]). ISG15 was identified as an IFN-inducible ubiquitin homolog. Like ubiquitin, its carboxy-terminal LRLRGG motif is required both for recognition by processing enzymes and covalent conjugation to lysine residues of target proteins. ISG15 modification (also termed ISGylation) is an IFN-stimulated and -regulated process that is found only in higher vertebrate animals and appears to modulate the function of target proteins (for review [2]). UBE1L is the E1 activating enzyme for ISG15 [3], and UbcH8, an ubiquitin E2 conjugating enzyme, also acts as the ISG15 E2 conjugating enzyme [4, 5]. HERC domain and RCC1-like domain containing protein 5 (Herc5), estrogen-responsive finger protein (EFP), and human homolog of Drosophila ariadne (HHARI) have been identified as E3 ligases for ISGylation in human cells [6–9]. ISG15, UBE1L, UbcH8, and Herc5 are IFN-inducible [4, 6, 10, 11]. Conjugated ISG15s are removed by an ISG15-specific protease, UBP43 (also known as USP18) [12]. Interestingly, UBP43 is also IFN-inducible [13, 14] and acts as a negative regulator of innate immune responses independent of its protease activity but dependent on its direct interaction with IFNAR2, a subunit of the type I IFN receptor [15]. Antiviral responses involving protein ISGylation have been reported against diverse viruses. Several cellular proteins involved in antiviral signaling, including RIG-I, MDA-5, STAT1, JAK1, IRF3, PKR, Mx1, and RNase L, were also identified or suggested as substrates for ISGylation [16–21]. ISGylation suppressed replication of diverse viruses, such as influenza virus (type A and B) [22–25], human immune deficiency virus (HIV) [26, 27], hepatitis C virus (HCV) [28–30], Japanese encephalitis virus [31], Sindbis virus [23, 32, 33], Ebola VP40 virus-like particle [34, 35], herpes simplex virus type-1 [23], murine γ-herpesvirus 68 [23], vaccinia virus [36], dengue and West Nile viruses [37], porcine reproductive and respiratory syndrome virus [38], Kaposi’s sarcoma-associated herpesvirus (KSHV) [39], and respiratory syncytial virus [40]. However, the antiviral mechanism of ISGylation against specific viruses is poorly understood. Herc5 associates with polyribosomes, and ISGylation appears to be restricted largely to newly synthesized proteins, suggesting that newly synthesized viral proteins may be primary targets of ISG15 [41]. ISGylation of NS1A in influenza A virus disrupted its association with importin-α, which mediates the nuclear import of NS1A, thus inhibiting viral replication [42]. ISGylation also suppressed the release of retrovirus particles by disrupting the budding process-related protein complex (for review [43]). An ISGylation-independent antiviral effect of ISG15 was also demonstrated in Chikungunya virus infection [44]. Although several studies have suggested a general role for ISG15 as an antiviral molecule, proviral effects of ISGylation have also been reported for certain viruses. In Newcastle disease virus (NDV) infection, ISGylation of RIG-I reduced IFN responses as a negative feedback regulation mechanism [45], whereas ISGylation of IRF3 stabilized IRF3 [21]. In addition, enhanced ISGylation by ISG15 overexpression promoted HCV production, while reduced ISGylation by UBE1L- or ISG15-knockdown inhibited HCV production [46, 47]. Therefore, the effects of global protein ISGylation appear to vary among the different viruses. In addition, ISG15 is secreted to the extracellular space as a free unconjugated form [48, 49]. While secretion of ISG15 in granulocytes is shown to activate T cells and natural killer cells to produce IFNγ in mycobacterial infection [50, 51], the role of secreted ISG15 in viral infection is not clear. Furthermore, a role of free ISG15 as a negative regulator that prevents IFNα/β overamplification and auto-inflammation by sustaining UBP43 levels has been suggested in humans but not in mice [52, 53]. Human cytomegalovirus (HCMV) is an opportunistic pathogen that causes severe disease complications and pathologies in newborns and immunocompromised individuals [54]. During productive infection, HCMV gene expression occurs sequentially in three phases: immediate-early (IE), early, and late. IE proteins and virion-associated tegument proteins play key roles in initiating viral gene expression and modulating host cell functions. HCMV employs several mechanisms to counteract IFN production and subsequent ISG activation. IE2 and pp65 inhibit IFN production [55–57], whereas IE1 suppresses the IFN response by directly binding to STAT2 [58–60] and PML [61, 62]. HCMV infection results in a decrease in the levels of JAK1 and p48, two components of the type I IFN signaling pathway [63, 64]. Modulation of the stability and phosphorylation of STAT proteins during HCMV infection was also reported [65, 66]. ISG15 transcription is induced in HCMV infection; however, its regulation during infection, the role of ISGylation in viral growth, and viral targets of ISG15 have not been characterized. In this study, we show that ISG15 expression and ISGylation are initially induced after HCMV infection but later suppressed by viral responses, and that IE1, a viral inhibitor of STAT signaling, plays an important role in reducing ISG15 transcription. By silencing the expression of E1, E2, and E3 ISGylation enzymes and of an ISG15 protease, we also demonstrate that ISGylation inhibits HCMV growth at multiple steps, including viral gene expression and virion release. Furthermore, we show that pUL26, a viral tegument protein, interacts with ISG15, E1, and E3 and is modified by ISG15, which inhibits pUL26 activity to promote viral growth. Moreover, we reveal that expression of pUL26 is able to suppress ISGylation induced by virus infection. Our results indicate that HCMV has evolved countermeasures to suppress ISG15 transcription and protein ISGylation, highlighting the important of the interplay between virus and ISG15 signaling during virus infection. Results Time course of ISG15 expression and protein ISGylation during HCMV infection The time course of ISG15 expression and protein ISGylation during HCMV infection were investigated with different multiplicity of infections (MOIs). In human fibroblast (HF) cells infected with HCMV (Towne), the levels of ISG15 and protein ISGylation were elevated by 24 h at all MOIs tested (MOIs of 0.2 to 10) (Fig 1A, lanes 2–6). At 48 and 72 h after infection, even greater levels of ISG15 expression and protein ISGylation were observed at relatively low MOIs (0.2, 0.5, and 1) (Fig 1A, lanes 13–15 and 24–26); however, the levels of free ISG15 and ISG15 conjugates at high MOIs (3 and 10) were much lower than those at low MOIs (Fig 1A, lanes 16–17 and 27–28). The time course of ISG15 expression and protein ISGylation was also examined in cells infected with UV-inactivated virus (UV-HCMV). In UV-HCMV infection, the levels of ISG15 and ISG15 conjugates were elevated at 24 and 48 h and correlated proportionally with MOI (Fig 1A, lanes 7–11 and 18–22). Levels of free ISG15 at 72 h induced by UV-HCMV were lower than those at 48 h, probably due to the termination of signaling (Fig 1A, compare lanes 29–33 and 18–22). The lack of viral gene expression in UV-HCMV infection was verified by the absence of viral IE protein expression. Collectively, these results comparing HCMV and UV-HCMV infection demonstrate that ISG15 expression and protein ISGylation are initially induced by HCMV infection, but are subsequently suppressed in a manner dependent on viral gene expression. 10.1371/journal.ppat.1005850.g001Fig 1 Time course of ISG15 expression and protein ISGylation during HCMV infection. (A) HF cells were mock-infected (M) or infected with HCMV (Towne) or UV-HCMV at different MOIs (0.2 to 10) as indicated. Cell lysates were prepared at 24, 48, and 72 h after infection and immunoblotted with antibodies for ISG15, IE1/IE2, or β-actin (a loading control). (B) HF cells were mock-infected or infected with HCMV or UV-HCMV at an MOI of 0.5 for 48 h. Cells were fixed with methanol and double-label IFA was performed with antibodies for ISG15 and UL112-113. Hoechst stain was used to stain cell nuclei. The images were obtained by confocal microscopy. Notably, we observed a greater induction of ISG15 and protein ISGylation with HCMV than with UV-HCMV at low MOIs (MOIs of 0.2, 0.5, and 1) (Fig 1A, compare lanes 2–4 and 7–9). We hypothesized that this was due to the induction of ISG15 at different level in uninfected cells that surround infected cells at low MOIs. To test this hypothesis and to examine the effect of HCMV infection on ISG15 expression at a single cell level, HF cells were infected with HCMV or UV-HCMV at a low MOI (0.5) and stained for ISG15 and viral UL112-113 proteins. We found that ISG15 expression was reduced in HCMV-infected cells, in which UL112-113 viral replication proteins were expressed at high level; however, it was markedly increased in neighboring uninfected cells compared to mock-infected cells. We also found that the levels of ISG15 in uninfected neighboring cells were higher in HCMV infection than in UV-HCMV infection (Fig 1B). This result suggests that an indirect effect of virus infection on neighboring uninfected cells is responsible for the greater induction of ISG15 and protein ISGylation by HCMV than by UV-HCMV at low MOIs. HCMV IE1 suppresses ISG15 transcription Since HCMV IE1 inhibits the activation of ISRE-containing promoters by sequestering STAT2 [58–60] and PML [61], IE1 expression may be responsible for the suppression of free ISG15 and ISG15 conjugate levels during HCMV infection. To test this hypothesis, we first compared the effects of wild-type HCMV, UV-HCMV, and IE1-deleted mutant virus (CR208) infection (MOI of 3) on ISG15 transcription by RT-PCR. The CR208 virus exhibited an MOI-dependent growth pattern with showing a severe growth defect in HF cells at low MOIs but normal growth at high MOIs [67]. All viruses increased ISG15 mRNA levels 12 h after infection; however, ISG15 induction was terminated earlier for HCMV than UV-HCMV, and not appreciably terminated for CR208, which continued to produce high levels of ISG15 transcripts even at a late stage of infection (72 h) (Fig 2A). ISG15 transcription induced by UV-HCMV infection might be gradually decreased due to the termination of the IFN signaling through several negative regulatory mechanisms. However, when cells were infected with CR208 at this high MOI, the replication of virus without IE1 appeared to lead to a robust activation of IFN signaling. This result indicates that IE1 indeed plays an important role in reducing ISG15 transcription during HCMV infection. 10.1371/journal.ppat.1005850.g002Fig 2 Effects of IE1 on ISG15 transcription and protein ISGylation. (A-B) HF cells were mock-infected (M) or infected with wild-type HCMV, UV-HCMV, or CR208 virus at an MOI of 3. Total RNAs were prepared at the indicated time points and the levels of ISG15 and β-actin transcripts were determined by RT-PCR (A). Cell lysates were also prepared and analyzed by immunoblotting as in Fig 1A (B). (C) Control and IE1-expressing HF cells produced by retroviral vectors were mock-infected or infected with wild-type HCMV, UV-HCMV, or CR208 virus at an MOI of 3, or treated with IFNβ (1,000 U/ml) for 48 h. Immunoblotting was performed with antibodies for ISG15, IE1/IE2, p52 (encoded by UL44) and β-actin. (D) 293T cells were co-transfected with plasmids expressing myc-ISG15 (wild-type, ISG15GG, or ISG15AA), HA-UBE1L (E1), Flag-UbcH8 (E2), or HA-Herc5 (E3) as indicated. At 48 h after transfection, cell lysates were prepared by boiling the cell pellets in sodium dodecyl sulfate (SDS) loading buffer and immunoblotted with anti-myc and anti-β-actin (a loading control) antibodies. (E and F) Co-transfection/ISGylation assays were performed in 293T cells with or without increasing amounts of plasmid expressing IE1 (E) or in control and IE1-expressing HF cells (F). Immunoblotting was performed with anti-myc, anti-IE1, and anti-β-actin antibodies. We also compared the levels of ISG15 and ISG15 conjugates in HCMV, UV-HCMV, and CR208 infection. When cells were infected with viruses at an MOI of 3, free ISG15 levels were elevated at 12 h by all viruses, whereas the level of ISG15 conjugates markedly increased after 24 h (Fig 2B). The delayed induction of ISGylation is similar to what was observed in IFNβ-treated cells [68] and may result from the delayed induction of the ISGylation machinery. Consistent with the results shown in Fig 1A, UV-HCMV induced more ISG15 conjugates than wild-type virus at this MOI (Fig 2B, compare lanes 3–5 and 8–10). Importantly, CR208 also resulted in greater ISGylation than wild-type virus (Fig 2C. compare lanes 3–5 and 13–15), indicating that IE1 is required for the suppression of ISG15 expression and ISGylation during HCMV infection. Immunoblot analysis demonstrated that IE1 and IE2 failed to be expressed in UV-HCMV infection and that IE1 failed to be expressed and IE2 levels were reduced in CR208 infection under these experimental conditions (Fig 2B and 2C). Although our results show that IE1 is largely responsible for the suppression of ISG15 transcription, it is notable that CR208 infection resulted in slightly lower levels of ISGylation compared to UV-HCMV infection (Fig 2B, compare lanes 10 and 15). Considering that ISG15 transcript levels remained high up to 72 h in CR208-infected cells (Fig 2A), this finding suggests that other viral processes, besides those mediated by IE1, may also be implicated in the downregulation of protein ISGylation. The inhibition of ISG15 expression by IE1 was further investigated using IE1-overexpressing HF cells generated using retroviral vectors. Control and IE1-overexpressing HF cells were infected with HCMV, UV-HCMV, or CR208. The results of immunoblot analysis showed that IE1 overexpression suppressed the induction of ISG15 and ISG15 conjugates by both virus infection and IFNβ treatment (as a control) (Fig 2C, compare lanes 2–5 and 7–10), further supporting the critical role of IE1 in reducing ISG15 expression. Although reduction of protein ISGylation during HCMV infection may be largely attributed to suppression of ISG15 transcription by IE1, it cannot be ruled out that IE1 also affects the ISGylation reaction. Therefore, we studied whether IE1 directly affects ISG15 conjugation reactions using co-transfection/ISGylation assays. To set up co-transfection/ISGylation assays, two different ISG15 forms, an active form with a stop codon immediately after the C-terminal double glycine residues (ISG15GG) and an inactive form with the double glycine residues substituted with alanine residues (ISG15AA), were employed. In cells transiently co-transfected with ISG15, UBE1L (E1), UbcH8 (E2), and Herc5 (E3), intact ISG15 and ISG15GG (an active form), but not ISG15AA (an inactive form), were conjugated to proteins (Fig 2D). When co-transfection/ISGylation assays were performed with or without IE1 overexpression, IE1 expression by co-transfection or retroviral transduction did not inhibit levels of ISG15 conjugates, indicating that IE1 does not inhibit the enzymatic cascade of reactions required for protein ISGylation (Fig 2E and 2F). ISGylation inhibits HCMV growth The role of protein ISGylation in HCMV infection was investigated by silencing expression of ISGylation enzymes using RNA interference. HF cells expressing control shRNA or shRNA for UBP43, an ISG15-specific protease, were generated using lentiviral vectors. IFNα treatment of normal and control shRNA (shC)-expressing HF cells induced the expression of UBP43 proteins; however, UBP43-specific shRNAs (shUBP43-1 and shUBP43-2) efficiently suppressed this induction (Fig 3A). UBP43 knockdown enhanced protein ISGylation in cells infected with UV-HCMV that stimulates interferon signaling (Fig 3B), but reduced the production of progeny virions in HCMV-infected cells to nearly 10% of that in control cells (Fig 3C), suggesting that enhanced ISGylation by UBP43 knockdown attenuates HCMV growth. However, UBP43 is also known to negatively regulate IFN signaling by downregulating JAK/STAT signaling [15, 69]. Consistently, we found that UBP43 knockdown enhanced STAT2 phosphorylation in HCMV-infected cells (Fig 3D). Therefore, it is possible that the reduction of HCMV growth in UBP43 knockdown cells is also a result of other aspects of the IFN response. 10.1371/journal.ppat.1005850.g003Fig 3 Effects of UBP43 or Herc5 knockdown on HCMV growth. (A) Normal HF cells or cells expressing control (shC) or UBP43-specific shRNAs (shUBP43-1 and -2) were treated with IFNα (2,000 U/ml) for 24 h and the UBP43 levels were analyzed by immunoblotting using anti-UBP43 antibody. The tubulin levels were shown as a loading control. Arrowheads indicate two UBP43 bands with different molecular weights and an open circle indicates non-specific bands. (B) Control and UBP43 knockdown HF cells were infected or not with UV-HCMV at an MOI of 3. At 24 h after infection, cell lysates were immunoblotted with antibodies for ISG15 and β-actin. (C) Control and UBP43-knockdown HF cells were infected with HCMV at an MOI of 3. At 5 days after infection, viral supernatants were collected and the levels of progeny virions were measured by infectious center assays. Scatter plots are shown. (D) Control and UBP43-knockdown cells were infected as in (C). Cell lysates were prepared at indicated time points and immunoblotted with antibodies for ISG15, STAT2, phosphorylated STAT2 (on Tyr 689), and β-actin. (E) Control HF cells or cells expressing Herc5-specific shRNA were mock-infected or infected with UV-HCMV at an MOI of 3. The Herc5 transcript levels were determined at 24 h after infection by qRT-PCR. The β-actin mRNA levels were used for normalization. Values are an average of duplicated assays; error ranges are indicated. (F) Control and Herc5-knockdown HF cells were infected or not with UV-HCMV at an MOI of 3. At 24 h after infection, cell lysates were immunoblotted with antibodies for ISG15 and β-actin. (G) Control and Herc5-knockdown HF cells were infected with HCMV at an MOI of 0.1. At 9 days after infection, virus titers in the culture supernatants were measured by infectious center assays. The effect of ISGylation on HCMV growth was further assessed by depleting Herc5, a major ISG15 E3 ligase in human [6, 7]. Herc5 knockdown HF cells were generated using lentiviral vectors expressing shRNAs. qRT-PCR assays confirmed the efficient reduction of Herc5 transcript levels in cells expressing Herc5-specific shRNAs (shHerc5-1 and shHerc5-2) compared to in cells expressing control shRNA (shC) after UV-HCMV infection (Fig 3E). We found that Herc5 knockdown markedly reduced protein ISGylation as expected (Fig 3F), but increased virus titers by 8- to 11-fold compared to in control cells (Fig 3G), demonstrating that the reduction of ISGylation by Herc5 knockdown facilitates HCMV growth. Similar enhancement of HCMV growth was observed in HF cells depleted of UBE1L (E1) or UbcH8 (E2) (S1 Fig). Collectively, our results with UBP43, Herc5, UBE1L, and UbcH8 knockdown cells demonstrate an inverse relationship between ISGylation and HCMV growth, indicating a general inhibitory role of ISGylation in HCMV infection. We also investigated whether enhanced ISGylation affects HCMV growth by ectopically expressing ISG15 together with ISGylation enzymes. HF cells co-transfected with plasmids expressing myc-ISG15GG or myc-ISG15AA and ISGylation enzymes were infected with HCMV and the production of progeny virions was compared. The results showed that the expression of myc-ISG15GG and ISGylation enzymes induced higher levels of ISG15 conjugates and reduced progeny virus titers to 50% of control, whereas the expression of ISG15AA did not significantly affect the levels of ISG15 conjugates or progeny virions (S2 Fig). These results are also consistent with our results using shRNA-expressing cells in which we showed an inverse relationship between levels of ISG15 conjugates and HCMV growth. ISGylation reduces HCMV gene expression and virion release To investigate the mechanisms by which ISGylation inhibits HCMV growth, we first compared the expression profile of viral proteins in control and Herc5 knockdown cells. When cells were infected with an MOI of 0.2, levels of major IE (IE1 and IE2), early (p52), and late (pp28) viral proteins produced at 24, 48, 72, and 96 h were higher in Herc5 knockdown cells than in control cells (Fig 4A, left panels). This effect of knocking down Herc5 diminished when cells were infected at higher MOIs. At an MOI of 1, IE1 expression was similar between control and Herc5 knockdown cells, although the levels of IE2, p52, and pp28 were slightly increased in Herc5 knockdown cells (Fig 4A, center panels). At an MOI of 5, levels of viral proteins were comparable between control and Herc5 knockdown cells (Fig 4A, right panels). Progeny virion titers measured in the culture supernatant correlated with the levels of expressed viral proteins (Fig 4B). These results demonstrate that reduced ISGylation by silencing Herc5 promotes viral gene expression, an effect that is more evident at lower MOIs. 10.1371/journal.ppat.1005850.g004Fig 4 Effect of Herc5-knockdown on HCMV gene expression and virion release. (A-B) Control HF cells expressing shC (C) and Herc5-knockdown cells expressing shHerc5-2 (H) were mock-infected or infected with HCMV at an MOI of 0.2, 1, or 5. Cell lysates were immunoblotted at the indicated time points with antibodies for IE1/IE2, p52, pp28 (encoded by UL99), ISG15, and β-actin (A). Virus titers produced in the culture supernatants for 24, 48, and 72 h were determined by infectious center assays (B). (C) Control (shC) and Herc5-knockdown HF cells were infected with HCMV at an MOI of 1 or 5. At 4 days after infection, virus titers produced within the cells (cell-associated) and in the culture supernatants (extracellular) were determined by infectious center assays. The total amounts of infectious units (cell-associated plus extracellular) produced in control and Herc5-knockdown cells were shown at the top. The percentages of cell-associated virions among total virions in control or Herc5-knockdown cells were shown at the bottom. The results are the mean values of three independent experiments with standard errors. We further investigated whether ISGylation affects the activation of viral promoters using reporter assays (S3 Fig). HF cells co-transfected with ISG15, UBE1L, UbcH8, and Herc5 exhibited substantially increased protein ISGylation, whereas cells co-transfected without Herc5 did not. In HF cells exhibiting enhanced ISGylation by co-transfection, the activity of viral MIE promoter was repressed to 50% of that observed in control cells. Similarly, IE2-mediated activation of viral early [UL112-113 and UL54 (POL)] and late [UL99 (pp28)] promoters was also suppressed in cells showing enhanced ISGylation. In control experiments, ectopic expression of Herc5 alone was not sufficient to increase ISGylation and therefore did not affect viral promoter activation. These results demonstrate that enhanced ISGylation inhibits the activation of viral promoters. We also assessed whether ISGylation inhibits HCMV virion release as is the case in HIV infection [43]. Control and Herc5 knockdown cells were infected with HCMV at an MOI of 1 or 5 and the levels of cell-associated and extracellular progeny virions produced were compared. The results showed that although more progeny virions were produced in Herc5 knockdown cells than in control cells (Fig 4B), the percentages of cell-associated virions to total virions were lower in Herc5-knockdown cells, 3.5% (MOI 1) and 3.8% (MOI 5), than in control cells, 19.7% (MOI 1) and 17% (MOI 5) (Fig 4C). These results indicate that, as observed in HIV infection, ISGylation inhibits HCMV virion release. Covalent and non-covalent interaction of pUL26 with ISG15 To identify potential HCMV proteins that interact with the ISG15 pathway, we screened the HCMV ORF library [70] for ISG15- or UBE1L-interacting proteins using yeast two-hybrid assays. Twenty HCMV proteins were identified as potential ISG15-interacting proteins, and five viral proteins (encoded from RL1, UL19, UL21A, UL26, and UL30) were found to interact with both ISG15 and UBE1L. Most of these ISG15 binding and all of UBE1L binding in yeast assays were also detected by co-immunoprecipitation (co-IP) assays (S4 Fig and S2 Table). Among them, the UL26 gene encodes the tegument proteins, p27 and p21, which are produced using two in-frame start codons and are shown to regulate viral gene expression, NF-κB signaling, and virion stability [71–74]. Since UL26 interacted with both ISG15 and UBE1L and its role in viral growth was relatively well reported compared to others, we further investigated the interaction of pUL26 with the ISG15 pathway. In co-IP assays, UL26-p21 interacted with ISG15AA, demonstrating that ISG15 can non-covalently interact with UL26 (Fig 5A). In similar co-IP assays, UL26-p21 also interacted with UBE1L and Herc5, but not UbcH8 (Fig 5B–5D). When co-IP assays were performed using HCMV (Towne)-infected cell lysates, immunoprecipitation with an anti-UL26 antibody co-precipitated unconjugated free ISG15 (Fig 5E), and in a reciprocal experiment, immunoprecipitation with an anti-ISG15 antibody co-precipitated UL26-p21 (Fig 5F), indicating that UL26-p21 interacts with ISG15 during virus infection. 10.1371/journal.ppat.1005850.g005Fig 5 Interactions of pUL26 with ISG15, UBE1L, and Herc5, and ISGylation of pUL26. (A-D) 293T cells were co-transfected with plasmids encoding myc-ISG15AA, HA-UBE1L, Flag-UbcH8, HA-Herc5, SRT-UL26-p21, Flag-UL26-p21, or HA-UL26-p21, as indicated. At 48 h after transfection, cell lysates were prepared and immunoprecipitated with anti-myc (A), anti-flag (B), or anti-HA (C and D) antibodies, followed by immunoblotting with anti-SRT (A), anti-HA (B), anti-flag (C), or anti-myc antibodies. To determine the expression levels of each protein, whole cell lysate were also immunoblotted. (E and F) HF cells were infected with HCMV(Towne) at an MOI of 3. At 48 h after infection, cell lysates were subjected to co-IP assays using anti-UL26 antibody and control IgG (E) or using anti-ISG15 antibody and control IgG (F), followed by immunoblotting with anti-ISG15 (E) or anti-UL26 (F) antibodies. Immunoblotting of whole cell lysates were performed to determine the expression levels of each protein. (G) HF cells were infected as in (E) for 48 h. Cell lysates were prepared and immunoprecipitated with anti-UL26 antibody and control IgG as described for co-IP assays to detect ISGylated protein (see Materials and Methods). Immunoprecipitated samples and total cell lysates were immunoblotted with anti-ISG15 and anti-UL26 antibodies. (H and I) HF cells were infected with wild-type HCMV (Toledo) or its recombinant virus expressing UL26-HA proteins at an MOI of 3. Cell lysates were prepared at the indicated time points and immunoblotted with anti-IE1/IE2, anti-HA, and anti-β-actin antibodies (H). The culture supernatants were collected at 3 days after infection and virus titers were determined by infectious center assays (I). (J) HF cells were mock-infected or infected with recombinant virus expressing UL26-HA at an MOI of 3 for 48 h. Cell lysates were subjected to co-IP assays using anti-HA antibody and anti-Flag antibody, followed by immunoblotting with anti-ISG15 antibody as in (G). Immunoblotting of whole cell lysates were performed to determine the expression levels of each protein. (K) HF cells were mock-infected or infected as in (J). Cell lysates were prepared and immunoprecipitated with anti-HA antibody as in (G). Immunoprecipitated samples and total cell lysates were immunoblotted with anti-ISG15 and anti-HA antibodies as indicated. It has been suggested that newly synthesized viral proteins may be broadly modified by ISG15 during virus infection [41]. Therefore, we also investigated whether UL26 is covalently conjugated by ISG15 during HCMV infection. Cell lysates prepared from HCMV (Towne)-infected cells were boiled in SDS-containing buffer and then immunoprecipitated with an anti-UL26 antibody. Immunoblotting of the sample with anti-ISG15 antibody revealed a band that is consistent with an ISG15-modified form of UL26-p21 (Fig 5G). To further investigate these non-covalent and covalent interaction of UL26 proteins with ISG15, we generated a recombinant HCMV (Toledo) expressing UL26 proteins tagged at their carboxyl termini with an HA tag (S5 Fig). Compared to the wild-type virus, the recombinant virus produced equivalent amounts of viral major IE (IE1 and IE2) and UL26 proteins, and progeny virions (Fig 5H and 5I). When co-IP assays were performed with lysates from UL26-HA virus infected cells, immunoprecipitation of UL26 proteins with an anti-HA antibody co-precipitated free ISG15 (Fig 5J). When HA-UL26 virus-infected cells were subjected to co-IP assays to detect ISGylated proteins, bands that are consistent with ISG15-modified forms of UL26 proteins were detected (Fig 5K). Since ISG15 can modify ubiquitin, forming ISG15-ubiquitin mixed chains [75] and a single lysine reside can be poly-ISGylated [20, 76], the smear bands of ISGylated UL26 appear to be UL26 proteins that contain ISG15-ubiquitin mixed chains or poly-ISG15 chains. Taken together, our data suggest that the UL26-encoded proteins non-covalently interact with ISG15 and are also covalently modified by ISG15. In a control experiment, we found that some viral proteins, which did not interact with ISG15, UBE1L, and Herc5 in co-IP assays, could be ISGylated in co-transfection/ISGylation assays (S6 Fig), supporting the concept that viral proteins may be broadly ISGylated during infection due to IFN-upregulated expression of Herc5 (E3) in polyribosomes [41]. ISGylation of UL26 regulates protein stability and inhibits UL26 activities to suppress NF-κB signaling and promote viral growth The UL26 ORF from the Towne strain contains three lysine residues (K54, K136, and K169) and K54 and K169 are conserved in human CMVs (Fig 6A). To determine the ISGylation sites of UL26 proteins, we performed co-transfection/ISGylation assays using wild-type UL26-p21 or its mutants in which lysine residues were replaced with arginines. The results showed that the K54R, K136R, and K169R mutants were still ISGylated; however, the K136/169R double mutant was not ISGylated, indicating that K136 and K169 are the major ISGylation sites (Fig 6B). When HF cells expressing wild-type UL26-p21 or K136/169R mutant were generated by retroviral vectors, expression level of the K136/169R mutant protein was higher than that of the wild-type protein (Fig 6C), suggesting that ISGylation of UL26 proteins may regulate protein stability. 10.1371/journal.ppat.1005850.g006Fig 6 Roles of UL26 ISGylation in viral growth. (A) Three lysine residues of UL26-p21 from Towne strain and their conservation in other human (Toledo, Ad169, and Merlin) and primate CMVs (chimpanzee CMV, ChCMV; rhesus CMV, RhCMV; simian CMV, SCMV) are shown. (B) 293T cells were co-transfected with plasmid expressing SRT-UL26-p21 (wild-type or lysine to arginine mutants), myc-ISG15 (with GG or AA terminus), HA-UBE1L, Flag-UbcH8, or HA-Herc5 as indicated. At 48 h after transfection, cell lysates were immunoprecipitated with anti-SRT antibody. Immunoprecipitated samples and whole cell lysates were detected by immunoblotting with anti-myc and anti-SRT antibodies. (C) Control HF cells or cells expressing UL26-p21 (wild-type or K136/359R mutant) were produced by retroviral vectors. Cell lysates were prepared and immunoblotted with antibodies for UL26 and β-actin (a loading control). (D-F) Control HF cells or cells expressing UL26-p21(K136/169R) or UL26-p21(K136/169R)-ISG15AA were produced by retroviral transduction. To determine the expression levels of UL26 proteins, cell lysates were immunoblotted with antibodies for IE1/IE2, UL26, and β-actin (D). Cells were treated or not with 50 ng/ml of TNFα for the indicated times. Immunoblotting was performed to detect the levels of p65 and its phosphorylated form. The levels of UL26-p21, its ISG15AA-fused form, and β-actin were also shown (E). Cells were infected with the UL26-deleted mutant virus (AD169) at an MOI of 0.2. At 72 h after infection, virus titers in the culture supernatants were determined by infectious center assays (F). The SUMO fusion proteins were often used to study the function of SUMO modification of proteins [77–81]. To investigate the effect of UL26 ISGylation on its activity, we used the UL26-p21(K136/169R)-ISG15AA fusion protein as a surrogate for the ISG15-modified form of UL26-p21. We produced the ISG15 fusion to the lysine mutant form of UL26 to compare the activities of ISGylation-defective UL26 and its ISG15 fusion form. HF cells expressing the K136/169R mutant or the UL26-p21(K136/169R)-ISG15AA fusion protein were generated by retroviral vectors (Fig 6D). UL26 proteins have been shown to inhibit TNFα-induced NF-κB activation [74]. UL26-p21(K136/169R) moderately inhibited TNFα-induced NF-κB activation but UL26-p21(K136/169R)-ISG15AA did not, suggesting that ISGylation of UL26 inhibits its activity to downregulate NF-κB signaling (Fig 6E). We further investigated the effect of UL26 ISGylation on its role in viral growth. When control and UL26-expressing cells were infected with the UL26-deleted mutant HCMV (AD169) [72] and the levels of progeny virus titers were compared, pre-expression of the K136/169R mutant significantly increased the growth of mutant virus but its ISG15AA fusion protein did not (Fig 6F). Together, these results indicate that ISGylation of UL26 inhibits its activities to suppress NF-κB signaling and promote the growth of UL26-deleted mutant virus. UL26 inhibits protein ISGylation Since influenza virus NS1B antagonizes ISGylation in human cells via direct binding to ISG15 [3, 82], we tested whether UL26 can regulate protein ISGylation. We found that expression of UL26-p21 reduced the levels of ISG15 conjugates in cells co-transfected with UBE1L (E1), UbcH8 (E2), Herc5 (E3), and ISG15GG, an active form of ISG15 (Fig 7A). In a similar assay, ISGylation of charged multivesicular body protein (CHMP) 5, a component of the endosomal sorting complex required for transport (ESCRT) machinery, was inhibited by UL26-p21 expression (Fig 7B). When control and UL26-p21-expressing HF cells were infected with UV-HCMV, the levels of ISG15 conjugates were significantly reduced in UL26-p21-expressing cells compared to control cells (Fig 7C). This effect of UL26 was also found with the K136/169R mutant protein (Fig 7D), which still interacted with ISG15AA in co-IP assays (Fig 7E), suggesting that the inhibitory effect of UL26 on ISGylation is not dependent on its own ISGylation. In control experiments, viral proteins such as pUL85, pUL71, and IE2 did not inhibit ISGylation of cellular proteins in co-transfection/ISGylation assays (S6 and S7 Figs). 10.1371/journal.ppat.1005850.g007Fig 7 Inhibition of protein ISGylation by UL26. (A) 293T cells were co-transfected with expression plasmids as indicated. At 48 h after transfection, cell lysates were immunoblotted with anti-myc and anti-SRT antibodies. (B) 293T cells were co-transfected with expression plasmids as indicated. At 48 h after transfection, cell lysates were prepared and immunoprecipitated with anti-SRT antibody, followed by immunoblotting with anti-myc antibody. Expression levels of myc-ISG15, SRT-CHMP5, and β-actin in whole cell lysates were also determined by immunoblotting. (C) Control and UL26-p21-expressing HF cells produced by retroviral vectors were infected with UV-HCMV at an MOI of 1. The samples were prepared at the indicated time points and the levels of ISG15, UL26 and β-actin were determined by immunoblotting. The relative levels of ISGylated UL26 protein over unmodified protein (normalized with β-actin) are also shown as graphs. (D) Control or UL26 (wild-type or mutant)-expressing HF cells were infected with UV-HCMV and immunoblotting was performed as described in (C). (E) 293T cells were co-transfected with plasmids expressing SRT-UL26-p21 (wild-type or K136/169R) or myc-ISG15AA as indicated. At 48 h after transfection, cell lysates were immunoprecipitated with anti-SRT antibody. Immunoprecipitated samples and whole cell lysates were detected by immunoblotting. (F) HF cells expressing control shRNA (shC) or shRNA specific for IE1 (shIE1) produced by retroviral transduction were infected with wild-type or ΔUL26 mutant virus at an MOI of 5 for the indicated time points. Cell lysates were immunoblotted with antibodies for ISG15, viral proteins (IE1, IE2, and UL26), and β-actin. (G) HF cells were pre-treated or not with IFNβ (100 U/ml) for 24 h and then infected with wild-type or ΔUL26 mutant virus at an MOI of 1. The production of progeny virions in the culture supernatant at 6 days after infection was measured by infectious center assays. The results shown are the mean values and standard errors of four independent experiments. (H) Summary for the HCMV strategy that consecutively inhibits ISG15 transcription and protein ISGylation. IE1 inhibits ISG transcription, while UL26 inhibits protein ISGylation. To further investigate the inhibitory effect of UL26 on ISGylation during HCMV infection, we compared ISGylation levels between cells infected with wild-type and UL26-deleted mutant viruses. With an MOI of 0.2, we found that the ISGylation levels were initially increased until 48 h but gradually decreased at 72 h and 96 h, and that the UL26-deleted virus infection showed only minimally increased levels of ISGylation at late times of infection compared to wild-type virus infection (S8 Fig, compare lanes 5–7 and 11–13). We reasoned this minimal effect of UL26 to the suppression of ISG15 transcription by IE1. Therefore, to minimize the effect of IE1 expression we performed a similar experiment in cells expressing shRNA for IE1 (shIE1). Control shRNA (shC) and shIE1-expressing HF cells, which were produced by retroviral vectors, were infected with wild-type and UL26-deleted mutant viruses at an MOI of 5. The results of immunoblotting showed that the expression of shIE1 substantially reduced the IE1 protein accumulation compared to control cells (Fig 7F, compare lanes 2–5 and 7–10), and that in shIE1-expressing cells the ISGylation levels were substantially increased at 24 h after infection and gradually decreased at 48 h and 72 h under these experimental conditions (Fig 7F, lanes 8–10). Notably, we found that the UL26-deleted virus less effectively reduced the ISGylation levels than wild-type virus at the late phase of infection (Fig 7F, compare lanes 8–10 and 13–15). These results demonstrate an inhibitory effect of UL26 on ISGylation during virus infection. Since HCMV growth was suppressed by IFN-induced ISGylation and UL26 inhibited ISGylation, we tested whether UL26-deleted mutant virus is more susceptible to type I IFN treatment than wild-type virus. We found that while the titers of wild-type virus were reduced by 12-fold by IFNβ pre-treatment, those of UL26-deleted virus were reduced by 40-fold, indicating that UL26-deleted virus less effectively overcomes the IFNβ-mediated anti-viral responses than wild-type virus (Fig 7G). Overall, our results demonstrate that, in addition to IE1 that suppresses ISG15 transcription, UL26 plays an important role in evading the ISG15-associated antiviral responses by inhibiting protein ISGylation (Fig 7H). Discussion Our analysis with UV-HCMV and IE1-deleted mutant virus demonstrates that ISGylation is induced by HCMV infection and that IE1 plays a central role in downregulating ISGylation by reducing ISG15 transcription. The latter is consistent with previous findings that IE1 represses transcription of ISGs by sequestrating STAT2 [58–60] and PML [61, 62]. Notably, although IE1 effectively reduced ISG expression, the level of ISGylation during HCMV infection largely depended on the MOI. Our data provide evidence for the antiviral roles of ISGylation during HCMV infection. To discern the effects of free ISG15 expression and protein ISGylation on virus replication, we used HF cells in which a specific ISGylation enzyme was depleted by shRNA. The data consistently showed an inverse relationship between the level of ISG15 conjugates and HCMV growth; i.e., enhanced ISGylation by UBP43 knockdown decreased viral growth, while reduced ISGylation by depletion of UBE1L, UbcH8, or Herc5 increased viral growth. Therefore, we conclude that protein ISGylation in general inhibits HCMV growth. Free ISG15 has also been shown to inhibit the replication of certain viruses [34, 35, 44]. In our analysis, however, overexpression of ISG15GG with UBE1L, UbcH8, and Herc5 led to a mild reduction of HCMV growth, whereas ISG15AA, an inactive form, did not significantly affect viral growth, suggesting that expression of free ISG15 prior to HCMV infection may minimally affect viral growth in cultured cells. It should be noted that given the role of free ISG15 in stabilizing UBP43 (or USP18), a negative regulator of IFN signaling [53], overexpression or depletion of ISG15 might affect both positive and negative activities of ISG15 to viral growth. HCMV replication was inhibited by ISGylation at multiple steps. First, the expression of viral genes was inhibited under conditions where the level of ISG15 conjugates was increased. Viral regulators that promote viral gene expression may be a direct target of ISGylation. However, we could not observe ISGylation of IE1, which is responsible for the activation of MIE promoters, or IE2, a strong transactivator of viral early and late genes. ISGylation of cellular proteins that in particular play a role in innate immune responses may affect viral gene expression. Notably, ISGylation of IRF3 increases its stability and enhances IRF3-mediated transcriptional activation during Sendai virus infection [16, 21] and ISGylation of 4EHP, an mRNA 5' cap structure-binding translation suppressor, plays a role in IFN-induced innate immune response [9]. Second, HCMV virion release was inhibited by ISG15 conjugation. The inhibitory role of ISG15 expression in the budding process of enveloped viruses has been demonstrated in retrovirus infection. ISG15 expression inhibits ubiquitination of HIV Gag and tumor susceptibility gene-101 (Tsg101) proteins, leading to disruption of their interaction [26]. ISG15 conjugation of CHMP5, 2A, and 6 in the ESCRT machinery causes release of vacuolar protein-sorting 4 (VPS4) from the membrane, leading to inhibition of virion release [43]. The ESCRT machinery seems to be involved in the process of HSV-1 and HCMV maturation [83–86]. Therefore, ISGylation may affect the HCMV maturation process by targeting components involved in the ESCRT machinery. Although the antiviral role of ISG15 expression and ISGylation has been demonstrated in several viruses, studies on viral targets for ISGylation and viral strategies that interfere with the ISG15-mediated antiviral functions are limited to a few examples. ISGylation of NS1A of influenza A virus inhibits virus replication by interfering with NS1A nuclear import [42]. KSHV vIRF1 is ISGylated but it role on vIRF1 function is not known [39]. In the present study, we demonstrated that HCMV pUL26 is a target for ISGylation. UL26 ISGylation appears to regulate protein stability by competing with ubiquitination. Our analysis using the UL26-p21-ISG15 fusion protein demonstrated that ISGylation inhibits the activities of UL26-p21 to downregulate the TNFα-mediated NF-κB activation and to complement the growth defect of UL26-deleted virus. Therefore, ISGylation of pUL26 is thought to inactivate its function, suppressing HCMV growth. Given the notion that newly synthesized proteins are targeted extremely broadly, perhaps stochastically, by Herc5 in polyribosomes [41], several other HCMV proteins may be ISGylated during infection. We indeed observed that some viral proteins, which did not interact with ISG15, UBE1L, and Herc5 in co-IP assays, were ISGylated in our co-transfection/ISGylation assays. However, we think that ISGylation occurs in a manner dependent on the context of each protein. Furthermore, since several HCMV proteins were bound to ISG15 and/or UBE1L in yeast and co-IP assays, it is also likely that other viral proteins besides pUL26 may affect ISGylation. Identification of more HCMV proteins that interact with the ISGylation system and studies on their functional relevance are warranted. A few examples for viral regulation of ISGylation have been described. The NS1 protein of influenza B virus non-covalently binds to ISG15 and inhibits ISGylation [3]. Similarly, the vaccinia virus E3 protein interacts with ISG15 and disrupts its antiviral activity [36]. Nairoviruses and arteriviruses have shown to encode ovarian tumor domain (OTU)-containing proteases that hydrolyze ISG15 from target proteins [87] and severe acute respiratory syndrome (SARS) coronavirus encoded the papain-like protease that cleaves both ubiquitin and ISG15 [88]. Recently, KSHV-encoded vIRF1 was shown to interact with Herc5 [39]. In this study, we demonstrated that HCMV UL26-p21 is able to non-covalently bind to ISG15, UBE1L, and Herc5, and inhibit ISGylation. Whether binding all of ISG15, UBE1L, and Herc5 is critical for pUL26 to inhibit ISGylation is not clear and needs to be further investigated. Comparative analysis of wild-type and UL26-deleted mutant viruses demonstrate that de novo expression of UL26 in virus-infected cells correlated with reduced accumulation of ISG15 conjugates. In addition to IE1, the presence of additional viral functions that downregulate protein ISGylation was prompted by our observation that UV-HCMV infection resulted in a higher level of ISG15 conjugates than IE1-deleted mutant virus at late times (Fig 2B). UL26 deletion mutant virus shows a moderate growth defect at low MOIs [72, 89]. Notably, the UL26 virus (in AD169 strain) showed a reduced plaque formation at low MOIs and this defect could be rescued by IE1 overexpression [72]. More importantly, like IE1-deleted virus [58–60], the growth of UL26-deleted virus was more sensitive to pre-treatment of type I IFNs [74] (and this study), suggesting a role of UL26 in antagonizing type I IFN response. It is likely that a common ISG15-targeting mechanism is shared by influenza B virus, vaccinia virus, and HCMV. HCMV also encodes a tegument protein pUL48 that contains a deubiquitinating protease domain; however, its activity was specific for ubiquitin and did not cleave ISG15 [90]. In this study, we demonstrate that the cellular ISGylation system is a critical part of cellular innate immune response against HCMV infection. We also provide evidence that HCMV has developed several strategies to disarm the ISGylation-mediated antiviral activity; IE1 reduces ISG15 transcription and UL26 inhibits protein ISGylation. This interplay of HCMV with the cellular ISGylation system may be critical for the virus to successfully establish a persistent infection. Materials and Methods Cell culture, transfection, and virus Human foreskin fibroblast (HF) (ATCC) and human embryonic kidney (HEK) 293T cells (ATCC) were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, penicillin (100 U/ml), and streptomycin (100 μg/ml). DNA transfection of 293T cells was performed using the N,N-bis-(2-hydroxyethyl)-2-aminoethanesulfonic acid-buffered saline (BBS) version of the calcium phosphate procedure. Electroporation of HF cells was conducted using a Microporator MP-100 (Digital Bio), as described previously [59]. Stocks for the parent Towne virus and the CR208 mutant virus with IE1-deleted were prepared in ihfie1.3 cells as described previously [59]. The HCMV (Towne strain)-GFP virus was grown in HF cells after electroporation with the bacmid DNAs. Wild-type and UL26-deleted HCMVs (AD169 strain) were previously described [72] and grown in UL26-expressing HF cells. Wild-type and UL26-HA-expressing HCMVs (Toledo strain) generated in this study were grown in HF cells. To produce UV-inactivated HCMV (UV-HCMV), the virus stock was irradiated with UV light three times at 0.72 J/cm2 using a CL-1000 Crosslinker (UVP). Expression plasmids Mammalian expression plasmids for HA-IE1 (pDJK170) and HA-IE2 (pDJK171) were cloned using the pSG5 vector and plasmid for myc-ISG15 (pOK20) was cloned using the pCS3-MT (with a hexa-myc tag) vector using Gateway technology as previously described [59]. Plasmids for myc-ISG15GG (pYJ12), an active form of ISG15 with a termination codon added immediately after the double glycine residues, or myc-ISG15AA (pYJ14), a conjugation-defective mutant in which the double glycine residues are replaced with alanine residues, were produced using the Stratagene QuickChange site-directed mutagenesis protocol. The pSG5-driven plasmids expressing Flag-UbcH8 (pYJ23) and HA-Herc5 (pYJ29) were produced using Gateway technology. Plasmids for SRT-UL26-p21 (pSE124) and Myc-UL26-p21 (pSE107) were cloned using the pcDNA6 (Life Technologies) vector and pCS3-MT vector, respectively, using Gateway technology. For the SRT-UL26-p21 expression plasmid used in ISGylation assays, two lysine residues on the linker between the SRT tag and the UL26 ORF were changed to alanines to block their possible ISGylation, resulting in pYJ178. Site-directed mutagenesis was performed on the pYJ178 background to produce plasmids expressing the lysine to arginine mutant versions of SRT-UL26-p21; K54R (pYJ179), K136R (pYJ180), K136/169R (pYJ182), and K54/136/169R (pYJ183). Plasmids for HA-hUBE1L (pCAGGS-HA-hUBE1L) and Flag-hUbcH8 (pFlagCMV2-UbcH8) and plasmids for S-tagged Herc5 (pCI-neo-S2-Herc5) were kindly provided by Dong-Er Zhang (Moores Cancer Center, University of California, San Diego, La Jolla, CA 92093, USA). Yeast two-hybrid assays Yeast AH109 (MATa) cells were transformed with plasmid expressing the GAL4-DNA-binding (DB)-ISG15 (TRP+) or GAL4-DB-UBE1L (TRP+) fusion protein. Y187 (MATα) cells were transformed with plasmid expressing the GAL4-activation domain (A)-HCMV ORF (LEU2+) fusion proteins. Each transformant was selected on plates lacking tryptophan (SC-Trp) or leucine (SC-Leu). Trp+ and Leu+ transformants were mated with each other on YPD plates. Diploid cells (a/α) were selected on plates lacking both tryptophan and leucine (SC-TrpLeu). Trp+Leu+ colonies were tested for their growth on plates that lack tryptophan, leucine and histidine (SC-TrpLeuHis). Cells expressing bait and prey that interact with each other grow on SC-TrpLeuHis. Cells expressing both GAL4- DB-ISG15 and GAL4-A-UBE1L were used as a positive control, whereas cells expressing GAL4-DB-ISG15 and GAL4-A only were used as a negative control. Production of the recombinant virus expressing UL26-HA proteins The Toledo-BAC clone encoding UL26 proteins with a C-terminal HA tag was produced by using a counter-selection BAC modification kit (Gene Bridges). The scheme for bacmid mutagenesis is described in S5 Fig and the LMV primers used for mutagenesis are listed in S1 Table. Retroviral vectors Retroviral vectors expressing IE1 (pYH38) or UL26 (pYJ104), UL26(K136/169R) (pYJ176), and UL26(K54/136/169R) (pYJ177) were produced on the background of pMIN (murine leukemia virus-based retroviral vector) using Gateway technology as previously described [59]. Retroviral vectors expressing shRNA for UbcH8 (pMSCVpuro-shUbcH8) was previously described [5]. To produce retroviral vectors expressing shRNA for UBE1L (pMSCVpuro-shUBE1L-1), short hairpin RNA (shRNA) for UBE1L was amplified with U6 promoter by PCR with primers 5′- TTTGGATCCCAAGGTCGGGCAGGAAGAGGGCCTATTTCC-3′ and 5′-TTTGAATTCAAAAAGGATGATGACAGCAACTTCTCTCTTGAAGAAGTTGCTGTCATCATCCGGTGTTTCGTCCTTTCCACAAGATATATAA-3′ (target sequence underlined). The PCR product was digested with BamHI and EcoRI and ligated to MSCV-PGKpuro (BD Biosciences Clontech) digested with BglII and EcoRI. Recombinant retroviruses were prepared in 293T cells after co-transfection with retroviral vectors together with the packaging plasmids pHIT60 (Gag-Pol) and pMD-G expressing the envelope G protein of vesicular stomatitis virus (VSV) [59] using Metafectene reagents (Biotex). Viral supernatants were collected at 48 h after transfection. HF cells were transduced by retroviruses in the presence of polybrene (7.5 μg/ml). Cells were selected with G418 (0.5 mg/ml) (Calbiochem) and maintained in a medium containing G418 (0.1 mg/ml). Lentiviral vectors Lentiviral vector pLKO.1-TRC control expressing a non-hairpin control RNA (shC) was purchased from Addgene. pLKO.1-based lentiviral vectors expressing shRNA for UBP43 (shUBP43-1: TRCN0000004194 and shUBP43-2: TRCN0000004195) and Herc5 (shHerc5-1: TRCN0000004171 and shHerc5-2: TRCN0000004169) were purchased from Open Biosystems. To produce lentiviruses, 293T cells were transfected with lentiviral vectors together with plasmids pCMV-DR8.91 expressing the gag-pol, tat, and rev proteins of human immunodeficiency virus (HIV) and pMD-G. At 48 h, the viral supernatants were collected and used to transduce HF cells in the presence of polybrene (7.5 μg/ml). The transduced cells were selected with puromycin (1 μg/ml) and maintained in a medium containing puromycin (0.5 μg/ml). Antibodies Mouse monoclonal antibody (MAb) 810R, which detects epitopes present in both IE1 and IE2, was purchased from Chemicon. Mouse MAbs against UL44 (p52) and UL99 (pp28) were obtained from Virusys. Anti-β-actin and anti-α-tubulin mouse MAbs were purchased from Sigma. Anti-HA rat MAb 3F10 and anti-myc mouse MAb 9E10, conjugated with peroxidase or labeled with fluorescein isothiocyanate (FITC), were purchased from Roche. Anti-ISG15 (F-9) and anti-STAT2 mouse MAbs were obtained from Santa Cruz. Mouse MAb against SRT epitope was previously described [91]. UBP43 antibody was previously described [92]. Rabbit polyclonal Ab (PAb) for STAT2 (C-20) and STAT2 phosphorylated at Tyr689 were purchased from Santa Cruz and Upstate, respectively. Rabbit PAb for ISG15 was kindly provided by Chin Ha Chung (Seoul National University, Seoul, Republic of Korea). Immunoblot analysis and indirect immunofluorescence assay (IFA) For immunoblot analysis, cells were washed with phosphate-buffered saline (PBS) and total cell lysates were prepared by boiling the cell pellets in sodium dodecyl sulfate (SDS) loading buffer. Equal amounts of the clarified cell extracts were separated on a SDS-polyacrylamide gel or Gradi-Gel II (Elpis biotech, Republic of Korea) and electroblotted onto nitrocellulose membranes. The blots were blocked by incubation for 30 min at room temperature with PBS plus 0.1% Tween 20 (PBST) containing 5% nonfat dry milk. After being washed with PBST three times, the blots were incubated with the appropriate antibodies in PBST for 1 h at room temperature. After three 5 min washes with PBST, the blots were incubated with horseradish peroxidase-conjugated goat anti-mouse IgG or anti-rabbit IgG (Amersham) for 1 h at room temperature. The blots were then washed three times with PBST, and the protein bands were visualized with enhanced chemiluminescence system (Amersham). For IFA, cells were fixed in ice-cold methanol for 5 min and rehydrated in cold PBS. Then, the cells were incubated with appropriate primary antibodies in PBS at 37°C for 1 h, followed by incubation with appropriate secondary antibodies at 37°C for 1 h. The mounting solution containing Hoechst and anti-fade reagent (Molecular Probes) was used. For double-labeling, two different antibodies were incubated together. Slides were examined and with a Carl Zeiss LSM710Meta confocal microscope system. Luciferase reporter assay HF cells (2 × 105) were collected and incubated with 200 μl of lysis buffer [40 mM Tris-HCl (pH 7.8), 50 mM NaCl, 2 mM EDTA, 1 mM MgSO4, and 1% Triton X-100 plus 5 mM dithiothreitol] for 20 min on ice. The extracts were clarified in a microcentrifuge and 20 μl of extracts were incubated with 350 μl of reaction buffer A (25 mM Gly-Gly pH 7.8, 15 mM ATP and 4 mM EGTA) and then mixed with 100 μl of 1 mM luciferin (Sigma). A TD-20/20 luminometer (Turner Designs) was used for the 10-s assay of the photons produced (measured in relative light units). Infectious center assay The diluted samples were used to inoculate a monolayer of 4 × 104 HF cells in a 24-well plate. At 24 h post infection, cells were fixed with 500 μl of cold methanol for 10 min. The cells were then washed three times in phosphate-buffered saline (PBS) and incubated with anti-IE1 rabbit polyclonal antibody in PBS at 37°C for 1 h, followed by incubation with phosphatase-conjugated anti-rabbit immunoglobulin G (IgG) antibody in phosphate-buffered saline (PBS) at 37°C for 1 h. Finally, the cells were gently washed in PBS and treated with 200 μl of developing solution (nitroblue tetrazolium/5-bromo-4-chloro-3-indolylphosphate) at room temperature for 1 h. The positively stained cells were counted for at least three to five separate fields per well under a light microscope (× 200 magnification). Co-immunoprecipitation (co-IP) assays Co-transfected 293T cells (8 × 105) or virus-infected HF cells were harvested and sonicated in 0.7 ml co-IP buffer [50 mM Tris-Cl (pH 7.4), 50 mM NaF, 5 mM sodium phosphate, 0.1% Triton X-100, containing protease inhibitors (Sigma)] by a microtip probe (Vibra-Cell; Sonics and Materials, Inc., USA) for 10 s (pulse on: l s, pulse off: 3 s). Cell lysates were incubated with appropriate antibodies. After incubation for 16 h at 4°C, 30 μl of a 50% slurry of protein A- and G-Sepharose (Amersham) were added and then the mixture was incubated for 2 h at 4°C to allow adsorption. The mixture was then pelleted and washed 7 times with co-IP buffer. The beads were resuspended and boiled for 5 min in loading buffer. Each sample was analyzed by SDS-PAGE and immunoblotting with appropriate antibodies. Co-IP assays to detect ISGylated protein For co-transfection/ISGylation assays, 293T cells were co-transfected with plasmids expressing target protein and ISGylation enzymes. Co-transfected or virus-infected cells were treated with 0.5 mM NEM (N-ethylmaleimide) for 30 min before they were harvested. Cell pellets were resuspended with 10% SDS lysis buffer containing protease inhibitors and boiled for 10 min. Cell lysates were diluted 10-fold with co-IP buffer (50 Mm Tris-Cl [pH 7.4], 50 mM NaF, 5 mM sodium pyrophosphate, containing protease inhibitors) and sonicated by using a Microtip probe (Vibra cell; Sonics and Materials, Inc.). The clarified cell lysates were incubated with appropriate antibody for 16 h and then with 30 μl of a 50% slurry of protein G for 2 h. The mixture was pelleted and washed seven times with co-IP buffer. The bound proteins were boiled and analyzed by SDS-PAGE followed by immunoblot assays Reverse transcription-PCR (RT-PCR) and quantitative real-time RT-PCR (qRT-PCR) Total RNAs were isolated from 2 × 105 cells using TRIzol reagent (Invitrogen) and MaXtract High Density (Qiagen). First-strand cDNA was synthesized by using the random hexamer primers in the SuperScript III system (Invitrogen). Quantitative real-time TR-PCR (qRT-PCR) was performed using the Applied Biosystems ABI Prism SDS software and the following primers: for ISG15, 5ʹ-GCTGGGACCTGACGGTG-3ʹ (sense) and 5ʹ-TTAGCTCCGCCCGCCAG-3ʹ (anti-sense); for UBE1L, 5ʹ-AGGTGGCCAAGAACTTGGTT-3ʹ (sense) and 5ʹ-CACCACCTGGAAGTCCAACA-3ʹ (anti-sense); for UbcH8, 5ʹ-AACCTGTCCAGCGATGATGC-3ʹ (sense) and 5ʹ-TGGTGCAAGGCTTCCAGTTC-3ʹ (anti-sense); for Herc5, 5ʹ-GGGATGAAAGTGCTGAGGAG-3ʹ (sense) and 5ʹ-CATTTTCTGAAGCGTCCACA-3ʹ (anti-sense); for β-actin, 5ʹ-AGCGGGAAATCGTGCGTG-3ʹ (sense) and 5ʹ-CAGGGTACATGGTGGTGCC-3ʹ (anti-sense). Statistical analysis Statistical significances were determined using the Student’s t-test and are indicated by *P<0.05, **P<0.01, or ***P<0.001. Supporting Information S1 Fig Effects of UBE1L and UbcH8-knockdown on HCMV growth. (A) 293T cells were co-transfected with MSCV retroviral vectors expressing shRNA for UBE1L (shUBE1L-1 or shUBE1L-2) and plasmids encoding HA-UBE1L or HA-AML1b as indicated. At 48 h after transfection, cell lysates were immunoblotted with anti-HA antibody. The results showed that shUBE1L-1 specifically reduced UBE1L expression. (B) HeLa cells transduced by MSCV or shUBE1L-1 (hereafter shUBE1L)-expressing MSCV were treated or not with IFNxα (5,000 U/ml) for 24 h. Cell lysates were immunoblotted with antibodies for ISG15 and tubulin-a. (C) Control HF cells or cells expressing UbcH8-specific shRNA were mock-infected or infected with UV-HCMV. The levels of UbcH8 were determined by immunoblotting with anti-UbcH8 rabbit polyclonal antibodies. (D) Control HF cells or cells expressing UBE1L- or UbcH8-specific shRNA were mock-infected or infected with UV-HCMV. The levels of UBE1L and UbcH8 transcripts were determined at 24 h after infection by qRT-PCR. The β-actin transcript levels were used for normalization. Values are the averages of duplicated assays; error ranges are indicated. (E) HF-shRNA cells were infected or not with UV-HCMV at an MOI of 3. At 24 h after infection, immunoblotting was performed with antibodies for ISG15 and β-actin. (F) HF-shRNA cells were infected with the recombinant virus containing the GFP expression cassette (HCMV-GFP) at an MOI of 0.1. GFP images of cells were taken at 7 days after infection. (G) HF-shRNA cells were infected with HCMV at an MOI of 0.1. At 9 days after infection, the viral supernatants were collected and the levels of progeny virions were measured by infectious center assays. Statistical significances were determined using the Student’s t-test and are indicated by **P<0.01. (TIF) Click here for additional data file. S2 Fig Effect of ectopic expression of ISG15 and ISGylation enzymes on HCMV growth. HF cells (2 × 105) were transfected by electroporation with plasmids expressing myc-ISG15GG (or myc-ISG15AA), HA-UBE1L, flag-UbcH8, and HA-Herc5 in combinations as indicated. At 24 h after electroporation, cells were infected with HCMV at an MOI of 1. At 6 days after infection, viral titers in the culture supernatants were measured by infectious center assays (top). The results were shown as averages in two experiments. The levels of exogenously expressed myc-ISG15 proteins, myc-ISG15 conjugates, and β-actin were shown by immunoblotting (bottom). (TIF) Click here for additional data file. S3 Fig Effect of ISGylation on the activation of HCMV promoters. (A-B) HF cells were co-transfected with reporter plasmids containing MIE-Luc, UL112-113-Luc, Pol-Luc, or pp28-Luc reporter gene and effector plasmids as indicated. At 48 after transfection, cell lysates were prepared and assayed for the luciferase activity. The results shown are the mean values for the three independent experiments with standard errors. The expression levels of myc-ISG15 and myc-ISG15 conjugates, HA-IE2, and β-actin in transfected cells were shown by immunoblotting. Statistical significances were determined using the Student’s t-test and are indicated by *P<0.05 or **P<0.01. (TIF) Click here for additional data file. S4 Fig Interactions of HCMV proteins with ISG15 and UBE1L in co-IP assays. (A) 293T cells were co-transfected with plasmids encoding GFP-ISG15AA or myc-ORFs, as indicated. At 48 h after transfection, cell lysates were prepared and immunoprecipitated with anti-myc antibody, followed by immunoblotting with anti-GFP antibody. To determine the expression levels of each protein, whole cell lysate were also immunoblotted. (B) 293T cells were co-transfected with plasmids encoding HA-UBE1L or myc-ORFs, as indicated. At 48 h after transfection, cell lysates were prepared and immunoprecipitated with anti-myc antibody, followed by immunoblotting with anti-HA antibody. (TIF) Click here for additional data file. S5 Fig Production of the recombinant HCMV expressing UL26 proteins with a C-terminal HA tag. (A) The scheme for Toledo-bacmid mutagenesis. The HCMV (Toledo) bacterial artificial chromosome (BAC) clone (Toledo-BAC)[93] was a gift from Hua Zhu (UMDNJ-New Jersey Medical School, Newark, NJ, USA). The rpsL-neo cassettes were PCR-amplified using LMV1766/1767 primers containing homology arms consisting of 50 nucleotides upstream and downstream of the target region plus 24 nucleotides homologous to the rpsL-neo cassette. The amplified rpsL-neo fragments with homology arms were purified and introduced into E. coli GS243 containing wild-type Toledo-BAC for recombination by electroporation using a Gene Pulser II (Bio-Rad). The intermediate Toledo-BAC constructs containing the rpsL-neo cassette were selected on Luria Broth (LB) plates containing kanamycin. Next, the rpsL-neo cassette was replaced by annealed oligo DNAs (LMV1768/1769) consisting of only homology arms (50 nucleotides upstream and downstream of the target region). The ΔUL26 Toledo-BAC was selected on LB plates containing streptomycin. The mutated regions were amplified by PCR and sequenced to verify the desired mutation. The Toledo-BAC encoding UL26-HA was generated from the mutant Toledo-BAC. First, the rps-neo cassettes flanked by homology arms were inserted again into the mutant Toledo-BAC. Next, DNA fragments containing the wild-type UL26 gene with a HA tag at its C-terminus were PCR amplified by 2-steps using LMV1805/1812 and LMV1805/1772. The amplified UL26-HA gene was then inserted into the Toledo-BAC containing the rps-neo cassette by homologous recombination. The LMV primers used for mutagenesis are listed in S1 Table. (B) The regions containing the UL26 ORF from Wt, ΔUL26, and UL26-HA bacmid DNAs were PCR amplified with LMV1764/1765. (C) Wt, ΔUL26, and UL26-HA bacmid DNAs were digested with BglII and the digestion patterns were compared via agarose gel electrophoresis. The bands corresponding to 5,226 and 5,253 bp from wild-type and UL26-HA bacmids, respectively, and a band of 4,660 bp from ΔUL26 bacmid were indicated as arrowheads. (TIF) Click here for additional data file. S6 Fig Specific binding of pUL26 with ISG15, UBE1L, and Hec5 in co-IP assays and broad ISGylation of proteins in cotransfection/ISGylation assays. (A-C) 293T cells were co-transfected with plasmids encoding SRT-ISG15, HA-UBE1L, HA-Herc5, or myc-ORFs, as indicated. At 48 h after transfection, cell lysates were prepared and immunoprecipitated with anti-myc antibody, followed by immunoblotting with anti-SRT antibody (A) or anti-HA antibody (B and C). To determine the expression levels of each protein, whole cell lysate were also immunoblotted. (D) Co-transfection/ISGylation assays. 293T cells were co-transfected with plasmid expressing SRT-tagged ORF (UL26, UL85, and UL71), myc-ISG15 (with GG or AA terminus), HA-UBE1L, Flag-UbcH8, or HA-Herc5 as indicated. At 48 h after transfection, cell lysates were immunoprecipitated with anti-SRT antibody, followed by immunoblotting with anti-myc antibody. Whole cell lysates were immunoblotted with anti-SRT antibody to determine the expression levels of each protein. (TIF) Click here for additional data file. S7 Fig Lack of ISGylation and ISGylation inhibitory effect of IE2. Comparative co-transfection/ISGylation assays for UL26 and IE2 were performed in 293T cells with or without increasing amounts of plasmids expressing SRT-UL26-p21 or SRT-IE2 IE1 as in Fig 2D. Cell lysates were prepared and immunoprecipitated with anti-SRT antibody, followed by immunoblotting with anti-myc antibody. Whole cell lysates were immunoblotted with anti-SRT antibody to determine the expression levels of UL26-p21 and IE2, or with anti-myc antibody to determine the effect of UL26-p21 or IE2 expression on ISGylation. (TIF) Click here for additional data file. S8 Fig Comparison of ISGylation between wild-type and ΔUL26 virus infected cells. HF cells were mock-infected or infected with wild-type or ΔUL26 mutant virus (Ad169) at an MOI of 0.2. Cell lysates were immunoblotted at the indicated time points with antibodies for ISG15, viral proteins (IE1, IE2, and UL26), and β-actin. (TIF) Click here for additional data file. S1 Table PCR primers used for bacmid mutagenesis. (TIF) Click here for additional data file. S2 Table Summary of the HCMV proteins that interacted with ISG15 and UBE1L in yeast two-hybrid assays and co-IP assays. (TIF) Click here for additional data file. We thank Dong-Er Zhang and Chin Ha Chung for providing plasmids and antibodies. We also thank Gary S. Hayward for sharing the HCMV ORF library, Edward S. Mocarski for providing the CR208 virus, and Hua Zhu for providing the HCMV (Toledo) bacmid. ==== Refs References 1 Sadler AJ , Williams BR . Interferon-inducible antiviral effectors . Nat Rev Immunol . 2008 ;8 (7 ):559 –68 . 10.1038/nri2314 18575461 2 Jeon YJ , Yoo HM , Chung CH . ISG15 and immune diseases . Biochim Biophys Acta . 2010 ;1802 (5 ):485 –96 . 10.1016/j.bbadis.2010.02.006 20153823 3 Yuan W , Krug RM . Influenza B virus NS1 protein inhibits conjugation of the interferon (IFN)-induced ubiquitin-like ISG15 protein . Embo J . 2001 ;20 (3 ):362 –71 . 11157743 4 Zhao C , Beaudenon SL , Kelley ML , Waddell MB , Yuan W , Schulman BA , et al The UbcH8 ubiquitin E2 enzyme is also the E2 enzyme for ISG15, an IFN-alpha/beta-induced ubiquitin-like protein . Proc Natl Acad Sci U S A . 2004 ;101 (20 ):7578 –82 . 15131269 5 Kim KI , Giannakopoulos NV , Virgin HW , Zhang DE . Interferon-inducible ubiquitin E2, Ubc8, is a conjugating enzyme for protein ISGylation . Mol Cell Biol . 2004 ;24 (21 ):9592 –600 . 15485925 6 Dastur A , Beaudenon S , Kelley M , Krug RM , Huibregtse JM . Herc5, an interferon-induced HECT E3 enzyme, is required for conjugation of ISG15 in human cells . J Biol Chem . 2006 ;281 (7 ):4334 –8 . 16407192 7 Wong JJ , Pung YF , Sze NS , Chin KC . HERC5 is an IFN-induced HECT-type E3 protein ligase that mediates type I IFN-induced ISGylation of protein targets . Proc Natl Acad Sci U S A . 2006 ;103 (28 ):10735 –40 . 16815975 8 Zou W , Zhang DE . The interferon-inducible ubiquitin-protein isopeptide ligase (E3) EFP also functions as an ISG15 E3 ligase . J Biol Chem . 2006 ;281 (7 ):3989 –94 . 16352599 9 Okumura F , Zou W , Zhang DE . ISG15 modification of the eIF4E cognate 4EHP enhances cap structure-binding activity of 4EHP . Genes Dev . 2007 ;21 (3 ):255 –60 . 17289916 10 Haas AL , Ahrens P , Bright PM , Ankel H . Interferon induces a 15-kilodalton protein exhibiting marked homology to ubiquitin . J Biol Chem . 1987 ;262 (23 ):11315 –23 . 2440890 11 Nyman TA , Matikainen S , Sareneva T , Julkunen I , Kalkkinen N . Proteome analysis reveals ubiquitin-conjugating enzymes to be a new family of interferon-alpha-regulated genes . Eur J Biochem . 2000 ;267 (13 ):4011 –9 . 10866800 12 Malakhova O , Malakhov M , Hetherington C , Zhang DE . Lipopolysaccharide activates the expression of ISG15-specific protease UBP43 via interferon regulatory factor 3 . J Biol Chem . 2002 ;277 (17 ):14703 –11 . 11854279 13 Li XL , Blackford JA , Judge CS , Liu M , Xiao W , Kalvakolanu DV , et al RNase-L-dependent destabilization of interferon-induced mRNAs. A role for the 2-5A system in attenuation of the interferon response . J Biol Chem . 2000 ;275 (12 ):8880 –8 . 10722734 14 Kang D , Jiang H , Wu Q , Pestka S , Fisher PB . Cloning and characterization of human ubiquitin-processing protease-43 from terminally differentiated human melanoma cells using a rapid subtraction hybridization protocol RaSH . Gene . 2001 ;267 (2 ):233 –42 . 11313150 15 Malakhova OA , Kim KI , Luo JK , Zou W , Kumar KG , Fuchs SY , et al UBP43 is a novel regulator of interferon signaling independent of its ISG15 isopeptidase activity . EMBO J . 2006 ;25 (11 ):2358 –67 . 16710296 16 Shi HX , Yang K , Liu X , Liu XY , Wei B , Shan YF , et al Positive regulation of interferon regulatory factor 3 activation by Herc5 via ISG15 modification . Mol Cell Biol . 2010 ;30 (10 ):2424 –36 . 10.1128/MCB.01466-09 20308324 17 Arimoto K , Konishi H , Shimotohno K . UbcH8 regulates ubiquitin and ISG15 conjugation to RIG-I . Mol Immunol . 2008 ;45 (4 ):1078 –84 . 17719635 18 Giannakopoulos NV , Luo JK , Papov V , Zou W , Lenschow DJ , Jacobs BS , et al Proteomic identification of proteins conjugated to ISG15 in mouse and human cells . Biochem Biophys Res Commun . 2005 ;336 (2 ):496 –506 . 16139798 19 Malakhov MP , Kim KI , Malakhova OA , Jacobs BS , Borden EC , Zhang DE . High-throughput immunoblotting. Ubiquitiin-like protein ISG15 modifies key regulators of signal transduction . J Biol Chem . 2003 ;278 (19 ):16608 –13 . 12582176 20 Zhao C , Denison C , Huibregtse JM , Gygi S , Krug RM . Human ISG15 conjugation targets both IFN-induced and constitutively expressed proteins functioning in diverse cellular pathways . Proc Natl Acad Sci U S A . 2005 ;102 (29 ):10200 –5 . 16009940 21 Lu G , Reinert JT , Pitha-Rowe I , Okumura A , Kellum M , Knobeloch KP , et al ISG15 enhances the innate antiviral response by inhibition of IRF-3 degradation . Cell Mol Biol (Noisy-le-grand) . 2006 ;52 (1 ):29 –41 .16914094 22 Hsiang TY , Zhao C , Krug RM . Interferon-induced ISG15 conjugation inhibits influenza A virus gene expression and replication in human cells . J Virol . 2009 ;83 (12 ):5971 –7 . 10.1128/JVI.01667-08 19357168 23 Lenschow DJ , Lai C , Frias-Staheli N , Giannakopoulos NV , Lutz A , Wolff T , et al IFN-stimulated gene 15 functions as a critical antiviral molecule against influenza, herpes, and Sindbis viruses . Proc Natl Acad Sci U S A . 2007 ;104 (4 ):1371 –6 . 17227866 24 Tang Y , Zhong G , Zhu L , Liu X , Shan Y , Feng H , et al Herc5 Attenuates Influenza A Virus by Catalyzing ISGylation of Viral NS1 Protein . J Immunol . 2010 ;184 (10 ):5777 –90 . 10.4049/jimmunol.0903588 20385878 25 Lai C , Struckhoff JJ , Schneider J , Martinez-Sobrido L , Wolff T , Garcia-Sastre A , et al Mice lacking the ISG15 E1 enzyme UbE1L demonstrate increased susceptibility to both mouse-adapted and non-mouse-adapted influenza B virus infection . J Virol . 2009 ;83 (2 ):1147 –51 . 10.1128/JVI.00105-08 19004958 26 Okumura A , Lu G , Pitha-Rowe I , Pitha PM . Innate antiviral response targets HIV-1 release by the induction of ubiquitin-like protein ISG15 . Proc Natl Acad Sci U S A . 2006 ;103 (5 ):1440 –5 . 16434471 27 Kunzi MS , Pitha PM . Role of interferon-stimulated gene ISG-15 in the interferon-omega-mediated inhibition of human immunodeficiency virus replication . J Interferon Cytokine Res . 1996 ;16 (11 ):919 –27 . 8938567 28 Chua PK , McCown MF , Rajyaguru S , Kular S , Varma R , Symons J , et al Modulation of alpha interferon anti-hepatitis C virus activity by ISG15 . J Gen Virol . 2009 ;90 (Pt 12 ):2929 –39 . 10.1099/vir.0.013128-0 19656964 29 Kim MJ , Yoo JY . Inhibition of hepatitis C virus replication by IFN-mediated ISGylation of HCV-NS5A . J Immunol . 2010 ;185 (7 ):4311 –8 . 10.4049/jimmunol.1000098 20810994 30 Domingues P , Bamford CG , Boutell C , McLauchlan J . Inhibition of Hepatitis C Virus RNA Replication by ISG15 does not Require Its Conjugation to Protein Substrates by the HERC5 E3 Ligase . J Gen Virol . 2015 . 31 Hsiao NW , Chen JW , Yang TC , Orloff GM , Wu YY , Lai CH , et al ISG15 over-expression inhibits replication of the Japanese encephalitis virus in human medulloblastoma cells . Antiviral Res . 2010 ;85 (3 ):504 –11 . 10.1016/j.antiviral.2009.12.007 20035788 32 Giannakopoulos NV , Arutyunova E , Lai C , Lenschow DJ , Haas AL , Virgin HW . ISG15 Arg151 and the ISG15-conjugating enzyme UbE1L are important for innate immune control of Sindbis virus . J Virol . 2009 ;83 (4 ):1602 –10 . 10.1128/JVI.01590-08 19073728 33 Lenschow DJ , Giannakopoulos NV , Gunn LJ , Johnston C , O'Guin AK , Schmidt RE , et al Identification of interferon-stimulated gene 15 as an antiviral molecule during Sindbis virus infection in vivo . J Virol . 2005 ;79 (22 ):13974 –83 . 16254333 34 Malakhova OA , Zhang DE . ISG15 inhibits Nedd4 ubiquitin E3 activity and enhances the innate antiviral response . J Biol Chem . 2008 ;283 (14 ):8783 –7 . 10.1074/jbc.C800030200 18287095 35 Okumura A , Pitha PM , Harty RN . ISG15 inhibits Ebola VP40 VLP budding in an L-domain-dependent manner by blocking Nedd4 ligase activity . Proc Natl Acad Sci U S A . 2008 ;105 (10 ):3974 –9 . 10.1073/pnas.0710629105 18305167 36 Guerra S , Caceres A , Knobeloch KP , Horak I , Esteban M . Vaccinia virus E3 protein prevents the antiviral action of ISG15 . PLoS Pathog . 2008 ;4 (7 ):e1000096 10.1371/journal.ppat.1000096 18604270 37 Dai J , Pan W , Wang P . ISG15 facilitates cellular antiviral response to dengue and west nile virus infection in vitro . Virol J . 2011 ;8 :468 10.1186/1743-422X-8-468 21992229 38 Sun Z , Li Y , Ransburgh R , Snijder EJ , Fang Y . Nonstructural protein 2 of porcine reproductive and respiratory syndrome virus inhibits the antiviral function of interferon-stimulated gene 15 . J Virol . 2012 ;86 (7 ):3839 –50 . 10.1128/JVI.06466-11 22258253 39 Jacobs SR , Stopford CM , West JA , Bennett CL , Giffin L , Damania B . Kaposi's Sarcoma-Associated Herpesvirus Viral Interferon Regulatory Factor 1 Interacts with a Member of the Interferon-Stimulated Gene 15 Pathway . J Virol . 2015 ;89 (22 ):11572 –83 . 10.1128/JVI.01482-15 26355087 40 Gonzalez-Sanz R , Mata M , Bermejo-Martin J , Alvarez A , Cortijo J , Melero JA , et al ISG15 is upregulated in respiratory syncytial virus infection and reduces virus growth through protein ISGylation . J Virol . 2016 ;90 (7 ):3428 –38 . 10.1128/JVI.02695-15 26763998 41 Durfee LA , Lyon N , Seo K , Huibregtse JM . The ISG15 conjugation system broadly targets newly synthesized proteins: implications for the antiviral function of ISG15 . Mol Cell . 2010 ;38 (5 ):722 –32 . 10.1016/j.molcel.2010.05.002 20542004 42 Zhao C , Hsiang TY , Kuo RL , Krug RM . ISG15 conjugation system targets the viral NS1 protein in influenza A virus-infected cells . Proc Natl Acad Sci U S A . 2010 ;107 (5 ):2253 –8 . 10.1073/pnas.0909144107 20133869 43 Seo EJ , Leis J . Budding of Enveloped Viruses: Interferon-Induced ISG15-Antivirus Mechanisms Targeting the Release Process . Adv Virol . 2012 ;2012 :532723 10.1155/2012/532723 22666250 44 Werneke SW , Schilte C , Rohatgi A , Monte KJ , Michault A , Arenzana-Seisdedos F , et al ISG15 is critical in the control of Chikungunya virus infection independent of UbE1L mediated conjugation . PLoS Pathog . 2011 ;7 (10 ):e1002322 10.1371/journal.ppat.1002322 22028657 45 Kim MJ , Hwang SY , Imaizumi T , Yoo JY . Negative feedback regulation of RIG-I-mediated antiviral signaling by interferon-induced ISG15 conjugation . J Virol . 2008 ;82 (3 ):1474 –83 . 18057259 46 Chen L , Sun J , Meng L , Heathcote J , Edwards AM , McGilvray ID . ISG15, a ubiquitin-like interferon-stimulated gene, promotes hepatitis C virus production in vitro: implications for chronic infection and response to treatment . J Gen Virol . 2010 ;91 (Pt 2 ):382 –8 . 10.1099/vir.0.015388-0 19846672 47 Broering R , Zhang X , Kottilil S , Trippler M , Jiang M , Lu M , et al The interferon stimulated gene 15 functions as a proviral factor for the hepatitis C virus and as a regulator of the IFN response . Gut . 2010 ;59 (8 ):1111 –9 . 10.1136/gut.2009.195545 20639253 48 Bogunovic D , Boisson-Dupuis S , Casanova JL . ISG15: leading a double life as a secreted molecule . Exp Mol Med . 2013 ;45 :e18 10.1038/emm.2013.36 23579383 49 Campbell JA , Lenschow DJ . Emerging roles for immunomodulatory functions of free ISG15 . J Interferon Cytokine Res . 2013 ;33 (12 ):728 –38 . 10.1089/jir.2013.0064 24010825 50 Bogunovic D , Byun M , Durfee LA , Abhyankar A , Sanal O , Mansouri D , et al Mycobacterial disease and impaired IFN-gamma immunity in humans with inherited ISG15 deficiency . Science . 2012 ;337 (6102 ):1684 –8 . 22859821 51 Pohl C , Dikic I . Fighting mycobacteria through ISGylation . EMBO Rep . 2012 ;13 (10 ):872 –3 . 10.1038/embor.2012.129 22940737 52 Zhang X , Bogunovic D , Payelle-Brogard B , Francois-Newton V , Speer SD , Yuan C , et al Human intracellular ISG15 prevents interferon-alpha/beta over-amplification and auto-inflammation . Nature . 2015 ;517 (7532 ):89 –93 . 10.1038/nature13801 25307056 53 Speer SD , Li Z , Buta S , Payelle-Brogard B , Qian L , Vigant F , et al ISG15 deficiency and increased viral resistance in humans but not mice . Nature communications . 2016 ;7 :11496 10.1038/ncomms11496 27193971 54 Mocarski ES , Shenk T , Griffiths PD , Pass RF . Cytomegaloviruses, p. 1960 –2014 . In Knipe D. M. , Howley P. M. , Cohen J. I. , Griffin D. E. , Lamb R. A. , Martin M. A. , Racaniello V. R. , and Roizman B. (ed.), Fields virology . Philadelphia, PA : Lippincott Williams & Wilkins ; 2013 . 55 Browne EP , Shenk T . Human cytomegalovirus UL83-coded pp65 virion protein inhibits antiviral gene expression in infected cells . Proc Natl Acad Sci U S A . 2003 ;100 (20 ):11439 –44 . 12972646 56 Taylor RT , Bresnahan WA . Human cytomegalovirus immediate-early 2 gene expression blocks virus-induced beta interferon production . J Virol . 2005 ;79 (6 ):3873 –7 . 15731283 57 Abate DA , Watanabe S , Mocarski ES . Major human cytomegalovirus structural protein pp65 (ppUL83) prevents interferon response factor 3 activation in the interferon response . J Virol . 2004 ;78 (20 ):10995 –1006 . 15452220 58 Paulus C , Krauss S , Nevels M . A human cytomegalovirus antagonist of type I IFN-dependent signal transducer and activator of transcription signaling . Proc Natl Acad Sci U S A . 2006 ;103 (10 ):3840 –5 . 16497831 59 Huh YH , Kim YE , Kim ET , Park JJ , Song MJ , Zhu H , et al Binding STAT2 by the acidic domain of human cytomegalovirus IE1 promotes viral growth and is negatively regulated by SUMO . J Virol . 2008 ;82 (21 ):10444 –54 . 10.1128/JVI.00833-08 18701593 60 Krauss S , Kaps J , Czech N , Paulus C , Nevels M . Physical requirements and functional consequences of complex formation between the cytomegalovirus IE1 protein and human STAT2 . J Virol . 2009 ;83 (24 ):12854 –70 . 10.1128/JVI.01164-09 19812155 61 Kim YE , Ahn JH . Positive role of promyelocytic leukemia protein in type I interferon response and its regulation by human cytomegalovirus . PLoS Pathog . 2015 ;11 (3 ):e1004785 10.1371/journal.ppat.1004785 25812002 62 Scherer M , Otto V , Stump JD , Klingl S , Muller R , Reuter N , et al Characterization of Recombinant Human Cytomegaloviruses Encoding IE1 Mutants L174P and 1–382 Reveals that Viral Targeting of PML Bodies Perturbs both Intrinsic and Innate Immune Responses . J Virol . 2015 ;90 (3 ):1190 –205 . 10.1128/JVI.01973-15 26559840 63 Miller DM , Rahill BM , Boss JM , Lairmore MD , Durbin JE , Waldman JW , et al Human cytomegalovirus inhibits major histocompatibility complex class II expression by disruption of the Jak/Stat pathway . J Exp Med . 1998 ;187 (5 ):675 –83 . 9480977 64 Miller DM , Zhang Y , Rahill BM , Waldman WJ , Sedmak DD . Human cytomegalovirus inhibits IFN-alpha-stimulated antiviral and immunoregulatory responses by blocking multiple levels of IFN-alpha signal transduction . J Immunol . 1999 ;162 (10 ):6107 –13 . 10229853 65 Le VT , Trilling M , Wilborn M , Hengel H , Zimmermann A . Human cytomegalovirus interferes with signal transducer and activator of transcription (STAT) 2 protein stability and tyrosine phosphorylation . J Gen Virol . 2008 ;89 (Pt 10 ):2416 –26 . 10.1099/vir.0.2008/001669-0 18796709 66 Knoblach T , Grandel B , Seiler J , Nevels M , Paulus C . Human cytomegalovirus IE1 protein elicits a type II interferon-like host cell response that depends on activated STAT1 but not interferon-gamma . PLoS Pathog . 2011 ;7 (4 ):e1002016 10.1371/journal.ppat.1002016 21533215 67 Greaves RF , Mocarski ES . Defective growth correlates with reduced accumulation of a viral DNA replication protein after low-multiplicity infection by a human cytomegalovirus ie1 mutant . J Virol . 1998 ;72 (1 ):366 –79 . 9420235 68 Loeb KR , Haas AL . The interferon-inducible 15-kDa ubiquitin homolog conjugates to intracellular proteins . J Biol Chem . 1992 ;267 (11 ):7806 –13 . 1373138 69 Malakhova OA , Yan M , Malakhov MP , Yuan Y , Ritchie KJ , Kim KI , et al Protein ISGylation modulates the JAK-STAT signaling pathway . Genes Dev . 2003 ;17 (4 ):455 –60 . 12600939 70 Kim ET , Kim YE , Kim YJ , Lee MK , Hayward GS , Ahn JH . Analysis of Human Cytomegalovirus-Encoded SUMO Targets and Temporal Regulation of SUMOylation of the Immediate-Early Proteins IE1 and IE2 during Infection . PLoS One . 2014 ;9 (7 ):e103308 10.1371/journal.pone.0103308 25050850 71 Stamminger T , Gstaiger M , Weinzierl K , Lorz K , Winkler M , Schaffner W . Open reading frame UL26 of human cytomegalovirus encodes a novel tegument protein that contains a strong transcriptional activation domain . J Virol . 2002 ;76 (10 ):4836 –47 . 11967300 72 Lorz K , Hofmann H , Berndt A , Tavalai N , Mueller R , Schlotzer-Schrehardt U , et al Deletion of open reading frame UL26 from the human cytomegalovirus genome results in reduced viral growth, which involves impaired stability of viral particles . J Virol . 2006 ;80 (11 ):5423 –34 . 16699023 73 Mathers C , Spencer CM , Munger J . Distinct domains within the human cytomegalovirus U(L)26 protein are important for wildtype viral replication and virion stability . PLoS One . 2014 ;9 (2 ):e88101 10.1371/journal.pone.0088101 24505393 74 Mathers C , Schafer X , Martinez-Sobrido L , Munger J . The human cytomegalovirus UL26 protein antagonizes NF-kappaB activation . J Virol . 2014 ;88 (24 ):14289 –300 . 10.1128/JVI.02552-14 25275128 75 Fan JB , Arimoto K , Motamedchaboki K , Yan M , Wolf DA , Zhang DE . Identification and characterization of a novel ISG15-ubiquitin mixed chain and its role in regulating protein homeostasis . Scientific reports . 2015 ;5 :12704 10.1038/srep12704 26226047 76 Zou W , Wang J , Zhang DE . Negative regulation of ISG15 E3 ligase EFP through its autoISGylation . Biochem Biophys Res Commun . 2007 ;354 (1 ):321 –7 . 17222803 77 Kim ET , Kim YE , Huh YH , Ahn JH . Role of noncovalent SUMO binding by the human cytomegalovirus IE2 transactivator in lytic growth . J Virol . 2010 ;84 (16 ):8111 –23 . 10.1128/JVI.00459-10 20519406 78 Ferreira HC , Luke B , Schober H , Kalck V , Lingner J , Gasser SM . The PIAS homologue Siz2 regulates perinuclear telomere position and telomerase activity in budding yeast . Nat Cell Biol . 2011 ;13 (7 ):867 –74 . 10.1038/ncb2263 21666682 79 Hang LE , Liu X , Cheung I , Yang Y , Zhao X . SUMOylation regulates telomere length homeostasis by targeting Cdc13 . Nature structural & molecular biology . 2011 ;18 (8 ):920 –6 . 80 Wu CS , Ouyang J , Mori E , Nguyen HD , Marechal A , Hallet A , et al SUMOylation of ATRIP potentiates DNA damage signaling by boosting multiple protein interactions in the ATR pathway . Genes & development . 2014 ;28 (13 ):1472 –84 .24990965 81 Garg M , Gurung RL , Mansoubi S , Ahmed JO , Dave A , Watts FZ , et al Tpz1TPP1 SUMOylation reveals evolutionary conservation of SUMO-dependent Stn1 telomere association . EMBO Rep . 2014 ;15 (8 ):871 –7 . 10.15252/embr.201438919 24925530 82 Chang YG , Yan XZ , Xie YY , Gao XC , Song AX , Zhang DE , et al Different roles for two ubiquitin-like domains of ISG15 in protein modification . J Biol Chem . 2008 ;283 (19 ):13370 –7 . 10.1074/jbc.M800162200 18356159 83 Tandon R , AuCoin DP , Mocarski ES . Human cytomegalovirus exploits ESCRT machinery in the process of virion maturation . J Virol . 2009 ;83 (20 ):10797 –807 . 10.1128/JVI.01093-09 19640981 84 Tandon R , Mocarski ES . Viral and host control of cytomegalovirus maturation . Trends Microbiol . 2012 ;20 (8 ):392 –401 . 10.1016/j.tim.2012.04.008 22633075 85 Calistri A , Sette P , Salata C , Cancellotti E , Forghieri C , Comin A , et al Intracellular trafficking and maturation of herpes simplex virus type 1 gB and virus egress require functional biogenesis of multivesicular bodies . J Virol . 2007 ;81 (20 ):11468 –78 . 17686835 86 Crump CM , Yates C , Minson T . Herpes simplex virus type 1 cytoplasmic envelopment requires functional Vps4 . J Virol . 2007 ;81 (14 ):7380 –7 . 17507493 87 Frias-Staheli N , Giannakopoulos NV , Kikkert M , Taylor SL , Bridgen A , Paragas J , et al Ovarian tumor domain-containing viral proteases evade ubiquitin- and ISG15-dependent innate immune responses . Cell Host Microbe . 2007 ;2 (6 ):404 –16 . 18078692 88 Lindner HA , Lytvyn V , Qi H , Lachance P , Ziomek E , Menard R . Selectivity in ISG15 and ubiquitin recognition by the SARS coronavirus papain-like protease . Arch Biochem Biophys . 2007 ;466 (1 ):8 –14 . 17692280 89 Munger J , Yu D , Shenk T . UL26-deficient human cytomegalovirus produces virions with hypophosphorylated pp28 tegument protein that is unstable within newly infected cells . J Virol . 2006 ;80 (7 ):3541 –8 . 16537622 90 Kim ET , Oh SE , Lee YO , Gibson W , Ahn JH . Cleavage specificity of the UL48 deubiquitinating protease activity of human cytomegalovirus and the growth of an active-site mutant virus in cultured cells . J Virol . 2009 ;83 (23 ):12046 –56 . 10.1128/JVI.00411-09 19759126 91 Lee JM , Kang HJ , Lee HR , Choi CY , Jang WJ , Ahn JH . PIAS1 enhances SUMO-1 modification and the transactivation activity of the major immediate-early IE2 protein of human cytomegalovirus . FEBS Lett . 2003 ;555 (2 ):322 –8 . 14644436 92 Malakhov MP , Malakhova OA , Kim KI , Ritchie KJ , Zhang DE . UBP43 (USP18) specifically removes ISG15 from conjugated proteins . J Biol Chem . 2002 ;277 (12 ):9976 –81 . 11788588 93 Murphy E , Yu D , Grimwood J , Schmutz J , Dickson M , Jarvis MA , et al Coding potential of laboratory and clinical strains of human cytomegalovirus . Proc Natl Acad Sci U S A 2003 ;100 (25 ):14976 –81 . 14657367

PMC005xxxxxx/PMC5001723.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756498810.1371/journal.pone.0161733PONE-D-16-20344Research ArticleBiology and Life SciencesMolecular BiologyMolecular Biology TechniquesArtificial Gene Amplification and ExtensionPolymerase Chain ReactionResearch and Analysis MethodsMolecular Biology TechniquesArtificial Gene Amplification and ExtensionPolymerase Chain ReactionBiology and Life SciencesOrganismsPlantsAlgaeChlamydomonas ReinhardtiiResearch and Analysis MethodsModel OrganismsPlant and Algal ModelsChlamydomonas ReinhardtiiBiology and life sciencesMolecular biologyMolecular biology techniquesSequencing techniquesSequence analysisDNA sequence analysisResearch and analysis methodsMolecular biology techniquesSequencing techniquesSequence analysisDNA sequence analysisBiology and life sciencesMolecular biologyMolecular biology techniquesDNA constructionPlasmid ConstructionResearch and analysis methodsMolecular biology techniquesDNA constructionPlasmid ConstructionBiology and Life SciencesBiochemistryProteinsRecombinant ProteinsBiology and Life SciencesMolecular BiologyMolecular Biology TechniquesSequencing TechniquesSequence AnalysisResearch and Analysis MethodsMolecular Biology TechniquesSequencing TechniquesSequence AnalysisBiology and life sciencesGeneticsDNADNA recombinationTransformation Associated RecombinationBiology and life sciencesBiochemistryNucleic acidsDNADNA recombinationTransformation Associated RecombinationBiology and Life SciencesMolecular BiologyMolecular Biology TechniquesMarker GenesResearch and Analysis MethodsMolecular Biology TechniquesMarker GenesConstruction of Marker-Free Transgenic Strains of Chlamydomonas reinhardtii Using a Cre/loxP-Mediated Recombinase System Cre/loxP-Mediated Gene Excision System in Chlamydomonas reinhardtiiKasai Yuki *Harayama Shigeaki Department of Biological Sciences, Faculty of Science and Engineering, Chuo University, Bunkyo-ku, Tokyo, JapanCooney Austin John EditorUniversity of Texas at Austin Dell Medical School, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: SH. Data curation: YK SH. Funding acquisition: SH. Investigation: YK. Methodology: YK. Project administration: SH. Resources: SH. Software: YK. Supervision: SH. Validation: YK. Writing – original draft: YK. Writing – review & editing: SH. * E-mail: ykasai@kc.chuo-u.ac.jp26 8 2016 2016 11 8 e016173320 5 2016 10 8 2016 © 2016 Kasai, Harayama2016Kasai, HarayamaThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.The Escherichia coli bacteriophage P1 encodes a site-specific recombinase called Cre and two 34-bp target sites of Cre recombinase called loxP. The Cre/loxP system has been used to achieve targeted insertion and precise deletion in many animal and plant genomes. The Cre/loxP system has particularly been used for the removal of selectable marker genes to create marker-free transgenic organisms. For the first time, we applied the Cre/loxP-mediated site-specific recombination system to Chlamydomonas reinhardtii to construct marker-free transgenic strains. Specifically, C. reinhardtii strains cc4350 and cc124 carrying an aphVIII expression cassette flanked by two direct repeats of loxP were constructed. Separately, a synthetic Cre recombinase gene (CrCRE), the codons of which were optimized for expression in C. reinhardtii, was synthesized, and a CrCRE expression cassette was introduced into strain cc4350 carrying a single copy of the loxP-flanked aphVIII expression cassette. Among 46 transformants carrying the CrCRE expression cassette stably, the excision of aphVIII by CrCre recombinase was observed only in one transformant. We then constructed an expression cassette of an in-frame fusion of ble to CrCRE via a short linker peptide. The product of ble (Ble) is a bleomycin-binding protein that confers resistance to bleomycin-related antibiotics such as Zeocin and localizes in the nucleus. Therefore, the ble-(linker)-CrCRE fusion protein is expected to localize in the nucleus. When the ble-(linker)-CrCRE expression cassette was integrated into the genome of strain cc4350 carrying a single copy of the loxP-flanked aphVIII expression cassette, CrCre recombinase-mediated excision of the aphVIII expression cassette was observed at a frequency higher than that in stable transformants of the CrCRE expression cassette. Similarly, from strain cc124 carrying a single loxP-flanked aphVIII expression cassette, the aphVIII expression cassette was successfully excised after introduction of the ble-(linker)-CrCRE expression cassette. The ble-(linker)-CrCRE expression cassette remained in the genome after excision of the aphVIII expression cassette, and it was subsequently removed by crossing with the wild-type strain. This precise Cre-mediated deletion method applicable to transgenic C. reinhardtii could further increase the potential of this organism for use in basic and applied research. http://dx.doi.org/10.13039/501100001863New Energy and Industrial Technology Development OrganizationP11502725-0Harayama Shigeaki This work was supported by the New Energy and Industrial Technology Development Organization (NEDO, P11502725-0)(http://www.nedo.go.jp). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll sequence data files are available from the DDBJ database (accession number(s) LC150883, LC150884, LC150885).Data Availability All sequence data files are available from the DDBJ database (accession number(s) LC150883, LC150884, LC150885). ==== Body Introduction The green unicellular alga Chlamydomonas reinhardtii has been widely used as a model system for studying the genetic and molecular mechanisms of biological processes such as photosynthesis and flagellar motility [1, 2, 3]. Recently, this alga has also been used to manipulate metabolic pathways involved in biofuel and hydrogen production using the range of genetic manipulation tools available to this organism [4, 5, 6, 7]. However, the number of selectable marker genes used in C. reinhardtii is limited even though availability of multiple selectable markers is necessary for the sequential introduction of transgenes. C. reinhardtii is considered to be a model organism for basic research and an industrial biotechnology host [8]. For the large-scale deployment of transgenic C. reinhardtii for various industrial applications, there are public concerns regarding the spread of marker genes in the environment. Therefore, efficient methods for the removal of marker genes from transgenic C. reinhardtii are highly anticipated. Sexual crossing is a powerful tool for this purpose. However, this technique cannot be used if the linkage between a marker gene and a co-introduced transgene is tight. Such tight linkage between a marker gene and co-introduced transgenes was observed in transgenic rice and soybean generated by biolistic bombardment, in which most of the transgenes were co-integrated together with a marker gene at one or multiple loci [9, 10, 11]. Co-transformation of plants by Agrobacterium tumefaciens-mediated transformation using multiple plasmids also resulted in the integration of multiple T-DNAs at the same locus on plant chromosomes [12, 13, 14]. Thus, although the fate of multiple-plasmid co-transformation in C. reinhardtii was not examined systematically, we presume that multiple plasmids are frequently integrated at the same locus, leading to a tight genetic linkage between marker genes and co-introduced transgenes in transgenic C. reinhardtii. One strategy to increase co-transformation frequency is the use of a marker gene physically linked to a gene of interest [15]. Several vectors systems were developed for this purpose [16, 17], including those enabling sustained expression of transgenes in recipients [18]. In cases in which transgenes were obtained using such vectors, the marker gene and transgene are genetically linked and usually inherited together. The genomic sequence of C. reinhardtii includes numerous functionally uncharacterized genes [19]. Reverse genetics is a robust method for revealing the functions of such genes. Because C. reinhardtii displays an extremely low efficiency of homologous recombination [20, 21, 22], insertional random mutagenesis using selectable markers in C. reinhardtii was identified as a valuable tool for investigating diverse biological functions [23, 24, 25, 26, 27, 28, 29]. Although the removal of selectable markers from insertional mutants without the loss of mutant phenotypes is desired for further genetic manipulation or industrial application, marker rescue from insertional mutants using sexual crossing is not possible. To overcome the limitations of sexual crossing, several strategies have been developed to remove selectable markers from transgenic eukaryotic cells [30], including the use of site-specific DNA excision systems such as Cre/loxP from bacteriophage P1 [31, 32, 33, 34], Flp/frt from Saccharomyces cerevisiae [35, 36], R/RS from Zygosaccharomyces rouxii [37, 38], and Gin/gix from bacteriophage Mu [39]. In the bacteriophage P1 bipartite Cre/loxP-mediated site-specific DNA excision system, Cre recombinase specifically recognizes the loxP sequence of 34 bp in length and excises a DNA segment flanked by two direct repeats of loxP, leaving a single copy of loxP [40, 41, 42]. This system has been proven to be a powerful marker rescue tool in eukaryotes [43, 44, 45]. Curiously, no research on the use of Cre/loxP-mediated system in C. reinhardtii has been published. In this study, we discuss the exact excision of a marker gene from the nuclear genome of C. reinhardtii via Cre/loxP-mediated site-specific recombination. This report expands the list of available genetics tools in this organism. Materials and Methods Plasmid Construction PCR for plasmid construction was performed using PrimeSTAR Max DNA polymerase (Takara) and appropriate primers, the sequences of which are listed in Table 1. aphVIII from Streptomyces rimosus encodes aminoglycoside 3′-phosphotransferase type VIII and confers resistance to paromomycin. The pSI103 plasmid carries the aphVIII expression cassette consisting of the C. reinhardtii HSP70_RBCS2 promoter, aphVIII, and the RBCS2 terminator [46]. For PCR amplification of the aphVIII expression cassette flanked by two direct repeats of loxP (loxP-P-aphVIII-T-loxP), the pSI103 plasmid was used as a template, and loxPphsp70_F1 and loxPtrbcS_R1 were employed as primers. The amplified fragment was digested using SmaI and XbaI and inserted between the SmaI and XbaI sites of the pBluescript II SK (+) plasmid to construct the ploxP-aphVIII plasmid (Fig 1). 10.1371/journal.pone.0161733.g001Fig 1 Structures of plasmids used in this study. The abbreviations of genes and loci were as follows: aphVIII, the gene for aminoglycoside 3′-phosphotransferase type VIII conferring paromomycin resistance; CrCRE, the codon-optimized gene for Cre recombinase; ble, the gene for bleomycin/Zeocin-binding protein conferring bleomycin/Zeocin resistance; Phsp70-rbcS2, the artificial tandem promoter consisting of the HSP70A and RBCS2 promoters; TrbcS2, the terminator of RBCS2. 10.1371/journal.pone.0161733.t001Table 1 Primers used in this study. Primer Sequence (5′ to 3′) Underlined sequence Primers for plasmid construction phsp70_F1 TCCCCGGGATAACTTCGTATAGCATACATTATACGAAGTTATGAGCTCGCTGAGGCTTGACA SmaI site trbcS_R1 CCTCTAGAATAACTTCGTATAATGTATGCTATACGAAGTTATCGCTTCAAATACGCCCAGCC XbaI site phsp70_F2 TCCCCGGGGAGCTCGCTGAGGCTTGACA SmaI site prbcS_R1 TCAGCAGGTTGCTCATGCTTCGAAATTCTTCAGCACCGG Underlined sequence is complementary to the underlined sequence of Crcre_F1 trbcS_F1 CGCCTGCTGGAGGACGGCGACTAAGGATCCCCGCTCCG Underlined sequence is complementary to the underlined sequence of Crcre_R1 trbcS_R2 CACTCTAGAGCTTCAAATACGCCCAGCCC XbaI site Crcre_F1 GAAGAATTTCGAAGCATGAGCAACCTGCTGACCGTGCACC Underlined sequence is complementary to the underlined sequence of prbcS_R1 Crcre_R1 CGGAGCGGGGATCCTTAGTCGCCGTCCTCCAGCAGGCG Underlined sequence is complementary to the underlined sequence of trbcS_F1 phsp70_F3 CACAAGCTTGACGGCGGGGAGCTCGCTGA HindIII site ble_R TTCTGGTGCACGGTCAGCAGGTTGTCCTGCTCCTCGGCCACG Underlined sequence is complementary to the underlined sequence of Crcre_F2 Crcre_F2 GCCGAGGAGCAGGACAACCTGCTGACCGTGCACCAGAAC Underlined sequence is complementary to the underlined sequence of ble_R ble_R2 GCGGCCGCCGGAGCCGCCGTCCTGCTCCTCGGCCACGAAGTG Underlined sequence encodes a linker peptide, and is complementary to the underlined sequence of Crcre_F3 Crcre_F3 GGCGGCTCCGGCGGCCGCATGAGCAACCTGCTGACCGTGCACCA Underlined sequence encodes a linker peptide, and is complementary to the underlined sequence of ble_R2 Primers used for the detection of specific sequences aphVIII_F ATGGACGATGCGTTGCGT aphVIII_R TCAGAAGAACTCGTCCAAC loxP_F AGCCCGGGATAACTTCGTA loxP_R GGCCGCTCTAGAATAACTTCGT Crcre_F4 GAGCACACCTGGAAGATGCT Crcre_R2 CAGGTAGTTGTTGGGGTCGT trbcS_inv_F1 GCGGTGGATGGAAGATACTGCTCTC aphVIII_F2 CGACTTGGAGGATCTGGACG phsp70_inv_R2 CCGCCAAATCAGTCCTGTAGCTTCA trbcS_inv_F2 AGTTTTGCAATTTTGTTGGTTGT trbcS_inv_R GGGGCAAGGCTCAGATCAAC LPm1_F2 TCTGATTTTGACTGATTTCGAGGC LPm1_R4 GGACAGGTATCCGGTAAGCG LPm19_F AGCACCGTGCACCACCTGCCTGCGCA LPm19_R GCGTTGGCCGATTCATTAATGCAGCT The codons of the Cre recombinase gene were optimized on the basis of the nuclear codon usage of C. reinhardtii stored in the codon usage database at Kazusa DNA Research Institute (http://www.kazusa.or.jp/codon/). Codon optimization was performed using the OptimumGene™ algorithm, and the optimized gene (CrCRE) was synthesized by GenScript (New Jersey, USA). The CrCRE sequence was cloned into the pUC57 plasmid to create the pUCrcre plasmid. The pCrcre plasmid (Fig 1) carrying the CrCRE sequence flanked by the HSP70-RBCS2 promoter and RBCS2 terminator (hereafter referred as “the CrCRE expression cassette”) was constructed using an overlapping PCR method as follows. In the first step, three DNA fragments were amplified separately using PCR: the 0.7-kb HSP70-RBCS2 promoter sequence was amplified using phsp70_F2 and prbcS_R1 as primers and the pSI103 plasmid as a template; the 0.3-kb RBCS2 terminator sequence was amplified using trbcS_F1 and trbcS_R2 as primers and the pSI103 plasmid as a template; and the 1.0-kb CrCRE sequence was amplified using Crcre_F1 and Crcre_R1 as primers and the pUCrcre plasmid as a template. In the second step, the three fragments amplified in the first step were assembled into a single fragment by PCR using the three fragments as templates and phsp70_F2 and trbcS_R2 as primers. The amplified product was purified using a PCR purification kit (Qiagen), digested with SmaI and XbaI, and cloned between the SmaI and XbaI sites of the pBluescript II SK (+) plasmid. To facilitate the nuclear localization of CrCre recombinase, CrCRE was fused in frame to ble from Streptoalloteicus hindustanus, conferring bleomycin/Zeocin resistance [47], to generate ble-CrCRE expression cassette I as follows. A 1.2-kb fragment containing the HSP70-RBCS2 promoter fused to ble was amplified using phsp70_F3 and ble_R as the primers and the pMF59 plasmid [47] as a template. A 1.3-kb fragment containing CrCRE fused to the RBCS2 terminator was amplified using Crcre_F2 and trbcS_R2 as primers and the pCrcre plasmid as a template. The 1.2- and 1.3-kb fragments were assembled into a single fragment by PCR with the phsp70_F3 and trbcS_R2 primers to form ble-CrCRE expression cassette I. The cassette DNA was purified using a PCR purification kit (Qiagen), digested with HindIII and XbaI, and cloned between the HindIII and XbaI sites of the pBluescript II SK (+) plasmid to generate the pbleCrcre plasmid (Fig 1). The pbleLCrcre plasmid harboring ble-CrCRE expression cassette II (Fig 1), in which a DNA sequence encoding a short artificial linker peptide, GGSGGR [48], was inserted in-frame between the 3'-end of ble and the 5'-end of CrCRE, was constructed as follows. First, PCR amplification was conducted using ble_R2 and Crcre_F3 as primers and the pbleCrcre plasmid as a template. Next, the amplified 5.5-kb fragment was circularized using an In-Fusion Cloning kit (Clontech) according to the manufacturer’s instructions. C. reinhardtii Strains and Growth Conditions C. reinhardtii strains cc124 (mt−) and cc4350 (cw15 arg7-8 mt+, Chlamydomonas Resource Center) were used as recipients of the ploxP-aphVIII plasmid, whereas strain cc125 (mt+) was used in backcross experiments. Cells were cultivated mixotrophically at 25°C in Tris acetate phosphate (TAP) medium [49] supplemented with 10 μg ml−1 arginine if necessary under white fluorescent light (100 μmol photons m−2 s−1) with gentle shaking or on solid medium supplemented with 1.5% Bacto agar (BD Difco). Genetic Transformation of C. reinhardtii Strains Nuclear transformation was performed using electroporation as described previously [50]. Briefly, the cells were grown for approximately 24 h until the cell densities reached 1 × 106–2 × 106 cells ml−1 in TAP medium. Cells were harvested by centrifugation at 800 × g for 5 min and washed with EP solution (30 mM HEPES, 5 mM MgSO4, 50 mM potassium acetate, 1 mM calcium acetate, 60 mM sucrose, pH 7.4), and suspend in EP solution to a final density of 1 × 108–3 × 108 cells ml−1. Then, 4 μl of 500 μg ml−1 DNA were added to 121 μl of the cell suspension. The cell suspension was placed into an electroporation cuvette with a 2-mm gap (Bio-Rad) and incubated at 15°C for 2 min. An exponential electric pulse of 2000 V/cm was applied to the suspension of strain cc124 using a GenePulser XCell™ (Bio-Rad) electroporation apparatus. The capacitance was set at 25 μF, and no shunt resistor was used. For strain cc4350, an exponential electric pulse of 700 V/cm at a capacitance of 600 μF was applied. After electroporation, cells were incubated at 15°C for 1 h and transferred to 10 ml of fresh TAP medium containing 40 mM sucrose. After incubation for 18 h at 25°C under dim light, the cells were collected by centrifugation at 800 × g for 5 min and selected on TAP agar plates supplemented with 20 μg ml−1 paromomycin (Wako) or 10 μg ml−1 Zeocin (Invitrogen). Each single colony developed on the agar plates was screened by PCR to identify gene-positive clones as described previously [51, 52] with some modifications. Each paromomycin-resistant (Pmr) clone was suspended in 10 μl of distilled water, into which the same volume of ethanol and 100 μl of 50% Chelex-100 (Bio-Rad, USA) were added. After incubation at 100°C for 10 min, cell debris was removed by centrifugation at 6000 rpm for 10 min. PCR was then performed using the supernatant as a template and aphVIII_F and aphVIII_R as primers to detect a partial aphVIII sequence or loxP_F and loxP_R as primers to detect the loxP-P-aphVIII-T-loxP sequence. In addition, PCR was performed to detect a partial CrCRE sequence using the primers Crcre_F4 and Crcre_R2, whereas detection of the full-length sequence of the CrCRE expression cassette, ble-CrCRE expression cassette I, or ble-CrCRE expression cassette II on the pCrcre, pbleCrcre, or pbleLCrcre plasmid was performed by two PCR amplifications using two primer sets: phsp70_F2 plus Crcre_R2 and Crcre_F4 plus trbcS_R2. The sequences of the primers used for the detection of transgenes are listed in Table 1. Southern Blot Analysis to Detect aphVIII Insertions C. reinhardtii genomic DNA was extracted using a standard phenol-chloroform protocol [53]. Five micrograms of genomic DNA were digested with BamHI, separated on 0.8% (w/v) agarose gel, and blotted onto a Hybond-N+ membrane (GE Healthcare, UK) by a standard capillary transfer method using 20 × SSC as a transfer buffer. The blotted membrane was then baked at 80°C for 2 h. An aphVIII fragment prepared by PCR using aphVIII_F and aphVIII_R as primers and the pSI103 plasmid as a template was labeled using a DIG High Prime DNA labeling and detection kit (Roche Applied Science). Hybridization and signal detection were performed according to the manufacturer’s instructions. Isolation of the Flanking Region of loxP-P-aphVIII-T-loxP Insertions DNA regions flanking the loxP-P-aphVIII-T-loxP insertion were determined using inverse PCR as follows. Genomic DNA (0.5 μg) of transformants carrying a single copy of the loxP-P-aphVIII-T-loxP sequence was digested with BamHI or PvuII (Takara), both enzymes being single cutters of the ploxP-aphVIII plasmid (Fig 1). After inactivation of the restriction enzymes using phenol, digested DNA was ethanol-precipitated and dissolved in TE buffer. To amplify the 5′-flanking region of the loxP-P-aphVIII-T-loxP insertion, PvuII-digested DNA was self-ligated and used as a template for an inverse PCR using trbcS_inv_F and phsp70_inv_R as primers. Similarly, to amplify the 3′-flanking region of the loxP-P-aphVIII-T-loxP insertion, BamHI-digested DNA was self-ligated and used as a template for an inverse PCR using trbcS_inv_F and trbcS_inv_R as primers. The PCR reactions were conducted using Advantage-GC Genomic PCR mix (Clontech) using the step-down PCR protocol according to the manufacturer’s instruction. The resulting amplified fragments were purified using a QIAquick Gel Extraction kit (Qiagen) and cloned into the pGEMT-Easy plasmid (Promega). The nucleotide sequences of the fragments were then determined using dideoxy chain termination via a commercial service provided by Macrogen Japan Corp. The nucleotide sequences thus obtained were compared with the genome sequence of Chlamydomonas at a Joint Genome Institute site (https://phytozome.jgi.doe.gov/pz/portal.html#!info?alias=Org_Creinhardtii). To verify excision of the loxP-P-aphVIII-T-loxP sequence integrated in the genomes of C. reinhardtii strains cc124_LPm1 and cc4350_LPm19 by Cre/loxP-mediated recombination, the insertion/excision regions were PCR-amplified using the primers designed from the sequences outside the loxP-P-aphVIII-T-loxP cassette sequence, namely primers LPm1_F and LPm1_R for the derivatives of strain cc124_LPm1 and primers LPm19_F and LPm19_R for the derivative of strain cc4350_LPm19. The nucleotide sequences of the PCR-amplified fragments were then determined as described previously. Reverse Transcription (RT)-PCR for the Detection of CrCRE Expression Total RNA was extracted from cells grown in TAP medium to an OD750 of 2.0 using a TRIzol® plus RNA purification kit (Ambion), and the remaining DNA was digested using a TURBO DNA-free kit (Ambion) according to the manufacturer’s instructions. First-strand cDNA was synthesized using a PrimeScript™ RT reagent kit with gDNA Eraser (Perfect Real Time, TaKaRa) and an RT primer mix containing oligo (dT)18 and random hexamers. PCR to confirm the expression of CrCRE was performed using primers Crcre_F4 and Crcre_R2. Backcrossing and Segregation Analysis Strain BLCP30, a derivative of cc124_LPm1 containing a single copy of loxP after Cre-mediated excision of the loxP-P-aphVIII-T-loxP sequence, was backcrossed to cc125 (mt+) to remove the CrCRE expression cassette. Mating was performed as described previously [54]. The resulting Zeocin-sensitive progenies were tested for the presence of the loxP sequence and the absence of the CrCRE expression cassette by PCR using primer sets LPm1_F2/LPm1_R4 and Crcre_F4/Crcre_R2. Accession Numbers Sequence data from this study can be found in the DDBJ/NCBI data libraries under the accession numbers LC150884 (pCrcre), LC150885 (pbleCrcre), and LC150883 (pbleLCrcre). Results and Discussion Construction and Characterization of C. reinhardtii Transformants Carrying a Single loxP-P-aphVIII-T-loxP Insertion The EcoRI-linearized ploxP-aphVIII plasmid was introduced in strains cc124 and cc4350, and 16 and 79 Pmr transformants, respectively, were isolated. The sequences of the loxP-flanked aphVIII expression cassettes (loxP-P-aphVIII-T-loxP) integrated in the genomes of these transformants were analyzed by PCR with primers loxP_F and loxP_R, and the integration of the whole loxP-P-aphVIII-T-loxP sequence was confirmed in 6 cc124-derived and 13 cc4350-derived Pmr transformants (Fig 2). Southern blot analyses were performed to detect the aphVIII sequence in BamHI-digested DNAs isolated from 4 cc124-derived and 13 cc4350-derived transformants carrying the whole loxP-P-aphVIII-T-loxP sequence. The BamHI restriction endonuclease cuts ploxP-aphVIII plasmid once at the 3′-end of aphVIII; therefore, the number of bands revealed by the probe corresponds to the number of aphVIII insertions in the host genomes. The top band in each lane was thought to be non-specific signals as the band was also observed in the lanes for the wild type strains, cc124 and cc4350. The analyses thus revealed that most transformants contained a single aphVIII insertion, whereas the remainder carried two (Fig 3). The sizes of the majority of the bands were different from each other, indicating that most of the loxP-P-aphVIII-T-loxP insertions were located at different loci on the C. reinhardtii chromosomes. Two transformants, cc124_LPm1 and cc4350_LPm19, each carrying a single copy of the loxP-P-aphVIII-T-loxP insertion, were selected for further studies to demonstrate the excision of the loxP-P-aphVIII-T-loxP insertion by CrCre recombinase. 10.1371/journal.pone.0161733.g002Fig 2 PCR analysis of transgenes in Pmr transformants of strains cc124 and cc4350. (A) Map of the aphVIII expression cassette. The positions of two PCR primers that amplify the loxP-P-aphVIII-T-loxP sequence are shown below the map. (B) Agarose gel electrophoresis of the PCR-amplified loxP-P-aphVIII-T-loxP fragments. Lane M, DNA marker (λ-EcoT14 I digest) with molecular size in bp. The template DNAs were as follows: lane P, the ploxP-aphVIII plasmid; lanes 1–7, genomic DNAs of Pmr transformants of strain cc124; lane N, no template. (C) Agarose gel electrophoresis of the PCR-amplified loxP-P-aphVIII-T-loxP fragments. Lane M, DNA marker (λ-EcoT14 I digest) with molecular size in bp. The template DNAs were as follows: lane P, the ploxP-aphVIII plasmid; lanes 1–14, genomic DNAs of Pmr transformants of strain cc4350; lane N, no template. 10.1371/journal.pone.0161733.g003Fig 3 Analysis of the aphVIII copy number by Southern blotting. Genomic DNAs were isolated from Pmr transformants of strains cc124 and cc4350, digested with BamHI, and hybridized with a digoxigenin-labeled aphVIII fragment. (A) Southern blot analysis of the genomic DNAs of four Pmr transformants derived from strain cc124. Lane M: DNA marker (λ-EcoT14 I digest) with molecular size in bp; next five lanes, BamH1-digested genomic DNA of the strains indicated above the lanes; lane ploxP-aphVIII, the ploxP-aphVIII plasmid digested with BamHI. (B) Southern blot analysis of the genomic DNAs of 13 Pmr transformants derived from strain cc4350. Lane M, DNA marker (λ-EcoT14 I digest) with molecular size in bp; next 14 lanes, BamHI-digested genomic DNAs of the strains indicated above the lanes; lane ploxP-aphVIII, the ploxP-aphVIII plasmid digested with BamHI. To map the insertion sites of the loxP-P-aphVIII-T-loxP sequence in strains cc124_LPm1 and cc4350_LPm19, flanking DNA regions were amplified using inverse PCR and sequenced as described in the Materials and Methods. The nucleotide sequences of the flanking regions were then aligned to the C. reinhardtii genome sequence [gene model version JGI 5.5 (Phytozome 10), Joint Genome Institute: http://www.phytozome.net/chlamy]. In strain cc124_LPm1, the loxP-P-aphVIII-T-loxP sequence was inserted in a gene of unknown function (Cre08.g362400, 1,099,596…1,102,295 on chromosome 8) at location 1,099,903, whereas the insertion site in the cc4350_LPm19 genome was mapped to multiple locations in the genome, which could not be determined unequivocally (Table 2). 10.1371/journal.pone.0161733.t002Table 2 The mapped locations of genomic sequence flanking of the loxP-P-aphVIII-T-loxP sequence in the cc4350_LPm19 genome. Position start end identity 5' flanking sequence chromosome_2 8931710 8932113 399/404 (98.8) chromosome_3 6813235 6812832 404/404 (100) chromosome_3 6861657 6862060 403/404 (99.8) chromosome_4 803530 803127 397/404 (98.3) chromosome_4 1679437 1679840 404/404 (100) chromosome_4 1682491 1682894 402/404 (99.5) chromosome_4 2994518 2994921 404/404 (100) chromosome_9 5934421 5934824 404/404 (100) chromosome_13 4248888 4248485 404/404 (100) chromosome_14 2096415 2096818 404/404 (100) chromosome_15 1378457 1378054 404/404 (100) chromosome_17 804556 804959 404/404 (100) chromosome_17 816331 816734 404/404 (100) chromosome_17 2661431 2661028 404/404 (100) scaffold_22 160645 160242 401/404 (99.3) 3' flanking sequence chromosome_2 8932114 8934462 2324/2349 (98.7) chromosome_3 6862061 6864404 2344/2344 (100) chromosome_3 6812831 6810512 2339/2344 (98.8) chromosome_4 803126 800786 2324/2344 (99.0) chromosome_4 1682895 1685238 2340/2344 (99.8) chromosome_4 2994922 2997265 2343/2344 (99.9) chromosome_8 2874233 2872350 1883/1884 (99.9) chromosome_9 5934825 5937168 2344/2344 (100) chromosome_13 4248484 4246141 2342/2344 (99.9) chromosome_14 2096819 2099151 2342/2344 (99.4) chromosome_15 1378053 1375720 2342/2344 (99.5) chromosome_17 804960 807292 2341/2344 (99.4) chromosome_17 816735 819078 2341/2344 (99.9) chromosome_17 2661027 2658684 2343/2344 (99.9) scaffold_22 160241 157991 2229/2251 (99.0) Demonstration of CrCre Recombinase-Mediated Site-Specific Recombination in C. reinhardtii To examine excision of the loxP-P-aphVIII-T-loxP sequence by CrCre recombinase, the pCrcre plasmid carrying the CrCRE expression cassette was introduced into strain cc4350_LPm19 via co-transformation with the pMF59 plasmid carrying ble conferring Zeocin resistance (Zeor), and Zeor transformants were screened on TAP agar plates containing Zeocin. The existence of the CrCRE expression cassette in 226 Zeor transformants was then examined by PCR with two primer sets: phsp70_F2 plus Crcre_R2 and Crcre_F4 plus trbcS_R2 (Fig 4A). The entire CrCRE cassette sequence was detected in 46 Zeor transformants. We first expected that all transformants carrying the intact CrCRE expression cassette would be Pm-sensitive (Pms), as the loxP-P-aphVIII-T-loxP sequence might have been excised by CrCre recombinase. However, only 1 of the 46 transformants was Pms, and excision of the aphVIII sequence in the Pms transformant was confirmed by PCR (Fig 4B). 10.1371/journal.pone.0161733.g004Fig 4 PCR amplification of the CrCRE expression cassette sequence integrated in the genomes of ZeoR transformants. (A) The structure of the CrCRE expression cassette and the sizes of the PCR products (1 and 2) amplified using two different primer sets. (B) Agarose gel electrophoresis of three different PCR products. Panels 1 and 2, detection of PCR products 1 and 2; panel aphVIII, detection of aphVIII. Lane M, DNA marker (λ-EcoT14 I digest) with molecular size in bp. The template DNAs were as follows: lane P, the pCrcre plasmid for panels 1 and 2 and the ploxP-aphVIII plasmid for panel aphVIII; lanes 1–4, genomic DNAs of Zeor transformants of strain cc4350_LPm19; lane N, no template. For panels 1 and 2, the primer sets phsp70_F/Crcre_R2 and Crcre_F4/trbcS_R2, respectively, were used, whereas for panel aphVIII, the primer set aphVIII_F/aphVIII_R was used. (C) Reverse transcription-PCR analysis of the expression of CrCRE and aphVIII in the ZeoR transformants of strain cc4350_LPm19 containing the CrCRE expression cassette. Panel CrCRE, detection of CrCRE transcripts. Panel aphVIII, the detection of aphVIII transcripts. Lane M, DNA markers (λ-EcoT14 I digest) with molecular size in bp. Lane P, template DNAs were extracted from the pCrcre plasmid for CrCRE detection and the ploxP-aphVIII plasmid for aphVIII detection; lanes 1–9, RNAs were isolated from the ZeoR transformants of strain cc4350_LPm19 containing the CrCRE expression cassette; lane N, minus reverse transcriptase. The same lane numbers within Fig4(A) and 4(B) do not represent the same transformants. This unexpectedly low excision rate of the loxP-P-aphVIII-T-loxP sequence in the pCrcre transformants may be due to several reasons. The first possibility was the low expression of CrCRE from the CrCRE expression cassette. Then, CrCRE expression was examined in nine randomly selected pCrcre transformants by RT-PCR using PCR primers Crcre_F4 and Crcre_R2. CrCRE expression was detected in seven of nine strains, whereas aphVIII expression was detected in all strains (Fig 4C). To overcome potential problems including malfunction of translation and/or inefficient nuclear translocation of the CrCre protein, CrCRE was fused to ble to construct the pbleCrcre plasmid (Fig 1). There were two reasons for the construction of the Ble-CrCre fusion proteins: (i) As the level of resistance to Zeocin is proportional to the protein expression level of Ble [55], transformants expressing the Ble-CrCre fusion protein at high levels could readily be isolated by selecting for Zeor at higher levels. (ii) Ble is a bleomycin-binding protein that localizes in the nucleus [47]; thus, fusion with the Ble protein would further facilitate the nuclear translocation of the CrCre recombinase. When the pbleCrcre plasmid carrying ble-CrCRE expression cassette I was introduced into strain cc4350_LPm19 by selecting Zeor transformants, excision of the loxP-P-aphVIII-T-loxP sequence was not detected. We expect that the CrCre recombinase directly fused to the Ble protein was not functional in the Zeor transformants probably because two domains in the bifunctional fusion protein were not effectively separated each other [56, 57], or that the fusion protein had a high chance of misfolding [58]. A fusion gene encoding the Ble protein fused to CrCre recombinase via a flexible linker of six amino acids was then designed. The pbleLCrcre plasmid harboring ble-CrCRE expression cassette II ([the HSP-RBCS promoter]–[the ble-linker-CrCRE fusion protein gene]–[the RBCS terminator]) (Fig 1) was introduced into strain cc4350_LPm19, and Zeor transformants were screened on TAP agar plates containing Zeocin. Seventy-four Zeor transformants were obtained, and the existence of the ble-CrCRE expression cassette II sequence in the transformants was examined by PCR as described previously (Fig 5A). In the genomes of 12 of 74 transformants, the intact ble-CrCRE expression cassette II was integrated (Fig 5B). The absence of the aphVIII sequence in the genome of 12 transformants was next examined by PCR using primers aphVIII-F and aphVIII-R. The aphVIII sequence was not detected in four of the transformants (Fig 5B). These four aphVIII-free transformants were Pms, whereas the remaining eight transformants were Pmr. The four aphVIII-free transformants were named strains BLCP1, BLCP6, BLCP15, and BLCP17. From the remaining eight Pmr transformants, aphVIII-free descendants were isolated after single-colony isolation repeated 2–6 times, indicating that CrCre recombinase-mediated site-specific recombination could be delayed, requiring many generations to elapse before recombination. 10.1371/journal.pone.0161733.g005Fig 5 PCR amplification of the ble-CrCRE expression cassette II sequence integrated into the Zeor transformants. (A) The structure of ble-CrCRE expression cassette II, positions of two PCR primer sets, and PCR products 1 and 2 amplified with the two primer sets are shown. (B) Agarose gel electrophoresis of three different PCR products. Panels 1 and 2, detection of PCR products 1 and 2; panel aphVIII, detection of aphVIII. Lane M, DNA marker (λ-EcoT14 I digest) with molecular size in bp. The template DNAs were as follows: lane P, the pbleLCrcre plasmid for panels 1 and 2 and the ploxP-aphVIII plasmid for panel aphVIII; next 13 lanes, genomic DNAs of the strains indicated above the lanes; lane N, no template. The Zeor transformants of strain cc4350_LPm19 carrying ble-CrCRE expression cassette II were named BLPC. The aphVIII sequence was not detected in strains BLCP1, BLCP6, BLCP15, and BLCP17. Genomic DNA was extracted from strains BLCP6, BLCP15, and BLCP17, and Southern blotting with a probe specific for the aphVIII sequence was performed (Fig 6). The aphVIII signal was not detected in these three strains. The DNA sequences of strains BLCP6, BLCP15, and BLCP17 corresponding to the loxP-P-aphVIII-T-loxP integration site in their parental strain, cc4350_LPm19, were analyzed by PCR using primers LPm19_F and LPm19_R. A 2.7-kb fragment was amplified from strain cc4350_LPm19, whereas a 0.9-kb fragment was amplified from BLCP6, BLCP15, and BLCP17 (Fig 7). The nucleotide sequence of the 0.9-kb fragment revealed that the loxP-P-aphVIII-T-loxP sequence was accurately excised by Cre/loxP-mediated recombination, leaving a single copy of loxP. 10.1371/journal.pone.0161733.g006Fig 6 Southern blot analysis of transgenes in strains BLCP6, BLCP15, and BLCP17. All DNAs were digested with BamHI, electrophoresed, and hybridized with a digoxigenin-labeled aphVIII fragment. The arrow indicates the restriction fragment containing the loxP-P-aphVIII-T-loxP sequence in the genome of strain cc4350_LPm19. 10.1371/journal.pone.0161733.g007Fig 7 PCR analysis to detect Cre-mediated excision of the loxP-P-aphVIII-T-loxP sequence. (A) The box indicates a partial sequence of the ploxP-aphVIII plasmid, and solid lines indicate the genomic sequence. The loxP sites are indicated by short arrows. The two arrows beneath the image denote the PCR primer set used for the amplification of either a 2.7-kb fragment from DNA without the excision event or a 0.9-kb fragment from DNA with the Cre-mediated excision. (B) Agarose gel electrophoresis of the 2.7- and 0.9-kb fragments. Lane M1, DNA size marker (λ-EcoT14 I digest) with molecular size in bp; lane M2, DNA size marker (50–2500 bp, Lonza); next five lanes, PCR fragments amplified from genomic DNAs of the strains indicated above the lanes; lane N, no template control. ble-CrCRE expression cassette II was also introduced into strain cc124_LPm1. In the genomes of 5 of the 163 Zeor transformants, the entire cassette was integrated, and one of the five Zeor transformants was Pms. This strain was named BLCP30, and the aphVIII sequence was absent in its genome (Fig 8A). The DNA sequence of strain BLCP30 corresponding to the loxP-P-aphVIII-T-loxP insertion site in its parental strain, cc124_LPm1, was analyzed by PCR using primers LPm1_F2 and LPm1_R4. The 2.9-kb fragment was amplified from strain cc124_LPm1, whereas a 1.1-kb fragment was amplified from strain BLCP30 (Fig 8B). The accurate excision of loxP-P-aphVIII-T-loxP leaving a single copy of loxP was confirmed by nucleotide sequencing of the 1.1-kb fragment. 10.1371/journal.pone.0161733.g008Fig 8 PCR analysis to detect the Cre-mediated excision of the loxP-P-aphVIII-T-loxP sequence. (A) Agarose gel electrophoresis to detect aphVIII. Lane M, DNA size marker (λ-EcoT14 I digest) with molecular size in kb. The template DNAs were as follows: lane P, the ploxP-aphVIII plasmid; next five lanes, DNAs of the strains indicated above the lanes; lane N, no template. The Zeor transformants of strain cc124_LPm1 carrying ble-CrCRE expression cassette II were named BLCP. No aphVIII signal was detected in strain BLPC30 indicating the excision of the loxP-P-aphVIII-T-loxP sequence in this strain. (B) PCR amplification of the loxP-P-aphVIII-T-loxP integration sites of strains cc124, cc124_LPm1, and BLCP30. Removal of ble-CrCRE Expression Cassette II via Backcross to a Wild-Type Strain ble-CrCRE expression cassette II remained in the genomes of aphVIII-cured derivatives. The presence of ble-CrCRE expression cassette II in a host genome hinders the subsequent introduction of a loxP-flanked marker gene into the host. Furthermore, constitutive expression of CrCre recombinase may induce DNA damage at off-target sites [59, 60, 61, 62]. To remove ble-CrCRE expression cassette II from strain BLCP30, which is a descendent of strain cc124 (mt−), this strain was crossed to the wild-type strain cc125 (mt+). Seven tetrads were dissected, and 25 recombinant progenies were isolated. They were grown on TAP plates for 72 h and tested for their phenotypes. In total, 12 of 25 progeny were sensitive to Zeocin, and four carried the loxP sequence. The absence of CrCRE in the genomes of the four progeny was also confirmed by PCR (Fig 9). 10.1371/journal.pone.0161733.g009Fig 9 Removal of ble-CrCRE expression cassette II sequence by backcross. Agarose gel electrophoresis of PCR-amplified loxP and CrCRE sequences from the genomes of four progeny obtained by backcross between strains BLCP30 and cc125. Lane M, DNA size marker (λ-EcoT14 I digest) with molecular size in kb. Template DNAs were extracted from the following: lane 1, strain cc124; lane 2, strain BLCP30; lanes 3–6, progeny; lane N, no template. Conclusion In this study, we developed a method to obtain marker-free transgenic strains in C. reinhardtii, and the steps of the method are outlined in Fig 10. The Cre/loxP-mediated precise marker excision method applicable to transgenic C. reinhardtii could further increase the potential of this organism for use in basic and applied research. 10.1371/journal.pone.0161733.g010Fig 10 Summary of the method to obtain marker-free transgenic strains in C. reinhardtii. We thank Takako Minoura and Ritsu Kamiya for advice concerning the Chlamydomonas mating experiments and for the kind gifts of the pSI103 plasmid and strains cc124 (mt−) and cc125 (mt+). We also thank Jun Abe for the kind gift of the pMF59 plasmid. This work was supported by the New Energy and Industrial Technology Development Organization (NEDO, P11502725-0). ==== Refs References 1 Grossman AR , Lohr M , Im CS . Chlamydomonas reinhardtii in the landscape of pigments . Annu Rev Genet . 2004 ; 38 : 119 –173 . 15568974 2 Rochaix JD . Chlamydomonas reinhardtii as the photosynthetic yeast . Annu Rev Genet . 1995 ; 29 : 209 –230 . 8825474 3 Dutcher SK . Elucidation of basal body and centriole functions in Chlamydomonas reinhardtii . Traffic . 2003 ; 4 : 443 –451 . 12795689 4 Specht E , Miyake-Stoner S , Mayfield S . Micro-algae come of age as a platform for recombinant protein production . Biotechnol Lett . 2010 ; 32 : 1373 –1383 . 10.1007/s10529-010-0326-5 20556634 5 Siaut M , Cuiné S , Cagnon C , Fessler B , Nguyen M , Carrier P , et al Oil accumulation in the model green alga Chlamydomonas reinhardtii: characterization, variability between common laboratory strains and relationship with starch reserves . BMC Biotechnol . 2011 ; 11 : 7 10.1186/1472-6750-11-7 21255402 6 Fan J , Yan C , Andre C , Shanklin J , Schwender J , Xu C . Oil accumulation is controlled by carbon precursor supply for fatty acid synthesis in Chlamydomonas reinhardtii . Plant Cell Physiol . 2012 ; 53 : 1380 –1390 . 10.1093/pcp/pcs082 22642988 7 Iwai M , Ikeda K , Shimojima M , Ohta H . Enhancement of extraplastidic oil synthesis in Chlamydomonas reinhardtii using a type-2 diacylglycerol acyltransferase with a phosphorus starvation-inducible promoter . Plant Biotechnol J . 2014 ; 12 : 808 –819 . 10.1111/pbi.12210 24909748 8 Scaife MA , Nguyen GT , Rico J , Lambert D , Helliwell KE , Smith AG . Establishing Chlamydomonas reinhardtii as an industrial biotechnology host . Plant J . 2015 ; 82 : 532 –546 . 10.1111/tpj.12781 25641561 9 Chen L , Marmey P , Taylor NJ , Brizard JP , Espinoza C , D’Cruz P , et al Expression and inheritance of multiple transgenes in rice plants . Nat biotechnol . 1998 ; 16 : 1060 –1064 . 9831036 10 Wu L , Nandi S , Chen L , Rodriguez RL , Huang N . Expression and inheritance of nine transgenes in rice . Transgenic Res . 2002 ; 11 : 533 –541 . 12437084 11 Li R , Bancroft B , Hutcheon C , Zhao S , Zheng S , Blahut-Beatty L , et al Multiple inserts of gene of interest and selectable marker gene are co-integrated and stably transmitted as a single genetic locus in transgenic soybean plants . In Vitro Cell Dev Biol Plant . 2011 ; 47 : 274 –281 . 12 De Block M , Debrouwer D . Two T-DNA’s co-transformed into Brassica napus by a double Agrobacterium tumefaciens infection are mainly integrated at the same locus . Theor Appl Genet . 1991 ; 82 : 257 –263 . 10.1007/BF02190610 24213168 13 De Neve M , De Buck S , Jacobs A , Van Montagu M , Depicker A . T-DNA integration patterns in co-transformed plant cells suggest that T-DNA repeats originate from co-integration of separate T-DNAs . Plant J . 1997 ; 11 : 15 –29 . 9025300 14 Radchuk VV , Van DT , Klocke E . Multiple gene co-integration in Arabidopsis thaliana predominantly occurs in the same genetic locus after simultaneous in planta transformation with distinct Agrobacterium tumefaciens strains . Plant Sci . 2005 ; 168 : 1515 –1523 . 15 Periz G , Keller LR . DNA elements regulating alpha1-tubulin gene induction during regeneration of eukaryotic flagella . Mol Cell Biol . 1997 ; 17 : 3858 –3866 . 9199320 16 Heitzer M , Zschoernig B . Construction of modular tandem expression vectors for the green alga Chlamydomonas reinhardtii using the Cre/lox-system . Biotechniques . 2007 ; 43 : 324 –332 . 17907575 17 Lauersen KJ , Kruse O , Mussgnug JH . Targeted expression of nuclear transgenes in Chlamydomonas reinhardtii with a versatile, modular vector toolkit . Appl Microbiol Biotechnol . 2015 ; 99 : 3491 –3503 . 10.1007/s00253-014-6354-7 25586579 18 Rasala BA , Lee PA , Shen Z , Briggs SP , Mendez M , Mayfield SP . Robust expression and secretion of Xylanase1 in Chlamydomonas reinhardtii by fusion to a selection gene and processing with the FMDV 2A peptide . PLoS One . 2012 ; 7 : e43349 10.1371/journal.pone.0043349 22937037 19 Merchant SS , Prochnik SE , Vallon O , Harris EH , Karpowicz SJ , Witman GB , et al The Chlamydomonas genome reveals the evolution of key animal and plant functions . Science . 2007 ; 318 : 245 –250 . 17932292 20 Sodeinde OA , Kindle KL . Homologous recombination in the nuclear genome of Chlamydomonas reinhardtii . Proc Natl Acad Sci USA . 1993 ; 90 : 9199 –9203 . 8415677 21 Gumpel NJ , Rochaix JD , Purton S . Studies on homologous recombination in the green alga Chlamydomonas reinhardtii . Curr Genet . 1994 ; 26 : 438 –442 . 7702662 22 Zorin B , Hegemann P , Sizova I . Nuclear-gene targeting by using single-stranded DNA avoids illegitimate DNA integration in Chlamydomonas reinhardtii . Eukaryot Cell . 2005 ; 4 : 1264 –1272 . 16002652 23 Tam LW , Lefebvre PA . Cloning of flagellar genes in Chlamydomonas reinhardtii by DNA insertional mutagenesis . Genetics . 1993 ; 135 : 375 –384 . 8244002 24 Pazour GJ , Witman GB . Forward and reverse genetic analysis of microtubule motors in Chlamydomonas. Methods . 2000 ; 22 : 285 –298 . 11133235 25 González-Ballester D , de Montaigu A , Higuera JJ , Galván A , Fernández E . Functional genomics of the regulation of the nitrate assimilation pathway in Chlamydomonas . Plant Physiol . 2005 ; 137 : 522 –533 . 15665251 26 Pollock SV , Colombo SL , Prout DL Jr, Godfrey AC , Moroney JV . Rubisco activase is required for optimal photosynthesis in the green alga Chlamydomonas reinhardtii in a low-CO2 atmosphere . Plant Physiol . 2003 ; 133 : 1854 –1861 . 14605215 27 Matsuo T , Okamoto K , Onai K , Niwa Y , Shimogawara K , Ishiura M . A systematic forward genetic analysis identified components of the Chlamydomonas circadian system . Genes Dev . 2008 ; 22 : 918 –930 . 10.1101/gad.1650408 18334618 28 Johnson X , Steinbeck J , Dent RM , Takahashi H , Richaud P , Ozawa S , et al Proton gradient regulation 5-mediated cyclic electron flow under ATP- or redox-limited conditions: a study of DeltaATpase pgr5 and DeltarbcL pgr5 mutants in the green alga Chlamydomonas reinhardtii . Plant Physiol . 2014 ; 165 : 438 –452 . 10.1104/pp.113.233593 24623849 29 Houille-Vernes L , Rappaport F , Wollman FA , Alric J , Johnson X . Plastid terminal oxidase 2 (PTOX2) is the major oxidase involved in chlororespiration in Chlamydomonas . Proc Natl Acad Sci USA . 2011 ; 108 : 20820 –20825 . 10.1073/pnas.1110518109 22143777 30 Ebinuma H , Sugita K , Matsunaga E , Endo S , Yamada K , Komamine A . Systems for the removal of a selection marker and their combination with a positive marker . Plant Cell Rep . 2001 ; 20 : 383 –392 . 31 Dale EC , Ow DW . Genetic-transfer with subsequent removal of the selection gene from the host genome . Proc Natl Acad Sci USA . 1991 ; 88 : 10558 –10562 . 1660141 32 Gleave AP , Mitra DS , Mudge SR , Morris BA . Selectable marker-free transgenic plants without sexual crossing: transient expression of cre recombinase and use of a conditional lethal dominant gene . Plant Mol Biol . 1999 ; 40 : 223 –235 . 10412902 33 Hoa TT , Bong BB , Huq E , Hodges TK . Cre/lox site-specific recombination controls the excision of transgene from the rice genome . Theor Appl Genet . 2002 ; 104 : 518 –525 . 12582653 34 Sreekala C , Wu L , Gu K , Wang D , Tian D , Yin Z . Excision of a selectable marker in transgenic rice (Oryza sativa L.) using a chemically regulated Cre/loxP system . Plant Cell Rep . 2005 ; 24 : 86 –94 . 15662501 35 Lyznik LA , Michell JC , Hirayama L , Hodges TK . Activity of yeast FLP recombinase in maize and rice protoplasts . Nucleic Acids Res . 1993 ; 21 : 969 –975 . 8451196 36 Lloyd AM , Davis RW . Functional expression of the yeast FLP/FRT site-specific recombination system in Nicotiana tabacum . Mol Gen Genet . 1994 ; 242 : 653 –657 . 8152416 37 Sugita K , Matsunaga E , Kasahara T , Ebinuma H . Transgene stacking in plants in the absence of sexual crossing . Mol Breeding . 2000 ; 6 : 529 –536 . 38 Endo S , Sugita K , Sakai M , Tanaka H , Ebinuma H . Single-step transformation for generating marker-free transgenic rice using the ipt-type MAT vector system . Plant J . 2002 ; 30 : 115 –122 . 11967098 39 Maeser S , Kahmann R . The Gin recombinase of phage Mu can catalyze site-specific recombination in plant protoplasts . Mol Gen Genet . 1991 ; 230 : 170 –176 . 1836050 40 Sternberg N , Hamilton D . Bacteriophage P1 site-specific recombination. I. Recombination between loxP sites . J Mol Biol . 1981 ; 150 : 467 –486 . 6276557 41 Abremski K , Hoess R , Sternberg N . Studies on the properties of P1 site-specific recombination: evidence for topologically unlinked products following recombination . Cell . 1983 ; 32 : 1301 –1311 . 6220808 42 Hoess RH , Abremski K . Mechanism of strand cleavage and exchange in the cre-lox site specific recombination system . J Mol Biol . 1985 ; 181 : 351 –362 . 3856690 43 Sauer B , Henderson N . Site-specific DNA recombination in mammalian cells by the Cre recombinase of bacteriophage P1 . Proc Natl Acad Sci USA . 1988 ; 85 : 5166 –5170 . 2839833 44 Kühn R , Torres R . Cre/loxP recombination system and gene targeting . Methods Mol Biol . 2002 ; 180 : 175 –204 . 11873650 45 Mizutani O , Masaki K , Gomi K , Iefuji H . Modified cre-loxP recombination in Aspergillus oryzae by direct introduction of cre recombinase for marker gene rescue . Appl Environ Microbiol . 2012 ; 78 : 4126 –4133 . 10.1128/AEM.00080-12 22504800 46 Sizova I , Fuhrmann M , Hegemann P . A Streptomyces rimosus aphVIII gene coding for a new type phosphotransferase provides stable antibiotic resistance to Chlamydomonas reinhardtii . Gene . 2001 ; 277 : 221 –229 . 11602359 47 Fuhrmann M , Oertel W , Hegemann P . A synthetic gene coding for the green fluorescent protein (GFP) is a versatile reporter in Chlamydomonas reinhardtii . Plant J . 1999 ; 19 : 353 –361 . 10476082 48 Ohnishi N , Mukherjee B , Tsujikawa T , Yanase M , Nakano H , Moroney JV , et al Expression of a low CO2-inducible protein, LCI1, increases inorganic carbon uptake in the green alga Chlamydomonas reinhardtii . Plant Cell . 2010 ; 22 : 3105 –3117 . 10.1105/tpc.109.071811 20870960 49 Gorman DS , Levine RP . Cytochrome f and plastocyanin: their sequence in the photosynthetic electron transport chain of Chlamydomonas reinhardtii . Proc Natl Acad Sci USA . 1965 ; 54 : 1665 –1669 . 4379719 50 Shimogawara K , Fujiwara S , Grossman A , Usuda H . High-efficiency transformation of Chlamydomonas reinhardtii by electroporation . Genetics . 1998 ; 148 : 1821 –1828 . 9560396 51 Cao M , Fu Y , Guo Y , Pan J . Chlamydomonas (Chlorophycease) colony PCR . Protoplasma . 2009 ; 235 : 107 –110 . 10.1007/s00709-009-0036-9 19242652 52 Wan M , Rosenberg JN , Faruq J , Betenbaugh MJ , Xia J . An improved colony PCR procedure for genetic screening of Chlorella and related microalgae . Biotechnol Lett . 2011 ; 33 : 1615 –1619 . 10.1007/s10529-011-0596-6 21431847 53 Sambrook J , Fritsh EF , Maniatis T . Molecular Cloning: A Laboratory Manual New York : Cold Spring Harbor Laboratory Press ; 1989 . 54 Wang Q , Snell WJ . Flagellar adhesion between mating type plus and mating type minus gametes activates a flagellar protein-tyrosine kinase during fertilization in Chlamydomonas . J Biol Chem . 2003 ; 278 : 32936 –32942 . 12821679 55 Dumas P , Bergdoll M , Cagnon C , Masson JM . Crystal structure and site-directed mutagenesis of a Bleomycin resistance protein and their significance for drug sequestering . EMBO J . 1994 ; 13 : 2483 –2492 . 7516875 56 Bai Y , Ann DK , Shen WC . Recombinant granulocyte colony-stimulating factor-transferrin fusion protein as an oral myelopoietic agent . Proc Natl Acad Sci USA . 2005 ; 102 : 7292 –7296 . 15870205 57 Bai Y , Shen WC . Improving the oral efficacy of recombinant granulocyte colony-stimulating factor and transferrin fusion protein by spacer optimization . Pharm Res . 2006 ; 23 : 2116 –2121 . 16952003 58 Zhao H , Yao X , Xue C , Wang Y , Xiong X , Liu Z . Increasing the homogeneity, stability and activity of human serum albumin and interferon-alpha2b fusion protein by linker engineering . Protein Expr Purif . 2008 ; 61 : 73 –77 . 10.1016/j.pep.2008.04.013 18541441 59 Loonstra A , Vooijs M , Beverloo HB , Allak BA , van Drunen E , Kanaar R , et al Growth inhibition and DNA damage induced by Cre recombinase in mammalian cells . Proc Natl Acad Sci USA . 2001 ; 98 : 9209 –9214 . 11481484 60 Pfeifer A , Brandon EP , Kootstra N , Gage FH , Verma IM . Delivery of the Cre recombinase by a self-deleting lentiviral vector: efficient gene targeting in vivo . Proc Natl Acad Sci USA . 2001 ; 98 : 11450 –11455 . 11553794 61 Baba Y , Nakano M , Yamada Y , Saito I , Kanegae Y . Practical range of effective dose for Cre recombinase-expressing recombinant adenovirus without cell toxicity in mammalian cells . Microbiol Immunol . 2005 ; 49 : 559 –570 . 15965304 62 Janbandhu VC , Moik D , Fässler R . Cre recombinase induces DNA damage and tetraploidy in the absence of loxP sites . Cell Cycle . 2014 ; 13 : 462 –470 . 10.4161/cc.27271 24280829

PMC005xxxxxx/PMC5001724.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756372410.1371/journal.pone.0161752PONE-D-16-03604Research ArticleSocial SciencesLinguisticsSpeechBiology and Life SciencesPsychologyEmotionsSocial SciencesPsychologyEmotionsEngineering and TechnologySignal ProcessingSpeech Signal ProcessingResearch and Analysis MethodsMathematical and Statistical TechniquesStatistical MethodsForecastingPhysical SciencesMathematicsStatistics (Mathematics)Statistical MethodsForecastingBiology and Life SciencesNeuroscienceCognitive ScienceArtificial IntelligenceMachine LearningComputer and Information SciencesArtificial IntelligenceMachine LearningBiology and Life SciencesNeuroscienceCognitive ScienceCognitionDecision MakingBiology and Life SciencesBehaviorBiology and Life SciencesNeuroscienceCognitive ScienceCognitive PsychologyLearningBiology and Life SciencesPsychologyCognitive PsychologyLearningSocial SciencesPsychologyCognitive PsychologyLearningBiology and Life SciencesNeuroscienceLearning and MemoryLearningStrength Is in Numbers: Can Concordant Artificial Listeners Improve Prediction of Emotion from Speech? A Novel Paradigm for Speech Emotion RecognitionMartinelli Eugenio 1Mencattini Arianna 1http://orcid.org/0000-0002-0919-0475Daprati Elena 2*Di Natale Corrado 11 Department of Electronic Engineering, University of Rome Tor Vergata, Rome, Italy2 Department of System Medicine and CBMS, University of Rome Tor Vergata, Rome, ItalyKotz Sonja EditorMax Planck Institute for Human Cognitive and Brain Sciences, GERMANYCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: EM AM ED CD. Formal analysis: EM AM ED. Funding acquisition: AM EM ED. Investigation: EM AM. Methodology: EM AM CD. Project administration: EM. Software: EM AM CD. Validation: EM AM ED. Visualization: EM AM. Writing – original draft: EM AM ED. Writing – review & editing: EM AM ED CD. * E-mail: elena.daprati@uniroma2.it26 8 2016 2016 11 8 e016175226 1 2016 11 8 2016 © 2016 Martinelli et al2016Martinelli et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Humans can communicate their emotions by modulating facial expressions or the tone of their voice. Albeit numerous applications exist that enable machines to read facial emotions and recognize the content of verbal messages, methods for speech emotion recognition are still in their infancy. Yet, fast and reliable applications for emotion recognition are the obvious advancement of present ‘intelligent personal assistants’, and may have countless applications in diagnostics, rehabilitation and research. Taking inspiration from the dynamics of human group decision-making, we devised a novel speech emotion recognition system that applies, for the first time, a semi-supervised prediction model based on consensus. Three tests were carried out to compare this algorithm with traditional approaches. Labeling performances relative to a public database of spontaneous speeches are reported. The novel system appears to be fast, robust and less computationally demanding than traditional methods, allowing for easier implementation in portable voice-analyzers (as used in rehabilitation, research, industry, etc.) and for applications in the research domain (such as real-time pairing of stimuli to participants’ emotional state, selective/differential data collection based on emotional content, etc.). Research supported by PainTCare project (University of Rome Tor Vergata, Uncovering Excellence program).Mencattini Arianna Research supported by PainTCare project (University of Rome Tor Vergara, Uncovering Excellence program) (AM EM ED). Data AvailabilityAll relevant data relating to the algorithm and the simulations described in the paper are within the paper and its Supporting Information files. Speech samples used for the study can be obtained (academic use only) from (https://diuf.unifr.ch/diva/recola/download.html). References for the RECOLA database are available in Ringeval, F., Sonderegger, A., Sauer, J., & Lalanne, D. Introducing the RECOLA multimodal corpus of remote collaborative and affective interactions, in Proc. of Face & Gestures 2013, 2nd IEEE Inter. Workshop on Emotion Representation, Analysis and Synthesis in Continuous Time and Space (EmoSPACE), (2013).Data Availability All relevant data relating to the algorithm and the simulations described in the paper are within the paper and its Supporting Information files. Speech samples used for the study can be obtained (academic use only) from (https://diuf.unifr.ch/diva/recola/download.html). References for the RECOLA database are available in Ringeval, F., Sonderegger, A., Sauer, J., & Lalanne, D. Introducing the RECOLA multimodal corpus of remote collaborative and affective interactions, in Proc. of Face & Gestures 2013, 2nd IEEE Inter. Workshop on Emotion Representation, Analysis and Synthesis in Continuous Time and Space (EmoSPACE), (2013). ==== Body Introduction One of the most irritating features of virtual receptionists is their being utterly impermeable to the emotional outbursts of callers, who, consequently, feel more neglected and less satisfied than when interacting with human attendants. Indeed, despite complexity of the non-verbal signals conveyed by the voice, humans easily recognize them, and react accordingly. Conversely, machines do not detect the emotional information embedded in the voice and, consequently, the human partner may become annoyed by the apparent lack of empathy. Thus, it is not surprising that speech emotion recognition systems (SER) have recently become of interest to the domain of human-machine interfaces [1–2], although their application is relevant also for treatment of psychiatric and neurologic conditions affecting the emotional sphere (e.g. autism [3–4] Parkinson Disease [5–7], mood disorders [8]). Regardless the application, a priority for SER systems is obtaining fast, online labeling of long speech sequences [9]. On this respect, a promising opportunity comes from the domain of machine learning and, specifically, from semi-supervised and unsupervised learning machines, which are typically used when a huge amount of data requires labeling (i.e. in diagnostic imaging, remote sensing imaging, etc.)[10–11]. In standard supervised learning methods, an initial model is firstly trained using a set of (pre-)labeled data. This model is then employed to automatically describe unlabeled data, using the machine-generated examples to improve prediction capabilities. In contrast, in semi-supervised learning approaches, an initial model provides estimates for some of the unlabeled data. Then, these machine-labeled examples are added to the training set, the model is retrained, and the process iterated. In both cases the learning phase is usually performed on descriptors extracted from long speech sequences coming from different speakers [12–13]. Consequently, accuracy and speed of the system is largely dependent on the amount of computational and data resources involved: the larger the amount the longer (and more cumbersome) the computation. As a result, most of the available SER systems show limited performances, precluding their inclusion in applications where they could be extraordinarily useful [14–16], such as smartphones and tablets. In human behavior, strength is often in numbers. Truth of the “vox populi” concept was first demonstrated by Galton in the early 20th century [17]. At an annual fair, about eight hundred visitors estimated the weight of an ox. Individual guesses ranged widely and often wrongly. In contrast, the middlemost estimate of the distribution, i.e. the vox populi, was correct to within 1 per cent of the real value. That large groups are better problem-solvers than single individuals is now well established in a number of domains (see for instance [18]). This “wisdom of the crowds” mainly comes from the large diversity of opinions and has proved an efficient feature in many areas, including that of diagnostics, improving fracture classification reliability [19]. In a similar way, we reasoned that by simulating a ‘wise crowd’ we could devise an efficient model of group decision-making that could be applied to a novel SER system. Here we describe a cooperative learning strategy (Fig 1) that simulates human decision-making in the social domain [20–22]. Group decision-making is a complex process that takes into account the distribution of individual preferences within a group and combines them to reach a collective response (e.g. Social Decision Scheme theory [23]). The combinatorial process can follow different decision schemes (e.g. majority vote, consensus, etc.) and may vary as a function of factors such as task and context. In the algorithm we propose, speech samples are fed to a classifier modeled to work as a group of individuals reaching a decision through a consensus scheme. In this analogy, consensus could be viewed as the means to extract the middlemost estimate or the ‘vox populi’ of Galton’s example. Specifically, for a set of speakers whose speech sequences and corresponding annotations were known, we trained and optimized separate regression models (Single Speaker Regression Model, SSRM). The ensemble of these models is shown in Fig 1 (left side), each colored circular block exemplifying one SSRM. 10.1371/journal.pone.0161752.g001Fig 1 Schematic description of the prediction system. Upper part: speech is continuously recorded in a naturalistic environment. In small time windows, acoustic descriptors are extracted and processed. A Cooperative Regression model provides an estimation of speaker’s perceived emotional status in terms of arousal and valence, and plots the detected emotions onto a circumplex diagram. Lower part: a schematic description of the functioning of the Cooperative Regression model. A new speech sample (unlabeled speaker) is fed to the algorithm for evaluation. Each Single Speaker Regression Model (SSRM) in the ensemble provides its own prediction about the new sequence. Dimensions and colors of the circles represent heterogeneity of the SSRM ensemble, which is a critical step into the model’s functioning and adaptability to different scenarios. Dimensions indicate that each SSRM can be trained on differently sized speech sequences and that descriptors extracted from each sequence can differ, as well as model parameters. Similarly, colors indicate that the emotional content can differ for each model. Predictions expressed by each SSRM are combined via a consensus rule. Newly labeled samples are added to the pool unless redundant (i.e. their content emulates that of samples already in the pool)(RED-EX criterion). Unlabeled speakers (yellow question mark in Fig 1) are handled by the model as a human group would handle newcomers in a discussion: namely, individuals already part of the group make predictions on the opinions the newcomers may express. In the human case, different individuals exhibit different opinions, leading to a plethora of information. Similarly, in the algorithm each SSRM in the ensemble provides its own prediction about the unlabeled speech sequences. To combine these predictions we implemented a consensus rule: only responses exhibiting an average pair-wise concordance with the majority (high consensus) are retained and averaged to provide the final response (Machine Labeled Single Speaker Regression Model, ML-SSRM; situation C in Fig 1). We call this algorithm the Cooperative Regression Model (CRM) (Fig 1, central blue arrows). In addition to select the SSRMs to be averaged, consensus is used here also to estimate reliability of the description assigned by the system to each unlabeled speech sequence. Inclusion criteria play a crucial role in the algorithm because new sequences—once labeled—are eventually added to the original pool. While this step ensures that the classifier’s capabilities increase with dimensions of the pool, it also requires a procedure that prevents insertion of unreliable sequences as well as that of sequences whose affective content emulates that of speakers already in the pool at the time of acquisition (redundant speakers). In fact, inclusion in the pool of a large number of closely similar sequences would produce (in the algorithm) a problem similar to that described for human groups: when conversations are dominated by a limited number of opinionated individuals, ‘group intelligence’ is largely reduced [24–26]. To minimize this risk, we introduced in the algorithm a redundancy exclusion criterion (REDundancy EXclusion, RED-EX; Fig 1 top right part) aimed at keeping the system free from self-referencing behavior. Specifically, in the model unlabeled speakers that meet the RED-EX criterion will be automatically excluded from the pool (redundant speaker, situation B in Fig 1). Finally, to avoid unproductive iterations we implemented selection of an optimal window width to calculate consensus level and added a dynamic consensus threshold for each window (i.e. the level of concordance sufficient to provide a reliable prediction is adaptively fixed based on intrinsic variation of the input data and related emotional content). That is, we simulated the groups’ choice of fixing a time limit for the debate and the fluctuations occurring in the members’ opinions as long as novel information becomes available. In the following sections, we describe results from three tests performed on speech samples in order to compare performance of the present SER system to that of standard strategies. Results are discussed in terms of their theoretical and applicative implications. Indeed, it is easy to imagine one such system in laboratory settings (e.g. to match stimulus’ presentation to the speaker’s emotional state) or implemented in smartphones/tablets as an aid for patients unable to decode speakers’ emotional tone, or to facilitate activity of the numerous helplines. Materials and Methods Dataset All simulations were performed on speech samples extracted from a public database (REmote COLlaborative and Affective interactions, RECOLA [27]). RECOLA is a speech corpus that was collected during a videoconference and contains spontaneous interactions between dyads of participants involved in a collaborative task [27]. The database includes 9.5h of speech recordings, collected continuously and synchronously from 46 participants. All collected speeches were in French, although participants differed as to their mother tongue (17 French, 3 Germans, 3 Italians). The total naturalness of speech recorded, makes RECOLA extremely suitable for our tests. As reported by RECOLA authors [27] twenty-three speakers gave informed consent to share their data and make their speech samples publicly available (10 male—13 female; mean age 21.3 ± 4.1 years). Only these samples were used in the present study. The non-consecutive numeric labels used to identify each speaker in the present work originate from the RECOLA dataset and are due to the fact that not all original speakers gave consent to use their multimodal emotional data. Emotional ratings for the RECOLA database were performed by six French-speaking assistants (3 females) for the first five minutes of all recorded sequences using a time-continuous annotation system for each affective dimension (ANNEMO web-based annotation tool, available at http://diuf.unifr.ch/diva/recola/annemo; for details see also Ringeval and co-workers [27]). As these were the only data made publicly available, these time periods were used here for training and testing. Metric One of the crucial concepts in the developed strategy is concordance. Hence we firstly provide a quantitative description of a metric suitable to represent the concordance level between two sequences. In particular, given two time series y1(t) and y2(t), having mean values μy1 and μy2, and standard deviations σy1 and σy2, the Concordance Correlation Coefficient (CCC) [28] is computed as follows: CCC=2 ρy1y2σy1σy2[σ2y1+σ2y2+(μy1−μy2)2](1) This metric accounts for biasing and signal correlation simultaneously so that it globally provides an estimation of temporal agreement. The CCC ranges between values of -1 (perfect negative agreement) and 1 (perfect positive agreement). A value of 0 corresponds to no agreement. The CCC will be used both in the consensus rule indicated in Fig 1 as well as in the system assessment. Procedure All tests were carried out on a personal computer with an Intel Quad Core i7 processor using the Matlab environment. When working, the algorithm occupies less than 1% of CPU usage and carries out the classification procedure in less then 1sec. A schematic description of the method is shown in Fig 1. The new semi-supervised approach presented here is structured into three main blocks: i) construction of single SSRM using the labeled speech sequences, ii) implementation of a consensus strategy to derive machine-labels for each unlabeled sequence, and iii) enlargement of the pool with new models trained on machine labeled sequences. The consensus rule is in turn composed by three minor steps: ii-a) application of a cooperative aggregation rule for adaptively and dynamically averaging the responses provided by each SSRM, ii-b) construction of the Machine Labeled Single Speaker Regression Model (ML-SSRM), and ii-c) implementation of a REDundancy EXclusion (RED-EX) criterion to decide whether or not including the ML-SSRM in the pool. According to this formulation, each speech sequence can be placed in one of three possible states: labeled, unlabeled, and waiting for inclusion. Labeled Speech sequences (LS) are those for which a reliable manual annotation is available. Unlabeled Speech sequences (US) are those that must be predicted, and for which machine-labels are estimated. Waiting-For-Inclusion (WFI) speech sequences are those for which machine-labels have been estimated and a model is trained but no decision about their inclusion has been taken. WFI sequences are subjected to the RED-EX criterion before a decision is reached. In the following sections, we detail each step of the method. Step 1—Construction of single speaker regression model (SSRM) Each single SSRM is trained and optimized on a labeled speech sequence yL(t) and the corresponding acoustic speech feature matrix X(t). Based on previous work [29], we consider the 65 acoustic low-level descriptors (LLD) and their first order derivatives (producing 130 LLD in total) that were used in the last two INTERSPEECH Computational Paralinguistic challengEs (ComParE 2013–2014) [30–31] using the open source extractor openSMILE (release 2.0) [32]. This feature-set includes a group of 4 energy related LLD, 55 spectral related LLD, and 6 voicing related LLD providing a description of speech in time-frequency as well as in voice quality domain (for more details on the ComParE feature set, refer to [33–34]). Acoustic model learning of emotion requires estimation of a gold standard from time-continuous dimensions. Here, the gold standard is extracted from the available annotations using a concordance-based weighted average procedure that averages the six annotations after subtracting their weighted mean values. Weights used in the mean centering are pairwise average correlation coefficient of each annotation with the remaining ones. Optimization of each SSRM is performed by identifying the most significant features using a new quadrant-based procedure inspired by Russell’s circumplex model of affect [35–36], which describes emotions along a continuous two-dimensional domain in terms of valence and arousal. Based on what suggested for facial expressions [36–37], a dimensional model—which plots emotions along a continuous space—was preferred to a categorical one because it was more likely to accommodate the fluctuations of the human voice as well as its continuous presentation. Methodologically, the two-dimensional model applied here thus looked better fit to correlate acoustic features with valence and arousal using an independent procedure for each quadrant. On this basis, features-selection (based on the Correlation Features Selection criterion) and gold standard synchronization (based on the estimation of the annotators’ reaction lag for the labelled speech sequences) are both implemented quadrant by quadrant. Features selected in segments of negative and positive valence (and similarly for arousal) are then concatenated so that a unique average reaction lag is computed for the whole output range and can be used for gold standard synchronization. A Partial Linear Regression model is trained using the resulting optimized setting, producing the corresponding SSRM for each labeled speech sequence. The procedure is repeated for each speaker in the initial pool. Mathematical details of the approach can be found in the S1 File. Step 2—Construction of cooperative regression model (CRM) In order to test an unlabeled speech sequence (US), a cooperative aggregation rule is applied to the responses provided by each SSRM, performing steps 1–5 illustrated in Fig 2. In step 1 a dynamic windowing is applied to the available predictions, for a certain number of window widths: L1, …, LN. In step 2 the average pairwise concordance of each response with the others estimated in the same window is computed. In step 3, in order to select the most concordant responses among the available ones, a threshold is selected to maximize average concordance of the subset of responses with a CCC average over the threshold, minus the average concordance of the remaining subset of responses (i.e., this quantity can be seen as the sum of the average concordance of a group and the average disagreement of the remaining ones). Hence, step 3 produces the average CCC of concordant responses for each window. In step 4, the optimal window width is selected as the one presenting the maximum average CCC across all the tested window widths. Here, we considered window widths in the range [1s–8s] with a step size of 1s. 10.1371/journal.pone.0161752.g002Fig 2 Graphical illustration of Steps 1–5. A schematic description of how the dynamic and adaptive cooperative strategy is implemented. Color legend as follows: for STEPS 1 to 4 the red, yellow, brown, and purple curves describe four synthetic predictions of a given output. For STEP 5, the cyan curves indicate the two concordant predictions. The black curves represent the less concordant prediction. Only the cyan curves are averaged to construct the unique final linear prediction (represented by the red segment superimposed on the rightmost part of the chart). Finally, in step 5, provided the optimal window and the optimal threshold for that window, a linear fitting of the concordant responses over a time interval of 1s is used to extrapolate the predicted value at time tn+1 (identified by the red dot in the bottom panel of Fig 2). The procedure is repeated at each time instant of the entire observation period. If the optimal concordance is negative, the prediction is not computed and the output machine label at that time is missing. Note that the model for the speaker under test is constructed only if the number of available predictions is sufficiently high; else the speaker is excluded from the pool (and set to the EXCLUDED state, Fig 1). In all other cases, the extracted machine labels are used to construct the corresponding SSRM following the procedure described in Step 1 and in the S1 File. We denote each of these models as Machine Labeled Single Speaker Regression Model (ML-SSRM). These models are set to the WFI state, meaning that they are reliable but not yet able to significantly improve system performance. Step 3—Redundancy exclusion criterion (RED-EX) In order to autonomously decide whether or not to add the ML-SSRM of the WFI sequence to the pool, the system evaluates the opportunity to increase the overall prediction capability of the system, while avoiding redundancy with respect to the models already in the pool. This is done by the REDundancy EXclusion (RED-EX) criterion, which compares the prediction of the model constructed on the WFI sequence with that provided by the SSRMs in the pool, using a further unlabeled speech sequence for testing. If the global concordance of the responses provided by the ML-SSRM with at least one response provided by the SSRMs in the pool is higher than 0.99 then the ML-SSRM does not add improvement to the pool. Therefore, the model is not included and set in the state EXCLUDED (Fig 1, situation B). Conversely, if the maximum concordance with the responses provided by the pool is lower than 0.99 then the new ML-SSRM is set to the state INCLUDED (Fig 1) and added to the pool. The procedure described in Step 2 is then applied to the enlarged ensemble to predict the affective content of the additional unlabeled speech sequence. The RED-EX criterion thus reduces the risk that speech samples that are too similar to those already present into the pool (i.e. samples which are unlikely to provide novel information) are included. In this way, biases in the labeling procedure that may result from poorly diversified information are prevented, and the computational burden is kept minimal. Tests 1–3 In order to test the performance of the proposed strategy in a naturalistic speech environment and demonstrate its capabilities in larger scenarios, we performed three tests. In Test 1 performance of the present semi-supervised learning strategy was compared to that of a standard supervised approach using an identical initial set of labeled speech sequences. In Test 2 robustness of the algorithm was tested with respect to variations in entering order of the different speakers. In Test 3 sensitivity of the approach to self-referencing behavior was assessed by repeatedly providing the same speech sequence as input to the system. To run these three simulations, we divided the RECOLA [27] dataset into two distinct sets, one for the construction of the SSRM of the initial pool and one for the testing of the semi-supervised regression strategy. In line with the assumption that in semi-supervised approach there is a small amount of labeled data and a relatively larger amount of unlabeled data, we randomly select a pool of four speakers from RECOLA. Using a “leave one speaker out” cross-validation strategy, we extracted the frequency of inclusion of each speaker in the training set for the prediction of the speaker in test. Based on the results, we selected speakers P23 (F), P30 (F), P43 (F), and P65 (M) that were averagely included (labels in parenthesis indicate gender). Moreover, based on preliminary simulation results, we eliminated speakers P16 (M), P17 (M), P34 (M), and P62 (M), whose features presented intrinsic problems that caused them to be rarely included in testing and to exhibit low prediction performance (due to a general emotional flatness in their speech). The remaining 15 speakers were used to validate the proposed semi-supervised strategy. Performance was evaluated using the CCC metric defined above in order to quantify the discrepancy between expected and estimated responses for each speech sequence in test either in terms of signal correlation or in terms of mean square error. Boxplots of the CCC obtained over the 15 speakers in test are used to visualize the results, and when needed, t-test were performed to demonstrate the statistical significance of improvement achieved using the proposed strategy vs. the application of supervised learning strategy. Alpha level was set at .05 for all tests. For the sake of clarity, details for each of the three tests are provided in the corresponding Results sections. Results Test 1: Comparison of semi-supervised and supervised strategies We compared the present model and a standard supervised approach by feeding them both with the same set of pre-labeled data. The supervised approach is assumed to systematically rely on the existing data pool. Accordingly, it is expected to be stably accurate but strongly dependent on the quality of labeled data. In other words, it mimics the behavior of a very conservative group of people, which—if prejudiced—may bias final results. Conversely, the semi-supervised method allows for on-line inclusion of any newly generated machine-labeled models, resembling more to a liberal group, open to novel opinions. Hence it should be more dynamic and explorative, looking for patterns that were not previously anticipated (but that might also result uninteresting). As already mentioned, for the simulation, we used the annotated speech corpus RECOLA [27] and compared predictions on the emotional content of 15 test speakers as obtained by the two approaches. Acoustic emotional features were identified according to a quadrant-based procedure inspired by Russell’s circumplex model of affect [35–36] (see also S1 File), which represents emotions along a continuous two-dimensional domain in terms of valence and arousal. Performance of the two approaches was separately evaluated for arousal and valence in terms of Concordance Correlation Coefficient (CCC) [28], a type of metric that accounts for biasing and signal correlation simultaneously, providing an estimation of temporal agreement. The boxplot of CCC values obtained for arousal (left) and valence (right) are shown in Fig 3, together with p-values for the related t-tests. In the arousal dimension, the semi-supervised method strongly increased the CCC values (median ± interquartile range: 0.59±0.17) compared to the supervised system (0.52±0.33, t = 2.377, df = 14, p < .03), with variations up to 0.88. The same was true for the valence dimension, in spite of the lower values of CCC obtained (semi-supervised: 0.16±0.19; supervised: 0.08±0.10, t = 2.985, df = 14, p < .01). Namely, by allowing for on-line inclusion of additional information in the form of the newly generated machine-labeled models, the semi-supervised approach increased the sampling pool and significantly enhanced its predictive capabilities compared to the standard supervised model. A large set of accurately machine-labeled items, i.e. a ‘wise crowd’, led to a significantly improved performance. 10.1371/journal.pone.0161752.g003Fig 3 Box-plot of the concordant correlation coefficient (CCC) values for the semi-supervised strategy and supervised strategy. Results relating to arousal (left) and valence (right) were independently obtained. Test 2: Testing robustness to rearrangement of speakers in testing Behavioral studies show that when competing options are presented in a sequence, order of appearance affects final evaluation. For example, when a jury evaluates a series of candidates, ratings increase with their serial position [38–39]. In the present model, order of testing could influence final performance due to the role played in the method by concordance (between responses provided by the pool) and redundancy (implemented by the RED-EX criterion, see also Test 3). To assess robustness of the model with respect to this issue, we quantified dispersion of the obtained CCC values when order of testing repeatedly changed in different iterated simulations (S1 Fig). Ten different simulations were run by randomly rearranging order of speakers in testing and computing CCC values for each speech sequence in test when it appeared in a different position in the test sequence. Order of the sequences chosen at each iteration was the same for arousal and valence. As in Test 1, we compared performance under semi-supervised and supervised learning. Fig 4 shows the boxplot of the median CCC values computed over the 15 speech sequences in test for the 10 iterations for arousal (left) and valence (right). A significant advantage was found for the semi-supervised compared to the supervised machine (arousal: t = 2.686, df = 14, p < .02; valence: t = 3.704, df = 14, p < .002). The effect was stronger for the valence compared to the arousal dimension. For both dimensions, predictions obtained under the semi-supervised method were less permeable to order effects and maintained better performances when compared to supervised strategy. 10.1371/journal.pone.0161752.g004Fig 4 Effects of rearrangement of speakers’ order. Box-plot of the median concordant correlation coefficient (CCC) computed over the 15 speech sequences in test achieved during the 10 iterations for arousal (left) and valence (right) comparing semi-supervised and supervised strategies. P-values of the t-test (arousal: p < .02; valence: p < .002) demonstrate that the semi-supervised strategy significantly improves general performance of the emotion recognition system. Test 3: Testing robustness to inclusion of redundant speech sequences Groupthink [24–25] is a phenomenon described in the social domain whereby failures in collective decision-making occur as a consequence of excessive cohesion within the group. This principle was nicely stated by social psychology in the 70s with reference to the Bay of Pigs invasion: the strong cohesion of the White House members either limited the possibility for alternative opinions or censored them, leading to a less objective (and fruitful) evaluation [24–25]. In the present model a groupthink analog can arise when an excessively narrow range of inputs is provided or when samples are too similar to one another (e.g. a speech sequence is inserted whose affective content emulates that of a speaker already in the pool). This could lead to self-referencing behavior and failure to generate correct labeling. To explore this issue, efficacy of the ad-hoc inclusion criterion, the RED-EX criterion, in preventing degeneration was tested as follows (S2 Fig). A series of simulations were run in which the system was repeatedly fed with a single speech sequence. The effects of enabling/disabling the RED-EX criterion were compared. Fig 5 reports CCC values of the prediction achieved for the speech sequence in repeated tests with respect to its gold standard (P20 in this case) for arousal (left) and valence (right), comparing the case when RED-EX criterion was applied (blue lines) or not applied (red lines). Numbers indicate cardinality of the pool at each iteration for the two conditions. Higher CCC values of predictions were obtained with the reduced pool (five models in both dimensions). Importantly, when the RED-EX criterion was not used, performance worsened during the repeated input of the same speaker (“self-referencing”). Conversely, when the RED-EX criterion was used, after an initial transient, performance remained stable around an optimal value, as desired. Indeed, a significant difference emerged for predictions relative to both arousal (t = 5.405, df = 9, p < .0004) and valence (t = 4.433, df = 9, p < .002) when the RED-EX criterion was used compared to when it was excluded. This confirms that the RED-EX criterion prevents the exponential increasing of models in the pool (which may occur during semi-supervised learning) by providing a guiding principle that selects which model to include for improving the system’s prediction capability. 10.1371/journal.pone.0161752.g005Fig 5 Effects of inclusion of redundant speech sequences. Comparison of CCC values for the prediction of arousal (left) and valence (right) according to whether the semi-supervised strategy was used in combination with the RED-EX criterion (blue lines) or not (red lines) in the Test 3. Numbers indicate cardinality of the pool at each iteration. Discussion We describe a novel speech emotion recognition system that applies a semi-supervised prediction model based on consensus. The system allows both for assigning new labels to unlabeled sequences and for preventing self-referencing. This approach deeply departs from procedures like active learning [10–40] (which aims at identifying an optimal pool of labeled sequences among the available ones) and from self-training [20] (that aims at enhancing robustness of existing classifiers, embedding labeled and unlabeled data). Novelty of the method lies in the modular architecture of the cooperative system, its strong reliance on the concept of concordance and the dynamic application of consensus rule. On these respects, the model calls forth the mechanisms involved in human group decision-making, of which shares the many benefits, including a substantial increase in accuracy as a function of sample size (the so-called wisdom of the crowd effect)[17–18,26]. As in a real crowd, the present algorithm is improved by the addition of novel inputs, particularly when they significantly differ from those already contained in the pool. This would be in line with observations on the detrimental effects that ‘groupthink’ [24–25] plays in humans, and with recent findings showing that ‘group intelligence’ is largely reduced when conversations are dominated by a limited number of individuals [26]. Differently from human groups (in which the same individuals produce and classify the emotional content), the algorithm described here mainly act as a classifier (although the newly added speech sequences, i.e. the signal generators, when added to the pool become, they too, classifiers). Methodologically, the system largely differs from traditional approaches where the whole set of available data is sent as input to a prediction system (that embeds features extraction, reduction, and regression method). In these approaches, each addition to the training set is laborious, requiring re-training of the whole prediction system and, consequently, causing an exponential increase in both complexity and computational time. Conversely, the strategy proposed here is modular: accordingly, predictions over single speaker models are trained and labeled in parallel, benefiting from the advantages of cooperative learning. By simultaneously taking into account a plethora of opinions, the system increases its prediction capabilities and provides responses that go beyond mere averaging. To account for parallel processing, the present method strongly relies on the concept of concordance. On this respect it differs from previously described SER systems, which mainly depend on correlation. The main advantage is that concordance shares with correlation the capability to assess the statistical discrepancy between random processes or time series, but additionally accounts for biasing. Biasing is a known distortion effect produced by systematic contributions: in the case of speech emotion recognition, it may lead to disastrous consequences when independently manifested in the two correlated dimensions of valence and arousal. Based on the Russell’s two-dimensional depiction of affect assumed here [35], different biasing terms, independently occurring in valence and arousal, will produce a displacement in the circumplex diagram that may completely alter the predicted emotional state. To prevent this issue, the present method relies on concordance correlation to assess for a realistic comparison between predicted and expected emotional dimensions, i.e. on a metric that simultaneously evaluates discrepancy between two signals in terms of both biasing and correlation. Concordance is used here also to prevent undesirable behaviors and implement the consensus rule. Specifically, we anticipated the need for a redundancy exclusion criterion to avoid self-referencing of the prediction system. This rule automatically prevents inclusion of machine-labeled models with no additional capability to predict future speech sequences than the pool itself. In addition, concordance is used for the first time in the implementation of a consensus rule. Predictions provided by each SSRM are collected and dynamically averaged according to their mutual consensus, extending the logic of majority voting to the regression framework. At each time instant, the most concordant predictions over a temporal delayed interval are averaged; in contrast, outlier predictions are excluded. In such a way, consensus among predictions is evaluated accounting for their mutual biasing and statistical correlations. Such a dynamic application of the consensus rule proposes a more realistic and complete logic of dynamic sliding windowing: namely, a model in which duration of the temporal window devoted to concordance evaluation changes dynamically according to the predictions provided by the individual regression models. Accordingly, a high level of concordance can be reached at once (due for example to obviousness of perceived emotions or capability of the models involved) or following a long listening session. As a matter of fact, dynamicity provides a benchmark for implementing multimodal emotion recognition because it allows for the possibility that different communicative channels may require different temporal intervals to express their informative content. Central to our approach is the way semi-supervised strategy is implemented. During the past few years, semi-supervised learning appeared an unfeasible approach for a context as complex as that of emotion recognition. Difficulties in obtaining acceptable results with supervised strategies did not constitute, until now, a convincing starting point for experiencing semi-supervised strategies [41–42]. Hence the most recent attempts [20] investigated semi-supervised learning approaches mainly as a mean for optimizing performance and reducing the amount of human annotation via machine labeling. Our approach takes advantage from a restricted pool of labeled speech sequences to build a single regression model for each sequence in the pool. These models are now ready to predict in a dynamic cooperative way the emotional content of any new speech sequence. The final prediction, when meeting with a sufficiently high level of concordance, is deemed reliable to apply machine-labels to the novel (semi-supervised) sequences. Subsequently, the new models constructed on the machine-labeled sequences are fed to the initial pool. With this respect, the system is speaker-independent, as shown by the low dependency of the labeling processes from the order of testing used for presentation of new speech sequences to the pool. To conclude, we present an entirely novel approach to the SER problem, which benefits from a slender and fast-processing architecture, and is enhanced rather than encumbered by increased sampling. At present, classification is provided only on a dimensional model, but we are confident that these data could be next used—if appropriately treated—to fit the framework of discrete emotion theories (such as Ekman’s)[43]. In addition, although the algorithm now exclusively targets speech, we expect that it could be eventually incorporated into a multimodal emotion prediction system, i.e. one that uses video, speech and physiological signals (such as ECG and/or EDA) that could extend and enrich prediction capabilities. For its characteristics, the system is likely to fit the demands of a number of situations, including use of portable voice analyzers, an advantage that could be of interest to cognitive science and rehabilitation. As to the former, application of this method to speech analysis approaches could be valuable in studies on patients for whom affective disturbances of speech are known, such as autism spectrum disorder [3–4] or Parkinson Disease [5–7]. In addition, fast processing qualifies this algorithm as a possible method for automatically pacing stimulus presentation and/or selecting data collection based on the speakers’ emotional tone, granting novel possibilities to neurophysiological and neuroimaging research. As to translational research, the opportunity to implement this SER system into a smartphone or tablet would provide a valuable aid to affective rehabilitation. Training sessions in which feedback is given by the system as to whether prosody and/or affective tone is adequate to a presented context would represent a useful tool for at-home therapy in the case of neurologic patients and an excellent alternative to the human-to-human interaction in autistic individuals. Finally, being devised as a simile to human behavior, this method could stimulate development of novel frameworks for simulating human behavior in contexts of consensus seeking, stereotypes construction and group decision-making (e.g. the “like” and viral effects in social networks and similar phenomena). Supporting Information S1 Fig Schematic illustration of Test 2 to test the robustness to rearrangement of the order of speech sequences in testing. The same set of three sequences (on the left) are repeatedly rearranged and input to the semi-supervised learning machine based on the cooperative regression and RED-EX criterion. The performance in terms of CCC values for the comparative simulations are reported in the main text. (DOCX) Click here for additional data file. S2 Fig Schematic representation of the test run to demonstrate the effectiveness of the RED-EX criterion to prevent self-referencing. The cooperative regression system is fed with an unlabeled speech sequence (green circle on the left) that is already in the pool of labeled speakers (feed 1). The cooperative regression module applies on it, generate the M-labeled speech sequence, but then put it in the WFI condition (gray circle on the right). At this point the RED-EX criterion applies and evaluates if the inclusion of the new annotated speech sequence may add improvement to the system knowledge-base by computing its CCC with the speech sequences already in the pool. A too high CCC value makes the RED-EX criterion been verified and the sequence excluded (red arrow in the bottom-left). The feed is repeated (feed 2) with the same input speech sequence but without the application of the RED-EX criterion. In such case, after been machine-labeled the sequence is included in the cooperative model. To emphasize the effect, the two kind of simulation are repeated for 10 times each. (DOCX) Click here for additional data file. S1 File Supplementary Methods. In this Section, we provide mathematical details of the main steps involved in the construction of each Single Speaker Regression Model (SSRM). (DOCX) Click here for additional data file. The authors are grateful to Dr A Sirigu for helpful suggestions during preparation of the manuscript. ==== Refs References 1 Zeng Z , Pantic M , Roisman GI , Huang TS . A survey of affect recognition methods: Audio, visual, and spontaneous expressions, IEEE Trans . Pattern Anal. Mach. Intell . 2009 ; 31 : 39 –58 . 2 El Ayadi ME , Kamel M , Karray F . Survey on Speech Emotion Recognition: Features, Classification Schemes, and Databases , Pattern Recogn . 2011 ; 44 : 572 –587 . 3 Stewart ME , McAdam C , Ota M , Peppé S , Cleland J . Emotional recognition in autism spectrum conditions from voices and faces . Autism . 2013 ; 17 : 6 –14 . 10.1177/1362361311424572 23045218 4 Fan YT , Cheng Y . Atypical mismatch negativity in response to emotional voices in people with autism spectrum conditions . PLoS One . 2014 ; 9 (7 ): e102471 10.1371/journal.pone.0102471 25036143 5 Ariatti A , Benuzzi F , Nichelli P . Recognition of emotions from visual and prosodic cues in Parkinson's disease . Neurol Sci . 2008 ; 29 : 219 –27 . 10.1007/s10072-008-0971-9 18810595 6 Dara C , Monetta L , Pell MD . Vocal emotion processing in Parkinson's disease: reduced sensitivity to negative emotions . Brain Res . 2008 ; 1188 :100 –111 . 18022608 7 Ventura MI , Baynes K , Sigvardt KA , Unruh AM , Acklin SS , Kirsch HE , et al Hemispheric asymmetries and prosodic emotion recognition deficits in Parkinson's disease . Neuropsychologia . 2012 ; 50 :1936 –45 . 10.1016/j.neuropsychologia.2012.04.018 22564479 8 Eldar E , Yael N . Interaction between emotional state and learning underlies mood instability . Nat Commun . 2015 ; 1 21 ;6 :6149 10.1038/ncomms7149 25608088 9 Cowie R , Cornelius R . Describing the emotional states that are expressed in speech , Speech Commun . 2003 ; 40 : 5 –32 . 10 Settles B . Active learning literature survey , Dept. of Comput. Sci., Univ. of Wisconsin–Madison , Madison, WI, USA , Tech. Rep. 2009 . 11 Tur G , Hakkani-Tür D , Schapire RE . Combining active and semi-supervised learning for spoken language understanding . Speech Commun . 2005 ; 45 : 171 –186 . 12 Mencattini A , Martinelli E , Costantini G , Todisco M , Basile B , Bozzali M , et al Speech emotion recognition using amplitude modulation parameters and a combined feature selection procedure . Knowl-Based Syst . 2014 ; 63 : 68 –81 . 13 Mencattini A , Martinelli E , Ringeval F , Schuller B , Di Natale C . Continuous Estimation of Emotions in Speech by Dynamic Cooperative Speaker Models , IEEE T Affect Comput . In press; 10.1109/TAFFC.2016.2531664 14 Alghowinem S , Goecke R , Wagner M , Epps J , Breakspear M , Parke G . From joyous to clinically depressed: Mood detection using spontaneous speech In Proc. of the 25th Inter. Florida Artificial Intelligence Research Society Conf. (FLAIRS) , Marco Island FL 2012 : 141 –146 . 15 Ooi KEB , Lech M , Allen NB . Prediction of major depression in adolescents using an optimized multi-channel weighted speech classification system Biomed Signal Process Control . 2014 ; 14 : 228 –239 . 16 Marchi E , Ringeval F , Schuller B . Perspectives and Limitations of Voice-controlled Assistive Robots for Young Individuals with Autism Spectrum Condition In Neustein A , Markowitz J , editors. Speech and Automata in the Healthcare Environment . Berlin : De Gruyter ; 2014 . 17 Galton F . Vox populi . Nature . 1907 ; 75 : 450 –1 . 18 Surowiecki J . The Wisdom of Crowds . Anchor Books ; 2005 . 19 Bernstein J , Long JS , Veillette C , Ahn J . Crowd intelligence for the classification of fractures and beyond . PLoS One . 2011 ;6 (11 ):e27620 10.1371/journal.pone.0027620 22132118 20 Zhang Z , Coutinho E , Deng J , Schuller B . Cooperative Learning and its Application to Emotion Recognition from Speech , IEEE T Audio Speech . 2015 ; 3 : 115 –126 . 21 Martinelli E , Magna G , Vergara A , Di Natale C . Cooperative classifiers for reconfigurable sensor arrays . Sens. Actuators B-Chem . 2014 ; 199 : 83 –92 . 22 Magna G , Casti P , Jayaraman SV , Salmeri M , Mencattini A , Martinelli E , et al Identification of mammography anomalies for breast cancer detection by an ensemble of classification models based on artificial immune system . Knowl-Based Syst . 2016 ; 101 : 60 –70 . 23 Davis JH . Group decision and social interaction: A theory of social decision schemes . Psychol Rev . 1973 ; 80 : 97 –125 . 24 Janis IL . Victims of Groupthink . 1972 ; New York : Houghton Mifflin . 25 Janis IL . Groupthink: Psychological study of foreign-policy decisions and fiascos . 1982 ; New York : Houghton Mifflin . 26 Woolley AW , Chabris CF , Pentland A , Hashmi N , Malone TW . Evidence for a collective intelligence factor in the performance of human groups . Science . 2010 ; 330 : 686 –8 . 10.1126/science.1193147 20929725 27 Ringeval F, Sonderegger A, Sauer J, Lalanne D. Introducing the RECOLA multimodal corpus of remote collaborative and affective interactions, in Proc. of Face & Gestures 2013, 2nd IEEE Inter. Workshop on Emotion Representation, Analysis and Synthesis in Continuous Time and Space (EmoSPACE); 2013. 28 Lawrence I , Kuei L . A concordance correlation coefficient to evaluate reproducibility . Biometrics . 1989 ; 45 : 255 –268 . 2720055 29 Ringeval F , Eyben F , Kroupi E , Yuce A , Thiran J-P , Ebrahimi T , et al Prediction of asynchronous dimensional emotion ratings from audiovisual and physiological data , Pattern Recogn Lett . 2015 ; 66 : 22 –30 . 30 Schuller B, Steidl S, Batliner A, Kim SK. The Interspeech 2013 computational paralinguistics challenge: social signals, conflict, emotion, autism. In Proc. of INTERSPEECH 2013, 14th Annual Conf. of the InterSpeech Communication Association (ISCA), Lyon 2013. 31 Schuller B, Steidl S, Batliner A, Epps J, Eyben F, Ringeval F, et al. The Interspeech 2014 computational paralinguistics challenge: Cognitive & physical load. In Proc. of INTERSPEECH 2014, 15th Annual Conf. of the Inter. Speech Communication Association (ISCA), Singapore. 2014. 32 Eyben F, Weninger F, Gross F, Schuller B. Recent developments in openSMILE, the Munich open-source multimedia feature extractor. In Proc. ACM Multimedia (MM), Barcelona. 2013. 33 Weninger F , Eyben F , Schuller BW , Mortillaro M , Scherer KR . On the acoustics of emotion in audio: What speech, music and sound have in common . Front Psychol . 2013 ; 4 : 292 10.3389/fpsyg.2013.00292 23750144 34 Eyben F. Real-time speech and music classification by large audio feature space extraction, Ph.D. dissertation, Technische Universität München, 2014, in press. 35 Russell JA . A circumplex model of affect . J Pers Soc Psychol . 1980 ; 39 : 1161 –1178 . 36 Posner J , Russell J , Peterson B . The circumplex model of affect: an integrative approach to affective neuroscience, cognitive development, and psychopathology , Dev Psychopathol . 2005 ; 17 : 715 –734 . 16262989 37 Colibazzi T , Posner J , Wang Z , Gorman D , Gerber A , Yu S , et al Neural systems subserving valence and arousal during the experience of induced emotions . Emotion . 2010 ; 10 :377 –89 . 10.1037/a0018484 20515226 38 Bruine de Bruin W . Save the last dance for me: unwanted serial position effects in jury evaluations . Acta Psychol (Amst) . 2005 ; 118 : 245 –60 .15698823 39 Bruine de Bruin W . Save the last dance II: unwanted serial position effects in figure skating judgments . Acta Psychol (Amst) . 2006 ; 123 : 299 –311 .16542632 40 Chapelle O , Schölkopf B , Zien A . Semi-Supervised Learning . Cambridge : The MIT Press ; 2006 . 41 Schuller B , Batliner A , Steidl S , Seppi D . Recognising realistic emotions and affect in speech: State of the art and lessons learnt from the first challenge , Speech Commun . 2011 ; 53 : 1062 –1087 . 42 Vogt T , André E , Wagner J . Automatic recognition of emotions from speech: a review of the literature and recommendations for practical realisation In Peter C , Beale R , editors. Affect and emotion in human–computer interaction . Berlin : Springer ; 2008 pp. 75 –91 . 43 Ekman P . An argument for basic emotions . Cogn Emot . 1992 ; 6 : 169 –200 .

PMC005xxxxxx/PMC5001725.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756470510.1371/journal.pone.0161686PONE-D-16-00276Research ArticlePhysical sciencesChemistryChemical compoundsOrganic compoundsVitaminsVitamin DPhysical sciencesChemistryOrganic chemistryOrganic compoundsVitaminsVitamin DMedicine and Health SciencesPublic and Occupational HealthPhysical ActivityBiology and Life SciencesPhysiologyPhysiological ParametersBody WeightBody Mass IndexMedicine and Health SciencesPhysiologyPhysiological ParametersBody WeightBody Mass IndexPeople and PlacesPopulation GroupingsEthnicitiesMexican PeoplePhysical SciencesChemistryChemical CompoundsOrganic CompoundsVitaminsPhysical SciencesChemistryOrganic ChemistryOrganic CompoundsVitaminsMedicine and Health SciencesEndocrinologyEndocrine PhysiologyMenstrual CycleMenarcheBiology and Life SciencesPhysiologyEndocrine PhysiologyMenstrual CycleMenarcheMedicine and Health SciencesPhysiologyEndocrine PhysiologyMenstrual CycleMenarcheBiology and Life SciencesPhysiologyReproductive PhysiologyMenstrual CycleMenarcheMedicine and Health SciencesPhysiologyReproductive PhysiologyMenstrual CycleMenarcheMedicine and Health SciencesDiagnostic MedicineDiagnostic RadiologyMammographyResearch and Analysis MethodsImaging TechniquesDiagnostic RadiologyMammographyMedicine and Health SciencesRadiology and ImagingDiagnostic RadiologyMammographyResearch and Analysis MethodsMathematical and Statistical TechniquesStatistical MethodsRegression AnalysisLinear Regression AnalysisPhysical SciencesMathematicsStatistics (Mathematics)Statistical MethodsRegression AnalysisLinear Regression AnalysisSerum 25–Hydroxyvitamin D3 and Mammography Density among Mexican Women 25–Hydroxyvitamin D3 and Mammography DensityAmadou Amina 1Biessy Carine 1Rinaldi Sabina 1Fedirko Veronika 2Assi Nada 1Lajous Martin 34Ortiz-Panozo Eduardo 3Yunes Elsa 3Lopez-Ridaura Ruy 3Torres-Mejia Gabriela 3Romieu Isabelle 13*1 Nutrition and Metabolism Section, International Agency for Research on Cancer (IARC), Lyon, France2 Department of Epidemiology, Rollins School of Public Health, Winship Cancer Institute, Emory University, Atlanta, GA, United States of America3 Center for Research on Population Health, National Institute of Public Health, Cuernavaca, Mexico4 Department of Epidemiology, Harvard School of Public Health, Boston, MA, United States of AmericaSingh Brij EditorUniversity of North Dakota School of Medicine and Health Sciences, UNITED STATESCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: AA SR ML EO EY RL GT IR. Data curation: ML EO EY RL GT IR. Formal analysis: AA CB IR VF NA. Funding acquisition: IR. Investigation: AA IR SR. Methodology: AA CB IR VF NA. Resources: AA IR. Supervision: IR. Validation: AA CB IR. Writing – original draft: AA SR IR. Writing – review & editing: AA SR IR CR CB VF NA ML EO EY RL GT. * E-mail: romieui@iarc.fr26 8 2016 2016 11 8 e01616865 1 2016 10 8 2016 © 2016 Amadou et al2016Amadou et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Low circulating levels of vitamin D and high mammographic density (MD) have been associated with higher risk of breast cancer. Although some evidence suggested an inverse association between circulating vitamin D and MD, no studies have investigated this association among Mexican women. We examined whether serum 25−hydroxyvitamin D3 [25(OH)D3] levels were associated with MD in a cross-sectional study nested within the large Mexican Teacher's Cohort. This study included 491 premenopausal women with a mean age of 42.9 years. Serum 25(OH)D3 levels were measured by liquid chromatography/tandem mass spectrometry. Linear regression and non-linear adjusted models were used to estimate the association of MD with serum 25(OH)D3. Median serum 25(OH)D3 level was 27.3 (23.3–32.8) (ng/ml). Forty one (8%) women had 25(OH)D3 levels in the deficient range (< 20 ng/ml). Body mass index (BMI) and total physical activity were significantly correlated with 25(OH)D3 (r = −0.109, P = 0.019 and r = 0.095, P = 0.003, respectively). In the multivariable linear regression, no significant association was observed between 25(OH)D3 levels and MD overall. However, in stratified analyses, higher serum 25(OH)D3 levels (≥27.3 ng/ml) were significantly inversely associated with percent MD among women with BMI below the median (β = −0.52, P = 0.047). Although no significant association was observed between serum 25(OH)D3 and percent MD in the overall population, specific subgroups of women may benefit from higher serum 25(OH)D3 levels. Consejo Nacional de Ciencia y Tecnologia (CONACYT)115312American Institute for Cancer Research (AICR)10A035Romieu Isabelle Ministry of Health of Mexico, AVON, BanorteThis work was supported by the American Institute for Cancer Research (AICR), grant number 10A035; Consejo Nacional de Ciencia y Tecnologia (CONACYT), grant number 115312; and Ministry of Health of Mexico, AVON, Banorte. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Mammographic density (MD) has been identified among the strongest predictors of breast cancer (BC) risk. Women having more than 75% of dense tissue have 4 to 6 times greater risk of BC compared to women with little dense tissue [1–3]. MD represents the dense tissue of the breast, and is expressed as a percentage. Women with high MD have more proliferating epithelial tissue and connective tissue relative to women with low MD [4, 5]. Proliferating cells are more vulnerable to genetic damage, thereby increasing BC risk [6]. MD is correlated with several BC risk factors including age, anthropometry, reproductive, genetic, and hormonal factors, diet, and circulating micronutrients [5, 7], including low vitamin D levels [8–10]. Vitamin D is a fat-soluble vitamin that is naturally present in very few foods, but is available as dietary supplements, or produced in the skin in response to ultraviolet B (UVB) exposure [11, 12]. Vitamin D3 (cholecalciferol) is the precursor to the steroid hormone calcitriol, and it is activated to its active form by two cytochrome P450-mediated hydroxylation steps [13]. The first hydroxylation step mostly occurs in the liver to yield 25-hydroxvitamin D3 (25(OH)D3), which is catalyzed by the enzyme vitamin D-25-hydroxylase (predominantly CYP2R1). 25(OH)D3 is the circulating form of the hormone that is measured in the blood and clinically used to establish and monitor the vitamin D status of a patient. Circulating 25(OH)D3 is hydroxylated in the kidney by the cytochrome P450 enzyme CYP27B1 (1α-hydroxylase) to yield calcitriol [13, 14]. Vitamin D2 (ergocalciferol) is a form of vitamin D that is of plant origin, is derived from ergosterol and functions much like vitamin D3 but is less active [14]. Serum 25(OH)D (25(OH)D3 and 25(OH)D2) concentration is the major circulating form of vitamin D, and the best biomarker reflecting exposure to vitamin D from different sources, with a half-life of 2-3weeks [12, 15]. Studies have shown that vitamin D can promote apoptosis and cell differentiation, and inhibits breast cell proliferation that may have an effect on estrogen metabolism [14, 16]. There are a number of studies on the association between vitamin D (estimated from the diet, or measured as biomarker) and MD among both pre and postmenopausal women [8, 17–24]. Although some of them consistently reported significant inverse associations between vitamin D and MD [18–20, 24], the evidence of associations continue to be inconsistent [21–23, 25] and to this date, no study has investigated this association among Mexican women. It is hypothesized that the actions of vitamin D, calcium, insulin-like growth factor (IGF)-I, and IGF-binding protein-3 (IGFBP-3) on BC are interconnected [26]. IGF-I levels have been positively related to MD among premenopausal Canadian women while IGFBP-3 levels have been inversely related [27]. Both Epidemiological and molecular studies have shown that vitamin D, directly or indirectly, may inhibit IGF-I and enhance IGFBP-3 effects in breast tissue, that may reduce breast density and breast cancer risk [16, 28]. IGF1-stimulated cell growth was inhibited by vitamin D analogs and this effect was associated with increased release of IGF binding protein 3 (IGFBP3) [29]. Deeb et al demonstrated that vitamin D treatment can result in the upregulation of IGFBP3 and transforming growth factor‑β (TGF β)–SMAD3 signaling cascades and by downregulating the epidermal growth factor receptor (EGFR) signaling pathway [16]. Further evidence reported that, in breast tumours, vitamin D modulates the IGF-I/IGFBP ratio to decrease proliferation and increase apoptosis [30]. However, there is no data on these associations in the Hispanic population that might help better understand mechanisms by which vitamin D may affect MD and BC. To further clarify the role of vitamin D in MD, we examined whether serum 25(OH)D levels were associated with MD, and whether this association differed by obesity status in a cross-sectional study nested within the Mexican Teachers’ Cohort (MTC). A second objective was to explore the modifying effect and mediating role of IGF pathway (as IGF1 and IGFP3) in the association between 25(OH)D and MD. Materials and Methods Study population The Mexican Teachers’ Cohort (MTC) is a prospective study of 115,315 female teachers from 12 Mexican states aged 35 years who were invited to participate in a cohort study to evaluate lifestyle and chronic diseases. Detailed methods have been described elsewhere [31]. Information were obtained on socio-demographics, socio-economic status (SES), reproductive history, hormone contraceptive and menopausal hormone replacement therapy use, physical activity, alcohol consumption, smoking history, family history of breast cancer (FHBC), clinical history, and lifestyle (including a food frequency questionnaire). In 2007, a subsample of 2,084 MTC participants from two Mexican regions (Jalisco and Veracruz) participated in a clinical evaluation that included an interview, anthropometric measurements, mammogram and the collection of biological specimens. Fasting blood samples (~25 ml) were obtained through venipuncture by trained nurses. Selection of subjects Among the 2,084 participants who participated in the clinical sub-cohort, we excluded 230 women who had insufficient information on metabolic syndrome components (because of a parallel study on metabolic syndrome in the same population [32], 67 who had an unknown menopausal status and 624 who were postmenopausal at the time of their mammogram. Women were considered as pre-menopausal if they had menstruated at least once over the 12 months prior to the visit, and were considered as postmenopausal if they had no menstruation over the last 12 months prior to the visit, and those with surgical menopause who reported bilateral oophorectomy [33]. We then stratified women by 4 breast density categories: <10%, 10 to <25%, 25 to <50% and > = 50% [34]. Women were randomly selected from each group proportionally to its size. Thirty-five women were selected for the first group, 158 for the second, 247 for the third and 160 for the fourth group. Among those, 500 women were selected to perform blood vitamin D measurements after excluding women that were treated with exogenous hormones. Four subjects were not analysed because they had either insufficient sample volume or unreadable data. Five women were excluded because their age at mammography was over 55 years old. Our final analytic sample therefore included 491 premenopausal women. Among these women, 237 had mammography and blood samples obtained during the months of May/June (in Veracruz) and 243 during the months of October/November (in Jalisco). Informed consent was obtained from all participants and the study was approved by the Research, Biosafety and Ethics Committee at the National Institute of Public Health in Mexico, and by the IARC Ethics Committee. Mammographic density A radiology technician performed mammography using the Giotto Image M (Internazionale Medico Scientifica, Bologna, Italy) in Jalisco and the Hologic Lorad M-III (Hologic, Bedford, MA) in Veracruz. Mammograms were developed using the Agfa CP1000 (Agfa-Gevaert Group, Belgium) developer. Craniocaudal views were taken on each breast. An Astra 2400S scanner (Umax, Fremont, CA) was used to digitize the mammograms. A single observer measured MD on the left craniocaudal view using Mamgr, a computer-assisted program developed at the Department of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine [35, 36]. This thresholding software measures the total area as well as the total dense area on a mammogram. Percent MD is automatically calculated as the percent of “dense” pixels within the total breast area. Non-dense area was calculated by subtracting the dense area from the total breast area. Absolute dense and non-dense area values are converted to cm2 according to the pixel size used in the digitisation. In a reliability study of 100 ESMaestras mammograms, the intraclass correlation coefficient between MD measurements evaluated using the Mamgr software versus the Cumulus program developed at the University of Toronto was 0.87. In 108 duplicate mammograms, the intra-observer intraclass correlation was 0.84. Laboratory assays 25(OH)D Assessment Serum 25(OH)D2 and 25(OH)D3 levels were measured at Heartland Assays LLC (Ames, Iowa) using a liquid chromatography/tandem mass spectrometry (LC/MS/MS) method as previously described [37]. The average intra-assay coefficient of variation was 5.8% for 25(OH)D3, and 12.6% for 25(OH)D2. Only the analyses for serum 25(OH)D3, the primary exposure variable of interest, are presented because only 24 participants (4.8%) had detectable 25(OH)D2 concentrations above the limit of assay sensitivity of 1.56 ng/ml. Hormone analyses Analyses were performed on never-thawed serum samples continuously stored at −80°C as previously reported [38]. In brief, serum IGF-I and IGFBP3 concentrations were measured by immunoradiometric assays by Beckmann Coulter (Marseille, France) at the laboratory of hormone analyses, Biomarkers Group, International Agency for Research on Cancer (Lyon, France). The intra-assay and inter-assay coefficients of variations were 0.8% and 4.2%, respectively, for a concentration of 19.5 nmol/l for IGF-I, and 1.3% and 3.0%, respectively, for a concentration of 125 nmol/l for IGFBP-3. Statistical analysis Means and standard deviations (SDs), or percentages of selected baseline characteristics of the study population were estimated across predefined categories of serum 25(OH)D3 [39]. Chi-square and ANOVA tests were used to determine whether the distribution of the selected BC risk factors differed across categories of serum 25(OH)D3. Multivariable adjusted Spearman partial correlation coefficients were performed to investigate the association between serum 25(OH)D3 (ng/ml) and other continuous variables including age (years), body mass index (BMI, kg/m2), waist to hip ratio (WHR), and total physical activity (metabolic equivalent of energy expenditure, MetS per week). Multivariable linear regressions were used to estimate the association of different MD measures (percent MD, dense or non-dense areas) with serum 25(OH)D3. Based on population distribution, the quartiles (25th percentile, 50th percentile, 75th percentile) of percent MD that divide the MD set into four equal group were estimated (MD ≤ 22.6%, 22.6–37.7%,; 37.5–51.9%, and > 51.9%). A woman was considered having low MD if she was in the lower quartile (MD ≤ 22.6%), and high MD if she was in the upper quartile (MD >51.9%). A subset of women was selected based on whether they had low or high percent MD, and 25(OH)D3 concentrations were compared by multivariable logistic regressions. Restricting our analysis to these women allowed for a better discrimination between women at low and high BC risk. Additionally, different approaches to nonlinear modelling were used to explore the association between serum 25(OH)D3 and MD: 1) local polynomial regression (LOESS), which is a smoothing method that essentially summarizes the association between outcome and exposure by fitting a multitude of regression models to adjacent subsets of the data [40]; 2) fractional polynomial modelling, which fit models using various transformations of the predictor for which a non-linear association with the outcome is assumed [41]. Final multivariable models were adjusted for age (continuous), age at menarche (continuous), BMI (continuous), physical activity (continuous), and region/season. Serum 25(OH)D3 concentrations were used as continuous variable, or categorical variables (2 categories (≤ and > median), and predefined categories [39]). Separate analyses were performed adjusting for and then stratifying by BMI (< and ≥ median). The multiplicative interaction between serum 25(OH)D3 and BMI was tested by including a cross-product term in the multivariable model. Stratified analyses and tests for interactions were used to further evaluate possible effect modification of IGF1 and IGFBP3 on the association between serum 25(OH)D3 and percent MD. The Sobel-Goodman mediation method [42, 43] was used to assess the mediating role played by IGF1 and IGFP3 in the associations between 25(OH)D3 and percent MD, and to estimate the proportion of the effect that is mediated (%). Bias-corrected CIs for the percentage mediation were obtained through bootstrap techniques with 1,000 replications [44]. All statistical tests were two-sided and P-values < 0.05 were considered significant. All statistical analyses were conducted using STATA (version 11). Results Baseline characteristics of the study population across predefined categories of serum 25(OH)D3 are presented in Table 1. The mean age at recruitment of the 491 premenopausal women was 42.9 ± 3.7 years. Overall, the mean serum 25(OH)D3 levels was 28.12 ± 6.93 ng/ml. Forty one (8%) women had 25(OH)D3 levels in the deficient range (< 20 ng/ml), 265 (54%) were in the insufficient range (20–29 ng/ml) and 185 (38%) were in the sufficient range (≥ 30 ng/ml). Statistically significant differences in physical activity (P < 0.001), non-dense area (P = 0.005), and breastfeeding (P = 0.011) were observed across the three categories of 25(OH)D3. No significant differences in age at menarche, age at mammography, age at 1st full term pregnancy, parity, percent MD, dense area, region, BMI, alcohol intake, use of hormonal contraceptive, FHBC and socio economic status were observed (Table 1). 10.1371/journal.pone.0161686.t001Table 1 Characteristics of the study population across predefined cut-points of serum 25(OH)D3. Serum 25(OH)D3 (ng/ml) Characteristic < 20 (n = 41) 20–29 (n = 265) ≥ 30 (n = 185) P-value Means ± SD Age at mammography (years) 43.6±4.2 43.1±3.6 42.6±3.6 0.457 Age at menarche (years) 12.2±1.2 12.5±1.4 12.5±1.5 0.343 Age at 1st full term pregnancy (years) 24.8±4.4 24.8±4.3 24.6±4.8 0.407 parity 1.77±1.4 2.13±1.2 2.16±1.2 0.261 Total physical activity (MetS per week) 24.3±16.7 25.9±19.5 28.7±25.0 <0.001 Percent MD 36.4±18.6 37.6±17.5 37.1±17.1 0.778 Dense area (cm2) 46.8±30.6 50.5±34.6 48.1±31.8 0.343 Non dense area (cm2) 83.4±45.4 82.1±39.9 78.7±33.1 0.005 Frequency n (%) Region  Jalisco 26(10.6) 126(51.6) 92(37.7) 0.167  Veracruz 15(6.1) 139(56.3) 93(37.6) Body mass index (kg/m2)  < 30 kg/m2 22(7.1) 161(51.7) 128(41.2) 0.076  ≥ 30 kg/m2 19(10.5) 104(57.8) 57(31.7) Breastfeeding  Never 10(19.2) 25(48.1) 17(32.7) 0.011  Ever 31(7.1) 240(54.7) 168 (38.3) Alcohol intake  No 14(9.9) 65(46.1) 62(44.0) 0.085  Yes 27(7.7) 200(57.1) 123(35.1) Ever use of hormonal contraceptive  No 24(10.5) 124(54.1) 81(35.4) 0.364  Yes 17(6.9) 133(54.1) 96(39.0)  Missing 0(0) 8(50) 8(5) Family history of breast cancer  No 38(8.1) 252(53.7) 179(38.2) 0.462  Yes 3(13.6) 13(59.1) 6(27.3) Socio economic status  Low 5(7.1) 39(55.7) 26(37.1) 0.519  Medium 19(10.3) 102(55.1) 64(34.6)  High 15(8.9) 89(52.9) 64(38.1)  Missing 2(2.9) 35(51.5) 31(45.6) P value based on chi square and ANOVA tests. Table 2 shows Spearman partial correlation coefficients and P-values for the correlations between serum 25(OH)D3 levels and other factors. Serum 25(OH)D3 levels were inversely correlated with BMI (r = −0.109, P = 0.019) and directly related with total physical activity (r = 0.095, P = 0.035). No statistically significant correlations were found between serum 25(OH)D3 levels and others factors (age, WHR, dietary calcium, IGF1 and IGFBP3) (Table 2). 10.1371/journal.pone.0161686.t002Table 2 Correlation between serum 25(OH)D3 and other risk factors. Characteristic n Spearman correlation coefficient a P-value b Age (years) 491 -0.045 0.311 BMI (kg/m2) 454 -0.109 0.019 WHR 430 0.017 0.723 Total physical activity (MetS/week) 490 0.095 0.035 Dietary calcium intake (mg/day) 474 -0.041 0.361 IGF1 (ng/ml) 474 0.016 0.719 IGFBP3 (ng/ml) 474 0.082 0.069 BMI body mass index, WHR waist hip to hip ratio, MetS, metabolic equivalents of energy expenditure. a Spearman correlation coefficient adjusted for age, total physical activity and BMI. b P-value. In the adjusted multivariate linear regression, no significant associations were observed between serum 25(OH)D3 levels (in continuous) and different measures (percent MD, dense area and non-area) overall (Table 3). In stratified analyses according to the median of serum 25(OH)D3 levels (<27.3 ng/ml and ≥27.3 ng/ml), higher serum 25(OH)D3 levels (≥ 27.3 ng/ml) were marginally related to percent MD (β = −0.38, P = 0.059) (Table 3). When stratified by median BMI (< and ≥ 27.4 kg/m2), a borderline statistically significant inverse association was found between higher serum 25(OH)D3 levels (> median) and MD only among women with BMI below the median (β = −0.52, P = 0.047) compared to those with BMI above or equal to the median (β = −0.14, P = 0.699) (not shown). In the analyses stratified by the WHO cut-off point for normal BMI (< 25 kg/m2) and overweight BMI (≥ 25 kg/m2), no significant heterogeneity association was observed, although in overweight women (BMI ≥ 25 kg/m2) the association with non-dense area was significant (S1 Table). 10.1371/journal.pone.0161686.t003Table 3 Multivariable linear regression estimates of percent MD (%), dense area (cm2), and non-dense area (cm2). Characteristic Percent Mammographic density (%) Dense area (cm2) Non-dense area (cm2) β coefficient (95% CI) P-value β coefficient(95% CI) P-value β coefficient (95% CI) P-value Serum 25(OH)D3 (ng/ml)  Overall -0.02(-0.24, 0.20) 0.840 0.07(-0.37, 0.50) 0.754 -0.13(-0.53, 0.28) 0.537  < median (27.3 ng/ml) 0.25(-0.42, 0.91) 0.469 1.02(-0.27, 2.31) 0.121 -0.2(-1.45, 1.04) 0.747  ≥ median(27.3 ng/ml) -0.39(-0.81, 0.01) 0.059 -0.42(-1.21, 0.37) 0.295 0.34(-0.36, 1.04) 0.338  Predefined categories  < 20 1 (ref) 1 (ref) 1 (ref)  20–30 1.98(-3.67, 7.63) 0.491 7.03(-3.92, 17.99) 0.208 0.05(-10.23, 10.33) 0.992  ≥ 30 1.26(-4.67, 7.19) 0.676 6.38(-5.11, 17.88) 0.276 -0.71(-11.50, 10.08) 0.897 Region  Jalisco -0.05(-0.35, 0.24) 0.725 0.05(-0.52, 0.63) 0.850 0.19(-0.46, 0.50) 0.937  Veracruz 0.05(-0.27, 0.37) 0.759 0.19(-0.44, 0.82) 0.550 -0.07(-0.73, 0.58) 0.827 Age -1.73(-3.63, 0.17) 0.074 -2.97(-6.67, 0.73) 0.116 0.46(-3.00, 3.93) 0.793 Age at menarche 0.41(-0.59, 1.41) 0.423 0.63(-1.31, 2.59) 0.523 -0.26(-2.09, 1.57) 0.779 Body mass index (kg/m2) -0.52(-0.80, -0.23) <0.001 1.42(0.87, 1.97) <0.001 4.32(3.81, -4.83) <0.001 Total physical activity (MetS per week) 0.008(-0.06, 0.07) 0.812 -0.03(-0.16, 0.09) 0.627 -0.04(-0.16, 0.08) 0.533 Multivariable models were adjusted for age, age at menarche, body mass index, total physical activity and region/season. No statistically significant associations were observed when we assessed serum 25(OH)D3 levels as predefined categories (<20, 20–29, and ≥30 ng/ml) (Table 3). There were no significant differences in associations between serum 25(OH)D3 and region (P interaction = 0.779). In the analyses restricted to women classified as having low MD (≤22.6%) and high MD (>51.9%), there was limited evidence of an inverse association between high serum 25(OH)D3 and MD. The odds ratio (OR) of having high MD for women with serum 25(OH)D3 above the median was 0.91(95%CI: 0.83–0.99) (not shown). This inverse association was significant only among women with BMI below the median (OR = 0.87, 95%CI: 0.78–0.98) (Fig 1). However there was no significant interaction between serum 25(OH)D3 (continuous) and BMI (continuous) (P interaction = 0.283). This could be due to the small sample size. 10.1371/journal.pone.0161686.g001Fig 1 Association of serum 25(OH)D3 and percent MD, stratified by BMI. Based on population distribution, the quartiles (25th percentile, 50th percentile, 75th percentile) of percent MD that divide the MD set into four equal group were estimated (MD ≤ 22.6%, 22.6–37.7%,; 37.5–51.9%, and > 51.9%). A subset of women was selected based on whether they had low percent MD (≤ 22.6%) or high percent MD (> 51.9%), and 25(OH)D3 concentrations were compared by multivariable logistic regressions. Multivariable models were adjusted for age, age at menarche, total physical activity and season of blood draw. Median BMI = 27.4 kg/m2, median serum 25(OH)D = 27.3ng/ml. N (high/low): number (high MD/ low MD), OR: odds ratio; 95% CI 95% confidence interval. In non-linear models including fractional polynomial modelling, there was a trend of an inverse the association between serum 25(OH)D3 levels and MD only among women with BMI < median (Fig 2). However the test of non-linearity was not significant. 10.1371/journal.pone.0161686.g002Fig 2 Fractional polynomial modelling of the association of serum 25(OH)D3 (ng/ml) with percent mammography density (MD). A model with 95%CI among women with: A) BMI < median (< 27.4 kgm2), B) BMI ≥ median (≥ 27.4 kgm2). Finally, we investigated whether IGF1 and IGFBP3 modified the relation of serum 25(OH)D3 with MD, and no statistically significant interactions were found (P interaction were 0.764 and 0.398, respectively for IGF1 and IGFBP3, S2 Table). In the mediation analyses, no evidence of the indirect effects of serum 25(OH)D3 on MD through IGF1 and IGFBP3 was observed (all P value > 0.05) (S2 Table). Discussion In this study, we observed a non-significant inverse association between serum 25(OH)D3 levels and MD overall. However, in the analyses restricted to women classified as having low MD and high MD, we observed an inverse association between high serum 25(OH)D3 and MD although of borderline significance. When stratified by BMI, this inverse association appeared statistically significant only among women with BMI below the median. Our finding of non-significant association overall between serum 25(OH)D3 levels and MD among premenopausal women is consistent with results from several studies [21–23, 25]. More recently the study by Crew et al on 195 women aged 40–60 years found no association between serum 25(OH)D and different measures of MD, however [22]. In a recent Korean study, the authors reported a small but statistically significant association between serum 25(OH)D and MD in correlation analyses, but no significant association was observed in the multivariable regression analyses [23]. A cross-sectional study by Chai et al., showed that serum 25(OH)D was not associated with MD among 182 premenopausal women [21]. Indeed, the associations between vitamin D and MD are not fully understood, as previous findings have shown inconsistent results. Some studies reported no significant associations [21, 23, 25] with lower MD among both pre- and post-menopausal women [19, 24]. In contrast to our study, Brisson et al. reported that changes in circulating vitamin D (25(OH)D) were inversely associated with changes in MD with a lag time of about four months [18]. This study showed the importance of the lag time that could be a critical concept when assessing the relation of changes in circulating 25(OH)D to changes in MD. This may explain the lack of significant association observed overall in our study. However, our blood sample collections were done over 4 months in two regions (May/June in Jalisco and October/November in Veracruz) do not allow a lag time analysis. Comparing high MD (highest quartile) and low MD (lowest quartile) we observed a significant inverse association between serum 25(OH)D3 and MD among the group with lower BMI (< median) and higher serum 25(OH)D3 (> median). This suggests that the effect of vitamin D may be observed after a certain level in the blood only in women with BMI below the median. The putative inverse relationship between vitamin D and obesity was firstly described by Rosenstreich et al. in 1971 [45]. The notion that vitamin D level is related to obesity is also consistent with a recent meta-analysis of twelve studies, which reported a pooled relative risk of 1.52 (95% CI: 1.33–1.73) for risk of vitamin D deficiency (< 50 nmol/L (equivalent to 15.7 ng/ml)) in obese people (BMI > 30 kg/m2) [46]. Likewise, although Kuhn et al. reported that overall levels of serum 25(OH)D were not associated with the risk of BC, they found a non-significant inverse association among women with BMI < 25 kg/m2 (OR = 0.83; 95% CI: 0.67–1.03, p = 0.09) [47]. Furthermore, several studies reported different strength of the association between vitamin D and MD according to some circulating growth factors levels [26, 48]. This was shown in the study of Diorio et al. which reported a stronger inverse association between MD and dietary vitamin D level in women with higher IGF-1 and lower IGFBP3 levels than in those with low levels, suggesting that the association between vitamin D, calcium and MD may be limited to premenopausal women, who have higher levels of calcium, IGF-1 and IGFBP-3 [26]. In contrast, our results showed no effect modification by these biomarkers. The main classical roles of vitamin D involve the regulation of calcium metabolism and skeletal remodelling [11, 12]. Vitamin D has also significant anticancer effects including inhibition of proliferation, induction of differentiation, and promotion of apoptosis in breast cells [49, 50]. In addition, vitamin D actions involve: transcriptional repression of aromatase via promoter II in BC cells and surrounding adipose tissue; decrease in prostaglandin E2 (PGE2), a major stimulator of aromatase transcription in BC cells; transcriptional repression of ER in BC cells to block oestrogen stimulus [14]. The protective effects of vitamin D have been shown to function mainly through the vitamin D receptor (VDR) present in breast cells [16, 48]. Vitamin D can also influence MD by indirect effects due to overlap with other pathways [16, 51]. Some studies have suggested that the effect of Vitamin D in the breast might result from its effect on the insulin growth factor signalling pathway [11, 19]. Vitamin D has been shown to stimulate and enhance the expression of IGFBP-3 [16, 29]. In contrast, it inhibits the mitogenic effect of IGF1, attenuates the antiapoptotic effect of IGF1, and down regulates the expression of IGF1 receptors [29, 52, 53]. However in our mediation analyses, we did not observe any evidence of the indirect effects of serum 25(OH)D3 on MD through IGF1 and IGFBP3. The current study had several strengths, including the use of serum 25(OH)D3 concentrations, which is the best biomarker reflecting the total body vitamin D levels and long-term vitamin D. The blood draws and mammograms were taken on the same date, and all mammograms were evaluated with a computer assisted method. We adjusted for potentials confounding factors and account for region/season. This is the first study conducted among Mexican women, however some limitations should be considered. We used a single measure of vitamin serum levels, which may conduct to a possible vitamin D exposure misclassification since circulating vitamin D is prone to seasonal variability. However, a previous study suggested that serum 25(OH)D concentration at a single time point may be a useful biomarker of long-term vitamin D status [54]. Another limitation is the potential misclassification of participants’ vitamin D status, due to excluding serum 25(OH)D2 concentrations. However this should be minimal, since unlike vitamin D3, vitamin D2 is present mostly in fungus-/yeast-derived products; thus, its contribution to overall vitamin D status is negligible. In addition, vitamin D3 is the most utilized form of vitamin D in clinical trials [55]. In conclusion, although, no significant association was observed between serum 25(OH)D3 and percent MD in the overall population, our study supports an inverse association between higher serum 25(OH)D3 levels (> median) and MD in premenopausal women with BMI below the median women. It is possible that the impact of 25 (OH)D3 is observed only after a certain threshold in the blood. Indeed, the Endocrine Society has stated that the desirable serum concentration of 25(OH)D was >30 ng/ml to maximize its effect on calcium, bone, and muscle metabolism [39]. More research is needed to understand this association and to see whether vitamin D supplementation may play a preventive role in BC. Supporting Information S1 Table Multivariable linear regression estimates of percent MD (%), dense area (cm2), and non-dense area (cm2), stratified by normal (< 25 kg/m2) and overweight BMI (≥ 25 kg/m2). Multivariable models were adjusted for age, age at menarche, body mass index, total physical activity and region/season. Serum 25(OH)D3 concentrations were used as continuous variable, or categorical variables (2 categories: ≤ and > median), and predefined categories. (DOCX) Click here for additional data file. S2 Table Multivariate Sobel–Goodman mediation tests and bias-corrected confidence interval. CI, confidence interval. Multivariable models were adjusted for age, age at menarche, body mass index, total physical activity and region/season. Tests for interactions were used to further evaluate possible effect modification of IGF1 and IGFBP3 on the association between serum 25(OH)D3 and percent MD. The Sobel-Goodman mediation method was used to assess the mediating role played by IGF1 and IGFP3 in the associations between 25(OH)D3 and percent MD, and to estimate the proportion of the effect that is mediated (%). Bias-corrected CIs for the percentage mediation were obtained through bootstrap techniques with 1,000 replications. (DOCX) Click here for additional data file. We are grateful first and foremost to all MTC participants for their time and commitment. We would like to thank Carrera Magisterial at the Ministry of Education, with special thanks to Victor Sastré, Director de Regulation. We thank ISSSTE’s (Institute for Social Security and Services for Civil Servants) Medical Directorate staff and Jalisco's regional office and the Veracruz Ministry of Health for technical and administrative support. ==== Refs References 1 Assi V , Warwick J , Cuzick J , Duffy SW (2012 ) Clinical and epidemiological issues in mammographic density . Nat Rev Clin Oncol 9 : 33 –40 2 Huo CW , Chew GL , Britt KL , Ingman WV , Henderson MA , Hopper JL , et al (2014 ) Mammographic density-a review on the current understanding of its association with breast cancer . Breast Cancer Res Treat 144 : 479 –502 10.1007/s10549-014-2901-2 24615497 3 White J (2000 ) Breast density and cancer risk: what is the relationship? J Natl Cancer Inst 92 : 443 10716955 4 Boyd NF , Martin LJ , Rommens JM , Paterson AD , Minkin S , Yaffe MJ , et al (2009 ) Mammographic density: a heritable risk factor for breast cancer . Methods Mol Biol 472 : 343 –360 10.1007/978-1-60327-492-0_15 19107441 5 Boyd NF , Rommens JM , Vogt K , Lee V , Hopper JL , Yaffe MJ , et al (2005 ) Mammographic breast density as an intermediate phenotype for breast cancer . Lancet Oncol 6 : 798 –808 16198986 6 Martin LJ , Boyd NF (2008 ) Mammographic density. Potential mechanisms of breast cancer risk associated with mammographic density: hypotheses based on epidemiological evidence . Breast Cancer Res 10 : 201 10.1186/bcr1831 18226174 7 Boyd NF , Melnichouk O , Martin LJ , Hislop G , Chiarelli AM , Yaffe MJ , et al (2011 ) Mammographic density, response to hormones, and breast cancer risk . J Clin Oncol 29 : 2985 –2992 10.1200/JCO.2010.33.7964 21709206 8 Chen P , Li M , Gu X , Liu Y , Li X , Li C , et al (2013 ) Higher blood 25(OH)D level may reduce the breast cancer risk: evidence from a Chinese population based case-control study and meta-analysis of the observational studies . PLoS One 8 : e49312 10.1371/journal.pone.0049312 23382798 9 Fedirko V , Torres-Mejia G , Ortega-Olvera C , Biessy C , ngeles-Llerenas A , Lazcano-Ponce E , et al (2012 ) Serum 25-hydroxyvitamin D and risk of breast cancer: results of a large population-based case-control study in Mexican women . Cancer Causes Control 23 : 1149 –1162 10.1007/s10552-012-9984-z 22622862 10 Yaghjyan L , Colditz GA , Drake B (2012 ) Vitamin D and mammographic breast density: a systematic review . Cancer Causes Control 23 : 1 –13 11 Khan QJ , Kimler BF , Fabian CJ (2010 ) The relationship between vitamin D and breast cancer incidence and natural history . Curr Oncol Rep 12 : 136 –142 10.1007/s11912-010-0081-8 20425599 12 Prentice A , Goldberg GR , Schoenmakers I (2008 ) Vitamin D across the lifecycle: physiology and biomarkers . Am J Clin Nutr 88 : 500S –506S 18689390 13 Jones G , Prosser DE , Kaufmann M (2014 ) Cytochrome P450-mediated metabolism of vitamin D . J Lipid Res 55 : 13 –31 . 10.1194/jlr.R031534 23564710 14 Feldman D , Krishnan AV , Swami S , Giovannucci E , Feldman BJ (2015) The role of vitamin D in reducing cancer risk and progression . Nat Rev Cancer 14 : 342 –357 . 2015 . 15 Thacher TD , Clarke BL (2011 ) Vitamin D insufficiency . Mayo Clin Proc 86 : 50 –60 10.4065/mcp.2010.0567 21193656 16 Deeb KK , Trump DL , Johnson CS (2007 ) Vitamin D signalling pathways in cancer: potential for anticancer therapeutics . Nat Rev Cancer 7 : 684 –700 17721433 17 Bertone-Johnson ER , Chlebowski RT , Manson JE , Wactawski-Wende J , Aragaki AK , Tamimi RM , et al (2010 ) Dietary vitamin D and calcium intake and mammographic density in postmenopausal women . Menopause 17 : 1152 –1160 10.1097/gme.0b013e3181e102d9 20613676 18 Brisson J , Berube S , Diorio C , Sinotte M , Pollak M , Masse B (2007 ) Synchronized seasonal variations of mammographic breast density and plasma 25-hydroxyvitamin d . Cancer Epidemiol Biomarkers Prev 16 : 929 –933 17507618 19 Berube S , Diorio C , Masse B , Hebert-Croteau N , Byrne C , Cote G , et al (2005 ) Vitamin D and calcium intakes from food or supplements and mammographic breast density . Cancer Epidemiol Biomarkers Prev 14 : 1653 –1659 16030097 20 Berube S , Diorio C , Verhoek-Oftedahl W , Brisson J (2004 ) Vitamin D, calcium, and mammographic breast densities . Cancer Epidemiol Biomarkers Prev 13 : 1466 –1472 15342447 21 Chai W , Maskarinec G , Cooney RV (2010 ) Serum 25-hydroxyvitamin D levels and mammographic density among premenopausal women in a multiethnic population . Eur J Clin Nutr 64 : 652 –654 10.1038/ejcn.2010.36 20216557 22 Crew KD , Campbell J , Reynolds D , Fulton L , Flom JD , Liao Y , et al (2014 ) Mammographic density and serum 25-hydroxyvitamin D levels . Nutr Metab (Lond) 11 : 18 24742098 23 Heo DS , Lee JG , Hwang HR , Lee SY , Cho BM , Kim SS , et al (2014 ) The association between 25-hydroxyvitamin D and mammographic density in healthy pre- and postmenopausal women regardless of the menstrual cycle phase: a cross-sectional study . Nutr Cancer 66 : 97 –103 10.1080/01635581.2014.851715 24328856 24 Thomson CA , Arendell LA , Bruhn RL , Maskarinec G , Lopez AM , Wright NC , et al (2007 ) Pilot study of dietary influences on mammographic density in pre- and postmenopausal Hispanic and non-Hispanic white women . Menopause 14 : 243 –250 17091096 25 Knight JA , Vachon CM , Vierkant RA , Vieth R , Cerhan JR , Sellers TA (2006 ) No association between 25-hydroxyvitamin D and mammographic density . Cancer Epidemiol Biomarkers Prev 15 : 1988 –1992 17035410 26 Diorio C , Berube S , Byrne C , Masse B , Hebert-Croteau N , Yaffe M , et al (2006 ) Influence of insulin-like growth factors on the strength of the relation of vitamin D and calcium intakes to mammographic breast density . Cancer Res 66 : 588 –597 16397276 27 Diorio C , Pollak M , Byrne C , Masse B , Hebert-Croteau N , Yaffe M , et al (2005 ) Insulin-like growth factor-I, IGF-binding protein-3, and mammographic breast density . Cancer Epidemiol Biomarkers Prev 14 : 1065 –1073 15894654 28 Peng L , Malloy PJ , Feldman D (2004 ) Identification of a functional vitamin D response element in the human insulin-like growth factor binding protein-3 promoter . Mol Endocrinol 18 : 1109 –1119 14963110 29 Colston KW , Perks CM , Xie SP , Holly JM (1998 ) Growth inhibition of both MCF-7 and Hs578T human breast cancer cell lines by vitamin D analogues is associated with increased expression of insulin-like growth factor binding protein-3 . J Mol Endocrinol 20 : 157 –162 9513092 30 Ameri P , Giusti A , Boschetti M , Murialdo G , Minuto F , Ferone D (2013 ) Interactions between vitamin D and IGF-I: from physiology to clinical practice . Clin Endocrinol 79 : 457 –463 31 Romieu I , Escamilla-Nunez MC , Sanchez-Zamorano LM , Lopez-Ridaura R , Torres-Mejia G , Yunes EM , et al (2012 ) The association between body shape silhouette and dietary pattern among Mexican women . Public Health Nutr 15 : 116 –125 10.1017/S1368980011001182 21875454 32 Rice MS , Biessy C , Lajous M , Bertrand KA , Tamimi RM , Torres-Mejia G , et al (2013 ) Metabolic syndrome and mammographic density in Mexican women . Cancer Prev Res (Phila) 6 : 701 –710 23682074 33 Angeles-Llerenas A , Ortega-Olvera C , Perez-Rodriguez E , Esparza-Cano JP , Lazcano-Ponce E , Romieu I , et al (2010 ) Moderate physical activity and breast cancer risk: the effect of menopausal status . Cancer Causes Control 21 : 577 –586 10.1007/s10552-009-9487-8 20084545 34 Boyd NF , Guo H , Martin LJ , Sun L , Stone J , Fishell E , et al (2007 ) Mammographic density and the risk and detection of breast cancer . N Engl J Med 356 : 227 –236 17229950 35 Byng JW , Boyd NF , Fishell E , Jong RA , Yaffe MJ (1994 ) The quantitative analysis of mammographic densities . Phys Med Biol 39 : 1629 –1638 15551535 36 Torres-Mejia G , De SB , Allen DS , Perez-Gavilan JJ , Ferreira JM , Fentiman IS , et al (2005 ) Mammographic features and subsequent risk of breast cancer: a comparison of qualitative and quantitative evaluations in the Guernsey prospective studies . Cancer Epidemiol Biomarkers Prev 14 : 1052 –1059 15894652 37 Saenger AK , Laha TJ , Bremner DE , Sadrzadeh SM (2006 ) Quantification of serum 25-hydroxyvitamin D(2) and D(3) using HPLC-tandem mass spectrometry and examination of reference intervals for diagnosis of vitamin D deficiency . Am J Clin Pathol 125 : 914 –920 16690491 38 Rinaldi S , Biessy C , Hernandez M , Lesueur F , dos-Santos-Silva I , Rice MS , et al (2014 ) Circulating concentrations of insulin-like growth factor-I, insulin-like growth factor-binding protein-3, genetic polymorphisms and mammographic density in premenopausal Mexican women: results from the ESMaestras cohort . Int J Cancer 134 : 1436 –1444 10.1002/ijc.28469 24037648 39 Holick MF , Binkley NC , Bischoff-Ferrari HA , Gordon CM , Hanley DA , Heaney RP , et al (2011 ) Evaluation, treatment, and prevention of vitamin D deficiency: an Endocrine Society clinical practice guideline . J Clin Endocrinol Metab 96 : 1911 –1930 10.1210/jc.2011-0385 21646368 40 Cleveland WS , Devlin SJ (1988 ) Locally weighted regression: an approach to regression analysis by local fitting . J Am Stat Assoc 83 : 596 –610 41 Royston P , Ambler G , Sauerbrei W (1999 ) The use of fractional polynomials to model continuous risk variables in epidemiology . Int J Epidemiol 28 : 964 –974 10597998 42 Baron RM , Kenny DA (1986 ) The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations . J Pers Soc Psychol 51 : 1173 –1182 3806354 43 De Neve JE , Oswald AJ (2012 ) Estimating the influence of life satisfaction and positive affect on later income using sibling fixed effects . Proc Natl Acad Sci U S A 109 : 19953 –19958 10.1073/pnas.1211437109 23169627 44 Haukoos JS , Lewis RJ (2005 ) Advanced statistics: bootstrapping confidence intervals for statistics with "difficult" distributions . Acad Emerg Med 12 : 360 –365 15805329 45 Rosenstreich SJ , Rich C , Volwiler W (1971 ) Deposition in and release of vitamin D3 from body fat: evidence for a storage site in the rat . J Clin Invest 50 : 679 –687 4322721 46 Shanmugalingam T , Crawley D , Bosco C , Melvin J , Rohrmann S , Chowdhury S , et al (2014 ) Obesity and cancer: the role of vitamin D . BMC Cancer 14 : 712 10.1186/1471-2407-14-712 25255691 47 Kuhn T , Kaaks R , Becker S , Eomois PP , Clavel-Chapelon F , Kvaskoff M , et al (2013 ) Plasma 25-hydroxyvitamin D and the risk of breast cancer in the European prospective investigation into cancer and nutrition: a nested case-control study . Int J Cancer 133 : 1689 –1700 10.1002/ijc.28172 23526380 48 Diorio C , Sinotte M , Brisson J , Berube S , Pollak M (2008 ) Vitamin D pathway polymorphisms in relation to mammographic breast density . Cancer Epidemiol Biomarkers Prev 17 : 2505 –2508 10.1158/1055-9965.EPI-08-0493 18768522 49 Krishnan AV , Feldman D (2011 ) Mechanisms of the anti-cancer and anti-inflammatory actions of vitamin D . Annu Rev Pharmacol Toxicol 51 : 311 –336 10.1146/annurev-pharmtox-010510-100611 20936945 50 Matthews D , LaPorta E , Zinser GM , Narvaez CJ , Welsh J (2010 ) Genomic vitamin D signaling in breast cancer: Insights from animal models and human cells . J Steroid Biochem Mol Biol 121 : 362 –367 10.1016/j.jsbmb.2010.03.061 20412854 51 Tseng M , Byrne C , Evers KA , Daly MB (2007 ) Dietary intake and breast density in high-risk women: a cross-sectional study . Breast Cancer Res 9 : R72 17949495 52 Colston KW , Hansen CM (2002 ) Mechanisms implicated in the growth regulatory effects of vitamin D in breast cancer . Endocr Relat Cancer 9 : 45 –59 11914182 53 Lowe L , Hansen CM , Senaratne S , Colston KW (2003 ) Mechanisms implicated in the growth regulatory effects of vitamin D compounds in breast cancer cells . Recent Results Cancer Res 164 : 99 –110 12908448 54 Hofmann JN , Yu K , Horst RL , Hayes RB , Purdue MP (2010 ) Long-term variation in serum 25-hydroxyvitamin D concentration among participants in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial . Cancer Epidemiol Biomarkers Prev 19 : 927 –931 10.1158/1055-9965.EPI-09-1121 20332255 55 Wolpowitz D , Gilchrest BA (2006 ) The vitamin D questions: how much do you need and how should you get it? J Am Acad Dermatol 54 : 301 –317 16443061

PMC005xxxxxx/PMC5001726.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756455010.1371/journal.pone.0161905PONE-D-16-23162Research ArticleEarth SciencesGeomorphologyTopographyLandformsBeachesBiology and Life SciencesOrganismsAnimalsInvertebratesBiology and Life SciencesOrganismsAnimalsVertebratesAmniotesBirdsBiology and Life SciencesEcologyCommunity EcologyTrophic InteractionsPredationEcology and Environmental SciencesEcologyCommunity EcologyTrophic InteractionsPredationEcology and Environmental SciencesAquatic EnvironmentsMarine EnvironmentsCoastsEarth SciencesMarine and Aquatic SciencesAquatic EnvironmentsMarine EnvironmentsCoastsEarth SciencesMarine and Aquatic SciencesBodies of WaterOceansBiology and Life SciencesEcologyEcological MetricsSpecies DiversityEcology and Environmental SciencesEcologyEcological MetricsSpecies DiversityEcology and Environmental SciencesHabitatsThe Early Shorebird Will Catch Fewer Invertebrates on Trampled Sandy Beaches Trampling Kills Beach InvertebratesSchlacher Thomas A. 1*Carracher Lucy K. 2Porch Nicholas 2Connolly Rod M. 3Olds Andrew D. 1Gilby Ben L. 1Ekanayake Kasun B. 12Maslo Brooke 4Weston Michael A. 21 School of Science and Engineering, The University of the Sunshine Coast, Q-4558, Maroochydore, Australia2 Centre for Integrative Ecology, School of Life and Environmental, Deakin University, Geelong, Australia3 Australian Rivers Institute - Coast & Estuaries, and School of Environment, Gold Coast Campus, Griffith University, Southport, Australia4 Department of Ecology, Evolution and Natural Resources, Rutgers, The State University of New Jersey, 55 Commercial Ave, New Brunswick, NJ, 08901, United States of AmericaBersier Louis-Felix EditorUniversity of Fribourg, SWITZERLANDCompeting Interests: The authors have declared that no competing interests exist. Conceptualization: TAS LKC MAW. Data curation: TAS LKC MAW. Formal analysis: TAS LKC RMC ADO BLG MAW. Funding acquisition: MAW. Investigation: LKC NP MAW. Methodology: LKC NP MAW. Project administration: MAW. Resources: LKC NP MAW. Supervision: MAW. Validation: TAS LKC NP MAW. Visualization: TAS. Writing – original draft: TAS LKC MAW. Writing – review & editing: TAS LKC RMC ADO BLG KBE BM MAW. * E-mail: tschlach@usc.edu.au26 8 2016 2016 11 8 e01619059 6 2016 12 8 2016 © 2016 Schlacher et al2016Schlacher et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Many species of birds breeding on ocean beaches and in coastal dunes are of global conservation concern. Most of these species rely on invertebrates (e.g. insects, small crustaceans) as an irreplaceable food source, foraging primarily around the strandline on the upper beach near the dunes. Sandy beaches are also prime sites for human recreation, which impacts these food resources via negative trampling effects. We quantified acute trampling impacts on assemblages of upper shore invertebrates in a controlled experiment over a range of foot traffic intensities (up to 56 steps per square metre) on a temperate beach in Victoria, Australia. Trampling significantly altered assemblage structure (species composition and density) and was correlated with significant declines in invertebrate abundance and species richness. Trampling effects were strongest for rare species. In heavily trafficked plots the abundance of sand hoppers (Amphipoda), a principal prey item of threatened Hooded Plovers breeding on this beach, was halved. In contrast to the consistently strong effects of trampling, natural habitat attributes (e.g. sediment grain size, compactness) were much less influential predictors. If acute suppression of invertebrates caused by trampling, as demonstrated here, is more widespread on beaches it may constitute a significant threat to endangered vertebrates reliant on these invertebrates. This calls for a re-thinking of conservation actions by considering active management of food resources, possibly through enhancement of wrack or direct augmentation of prey items to breeding territories. The authors received no specific funding for this work. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction A bird came down the walk: He did not know I saw; He bit an angle-worm in halves And ate the fellow, raw. “A Bird Came Down” Emily Dickinson Globally, ocean beaches attract humans whose activities cause numerous detrimental impacts to beach ecosystems [1–3]. Beaches are prime sites for recreation[4]. From an environmental perspective, one of the ways that recreation manifests itself is through the impact of trampling by people, a near-ubiquitous impact which is most severe on developed coasts. The harmful consequences of trampling are well documented for coastal dunes, especially the impacts on vegetation [4]. For the non-vegetated part of the shore, seawards of the dunes, trampling impacts are either insufficiently quantified or are confounded with more diffuse pressures from urbanization, shore armouring, or grooming [5, 6]. The two studies that could attribute changes in shallow-buried beach invertebrates to human trampling show declines in abundance for species of the middle and lower shore [7, 8]. It is, however, largely unknown whether—and to which extent—trampling may impact invertebrates on the upper shore where many shorebirds feed and human foot traffic is concentrated during high tides [9]. Ocean beaches are habitat for diverse assemblages of invertebrates [10, 11]. On the upper shore near the dunes, these assemblages are mainly composed of insects (e.g. coleoptera, diptera) and smaller peracarid crustaceans (e.g. amphipods, isopods) that are buried in the sand or under wrack during the day and become surface active at night [12]. Beach invertebrates are an irreplaceable food source for beach-nesting birds, a group that contains several species of critical conservation concern [13]. Obligate beach-nesting birds depend on adequate invertebrate prey, particularly in the upper beach [14], actively select prey-rich sections of beaches [15], and have energetically-demanding reproductive strategies [16]. Given that food is often a limiting resource and that many shorebirds are in decline worldwide, processes that diminish prey abundance are likely to be of major importance to the conservation of beach-dwelling birds [9]. Plovers on ocean beaches, worldwide rely heavily on upper beach invertebrates as a food source (Fig 1). They have catholic tastes, consuming a diversity of prey items that generally includes amphipods, isopods, molluscs, worms and a broad range of insects (e.g. beetles, dipterans, ants) [17–19]. Plovers regularly forage across habitat boundaries, feeding on exposed shores, dunes and upland wetlands during the day and at night [20]. In hooded plovers, biparental care and low reproductive success also mean that adults have limited mobility for long periods when breeding (incubation: 30 days; brood-rearing: 35 days) and that breeding territories are used for up to eight months per year [21]. Thus, food availability is a pivotal factor in habitat selection in beach-associated birds, extending also to non-breeding periods [15, 19, 22, 23]. 10.1371/journal.pone.0161905.g001Fig 1 A hooded plover chick feeding on the strandline of beaches in the study region (Photo: Glenn Ehmke). Management interventions for threatened beach birds usually encompass reducing disturbance and direct physical threats through access restrictions, predator control, and provisioning of shelter [24, 25]. By contrast, management interventions focusing on the conservation and supplementation of food resources for obligate beach birds are rarely implemented. Trampling might reduce the abundance of prey available to beach-foraging birds and impact on the conservation of these species, but this hypothesis has rarely been tested with empirical data. “Hooded Plovers often breed in areas heavily used by recreationists [26] and their habitat is often heavily trampled (M.A. Weston pers. obs.).” Thus, here we examine whether intense human trampling acutely alters upper-shore invertebrate populations on a beach that supports a breeding population of the threatened Hooded Plover, Thinornis rubricollis. Methods Ethics Statement All field work was undertaken in in accordance with Permit No.10007183 issued for the work by the Victorian Department of Primary Industry and Environment. It did not involve the use of animals requiring animal ethics committee clearance by Deakin University or any other body with jurisdiction for the site or the taxa. In no instance was research undertaken on private land. No other permissions were necessary for the research reported here. Study Site We ran an experiment to determine the acute effects of human trampling on upper-shore invertebrates on an exposed (mean wave height 1.8 m) beach on the south coast of Australia. The experimental site was Venus Bay (-38.71°, 145.82°) in Victoria, 170 km south of Melbourne. The beach is ca. 30 km long, faces south-west, and is of the intermediate morphodynamic type (100–150 m wide intertidal zone at low tide that slopes relatively gently (3–8°) and is composed of sands with a median grain size of 200–260 microns). The site was chosen for two reasons: 1) it represents important breeding habitat for a sandy-shore obligate threatened species, the Hooded Plover, a species for which human trampling may detrimentally affect prey resources; and, 2) some beach sections have very few visitors (M. Weston pers. obs.) and hence had low human trampling impacts before the experiment. Layout of Field Experiments We dispersed fourteen 5 x 5-m plots (Fig 2) along the upper part of the shore between the last high-tide mark and the base of the dunes using four criteria: 1) maximize the distance from beach access points to avoid interference from beach visitors; 2) avoid known nest sites; 3) exclude very narrow sections of the beach where tides would likely inundate experimental plots; and 4) separate adjoining plots by at least 50 m alongshore (the actual mean distance between plots was 233 m). 10.1371/journal.pone.0161905.g002Fig 2 Illustration of plot lay-outs and surface disturbance caused by experimental human trampling. We randomly assigned a trampling intensity to each experimental plot, which ranged between 0 and 1,400 steps, applied at increments of 100. We effected trampling by allowing three-person teams (weight 55–80 kg) to walk barefoot on the beach inside the plots until the required number of steps was reached. The treatment was applied such that an approximately even spread of footprints occurred across the plot; this was achieved by having an observer directing the walkers to “spread out” the trampling and who was aware how many steps remained. The experiment was done on 8 Dec. 2014. Ethical considerations demanded that we limited any physical alterations to the beach habitat and avoided disturbing nesting birds; hence, all experimental trampling was limited to a maximum intensity of 1,400 steps per plot and to a single pulse event. While few studies of trampling on beaches report density of steps for prevailing recreational trampling, our trampling rates visually appear typical of heavy recreational use (Fig 2). The density of steps was, however, less than that associated with spatially constrained sports activities (i.e. volleyball). Sampling Surface-active invertebrates were sampled with pitfall traps (350 ml plastic cups, 75 mm diameter, 110 mm deep, half filled with a mix of seawater and detergent, sunk into the sand so the rim was flush with the surface). We placed 25 traps, separated by 1 m, within each plot. Traps were placed after trampling and in a manner so as not to alter trampling rates (i.e. from the edge and from a board positioned so that it enabled access to the centre of the plot). Traps were placed immediately after the trampling treatment near sunset and left out overnight, with a deployment duration of of 851 ± 9 (se) minutes; Cuttris et al. [23] showed that on very similar beaches in the same region, species saturation in pitfall collections is reached after approximately 360 minutes. Sediment compactness was measured by dropping a stainless steel rod (weight 450 g, diameter 10 mm, point blunt) five times before trampling and averaging the penetration distance. We visually estimated the percentage of plots covered with macroalgal wrack and ca. 100 g of sand was collected from the top 10 cm of each plot for grain size analysis. Data Analysis The chief question was how important foot trampling by walkers is in comparison to habitat features frequently reported to be associated with variations in invertebrate assemblages on upper parts of sandy beaches (i. e. wrack, sediment compactness, grain size). We addressed this with distance-based linear models, DISTLMs, [27], relating trampling intensity, deployment duration of traps and environmental predictors to three complementary fauna metrics (assemblage structure, abundance, species richness). For the fauna, similarity matrices were based on Bray-Curtis resemblances (calculated from untransformed abundance data) for multivariate assemblage structure n and Euclidean distances for total catch (individuals per trap, untransformed) and species richness (number of species per trap). All analyses were run on data pooled over all traps per plot, using the PRIMER software [28]. The contribution of individual species to the total dissimilarity between reference and trampled plots was examined with the SIMPER (similarity percentages) of Primer [28]. Model performance was evaluated using the corrected Akaike Information Criterion (AICc) based on all possible combinations of variables used in model building [29, 30]. A multi-model inference approach was employed to assess the contributions of individual variables based on their summed Akaike weights [31]: summed AICc weights (w+) provide relative probabilities of variable importance, with variables < 0.3 likely to be of minor or no importance [32]. As a diagnostic before the model runs, we checked for confounding between experimental treatments and habitat attributes and found that trampling intensity was not significantly correlated with any of the environmental variables (max. r = 0.47; min. P = 0.09). Results The intensity of experimental trampling was the most influential predictor for observed heterogeneity in invertebrate assemblages (Table 1, S1 Table). Trampling was the highest-ranked predictor in all models, based on either variable weights or the proportion of variance accounted for; no other tested variable accounted for significant fractions of variance in assemblage composition or total catch in our models, nor could they be considered important based on their relatively low variable weights (i.e. mostly <0.30; Table 1). The cover of macroalgal wrack that was naturally present on experimental plots was an important predictor of species richness, with a minor contribution from deployment duration (Table 1). 10.1371/journal.pone.0161905.t001Table 1 Contributions of variables in models relating three metrics of invertebrate assemblages (assemblage structure, total catch of individuals, species richness) to experimental trampling, key habitat attributes (wrack cover, sediment compactness, grain size) and the time pitfall traps were deployed during the experiments. Variable contributions are assessed in two complementary ways: i) a multi-model inference approach using cumulative weights, w+(j), and ii) the proportion of variance explained in distance-based linear models. Variable Weights w+(j) Proportion of Variance Explained Variable Assemblage Structure Catch No Species Assemblage Structure Catch No. Species Trampling 0.67 # 0.77 # 0.99 # 0.26 ** 0.34 * 0.54 ** Wrack Cover 0.31 0.39 0.97 # 0.10 0.20 0.42 * Deployment Duration 0.34 0.20 0.27 0.13 0.13 0.40 * Sediment Compactness 0.32 0.15 0.19 0.10 0.00 0.00 Grain Size 0.20 0.24 0.12 0.05 0.11 0.14 * P < 0.05, ** P < 0.01; marginal tests in DISTLM # included in best overall model based on lowest AICc value Experimental trampling was strongly linked to variations in assemblage structure of surface-active invertebrates, and substantially reduced both the size and diversity of catches (Figs 3 and 4). In plots that were not, or lightly, trampled we caught 28 species, declining to 22 species under more intense (> 50 steps m-2) foot traffic (Fig 4). Mean density of all species declined by half over the range of experimental trampling applied (Fig 4). Negative effects of trampling on invertebrate abundance were most pronounced for rare species that declined at significantly greater rates in trampled plots: many taxa that occurred at initially low numbers were not found in the more heavily trampled plots. Effect sizes for declines in species richness were highest for the least and the most abundant species (Fig 4). 10.1371/journal.pone.0161905.g003Fig 3 Ordination (non-metric multidimensional scaling) of experimental plots based on similarity (Bray Curtis) in assemblage structure (i.e. species composition and abundance of species) of surface-active invertebrates. Distance between plot symbols is scaled to similarity (i.e. nearby samples have more similar invertebrate catches) and the size of segments is proportional to the value of an environmental or experimental variable and illustrates relative influence (cf. Table 1 for actual model statistics). 10.1371/journal.pone.0161905.g004Fig 4 Changes in abundance (left column) and species richness (right column) of surface-active invertebrates in relation to experimental trampling on the upper part of an ocean-exposed beach (statistics inside top panels are for linear regression analysis). Bottom row illustrates trampling effect sizes (indexed as the slope (+/- se) of linear regressions of abundance/species richness vs the number of steps) for logarithmic abundance classes of species. Nearly half of the shift in assemblage structure observed under experimental trampling was driven by sizeable (-2.4 x) reductions in the density of the talitrid amphipod Bellorchestia sp. 1 (Table 2). All other taxa occurred at considerably lower abundance and accounted for comparatively smaller proportions of assemblage change: three species of beetles, an oniscoid isopod, ants, dipterans, a second species of talitrid amphipod and beetle larvae each accounted for > 1% of dissimilarity in assemblage structure observed between the trampled and the control plots (Table 2). 10.1371/journal.pone.0161905.t002Table 2 Summary of similarity percentage analysis (SIMPER) listing species that cumulatively contributed 90% to the dissimilarity (Bray Curtis) in community structure between trampled and un-trampled plots. Species (Higher Taxon) Control—Mean Abundance (ind. trap-1) Trampled Mean Abundance (ind. trap-1) Avg. Dissimilarity Diss / SD Contribution (%) Bellorchestia sp.1 (Amphipoda, Talitridae) 23.00 9.52 21.79 2.11 46.45 Phycosecis litoralis (Coleoptera-Phycosecidae) 1.25 5.01 5.73 2.99 12.21 Diptera sp.3 3.88 0.99 4.57 2.75 9.73 Sartallus signatus (Coleoptera, Staphylinidae) 3.00 1.51 2.47 1.83 5.25 Lycosid spider sp.1 (Araneae, Lycosidae) 0.38 1.72 2.16 0.61 4.60 Oxyteline cf. Carpelimus sp.1 (Coleoptera, Staphylinidae) 1.75 0.40 2.10 5.89 4.48 Actaecia thomsoni (Isopoda, Actaeciidae) 2.08 1.24 1.37 1.75 2.92 Phycosecis spp. Larvae (Coleoptera-Phycosecidae) 0.42 1.25 1.30 1.53 2.77 Bellorchestia sp.2 (Amphipoda, Talitridae) 0.46 0.84 0.66 0.99 1.41 Ant sp.1 (Hymenoptera, Formicidae) 0.50 0.08 0.66 5.38 1.41 Discussion Anthropogenic habitat change resulting from recreational activities is widespread on sandy coastlines [4]. Here, the ecological consequences of disturbance caused by vehicles and pedestrian trampling are extensive and well-documented for dunes and their biota [33]. By contrast, on the non-vegetated part of the shore, only impacts caused by vehicles have been comprehensively quantified whereas the effects of foot traffic are generally more sparsely covered [34–36]. We addressed this gap with a high-intensity, but short-term, experiment demonstrating acute ecological effects resulting from beach walkers. Detrimental ecological effects of physical impacts from vehicles are generally manifested over large areas and are persistent [37, 38]. By comparison, trampling effects can, on lightly used beaches, be more localized and transient [35], but lead to persistent reductions in invertebrates on more intensively used shores [5, 33, 35, 39]. Thus, whether trampling will have negative consequences for shorebirds will largely depend on the intensity of human foot traffic, particularly during the breeding season. Our focal species, the Hooded Plover, feeds extensively on the invertebrates observed to be most heavily impacted by trampling in our experiment. When feeding on ocean beaches, these plovers use both darting and foot-trembling as foraging tactics, and regularly chase and consume surface-active talitrid amphipods [14]. Importantly, sand hoppers constituted 43% of all individuals caught in our experiment and the abundance of this key prey item declined substantially with foot traffic (from a total catch of 563 individuals in the reference plot to 260 individuals over the range of trampling intensities tested). Thus, human foot traffic reduced, at least in the short-term, the main food source of a threatened bird species, which is the target for considerable conservation investment. Because energy requirements in birds are particularly high during nesting and brood-rearing, there are two plausible consequences of trampling-induced declines in food availability. At the level of nesting pairs, young-of-the-year birds could experience lower fitness. The precocial chicks of beach-nesting birds are flightless and require access to abundant local food resources to promote rapid growth and development [40]. Those not acquiring adequate resources in this critical growth phase experience poor first-year survival [41][42]. At the population level, trampling effects may exacerbate existing density-dependent effects of food availability and ultimately reduce overall carrying capacity of target species. Solitary beach-nesting bird species can exhibit a clumped dispersion where food resources are abundant. In these areas, territory size is small because adults can secure enough resources for the brood within a localized area. Where prey availability is limited, however, competition and territory size increase, reducing the number of pairs using a given habitat. At carrying capacity, there can be no sustained increase in population size, as survival and reproductive success are severely constrained. The ‘standard repertoire’ of beach-bird conservation usually encompasses any combination of the following five actions: i) restrict access (seasonal or spatial beach closures), ii) educate beach users (etc. signs, awareness campaigns), iii) control predators; iv) provide extra shelter; and v) enhance or restore habitat [24, 43]. These interventions generally have mixed success, chiefly because invasive species control is incomplete (save for small islands) and because compliance of humans with control measures or environmental messages can be low [44–47]. Our results demonstrate that human trampling might be a cryptic factor that limits the success of beach-nesting bird conservation, and highlight the need for fresh approaches to management. We propose two actions that may partially offset the negative impacts of trampling on threatened beach-nesting birds: (1) wrack relocation; and (2) invertebrate augmentation. Wrack supports invertebrate prey for birds through the creation of habitat and the supply of food [19, 48]. As a component of recreational management on beaches, wrack is often removed from beaches and discarded. These spoils may, however, be deposited in or near breeding territories of plovers (following appropriate protocols) to enhance invertebrate food resources. Wrack relocation and deposition could have the added benefit of reducing disturbance to breeding birds because walkers may be deterred by wrack piles placed strategically near nests. An alternative strategy for locations where trampling is very intense and the supply of wrack is limited could be to directly supplement invertebrate prey to the breeding territories of beach-nesting birds. Many plovers on ocean beaches have high site fidelity and limited movement [19, 23]. Theoretically, it may therefore be possible to obtain invertebrate prey by harvesting limited numbers of individuals from sites that are consistently not used by plovers and augment these to existing breeding territories. Before this is considered as a conservation tool, it needs to be demonstrated that this action does not significantly deplete invertebrate populations at harvest sites and that any augmented prey will be available to plovers for a meaningful periods of time. Furthermore, whether this method is cost-effective and practicable needs to be tested. It is critically important to note that neither wrack relocation nor invertebrate augmentation are likely to replace more traditional approaches that restrict access and provide nest shelter, but both techniques can be explored as a useful complement, particularly for beaches that support prime breeding habitat and those where human compliance with regulations is particularly poor. The ecological impacts of foot traffic on ocean beaches will in most situations depend on a combination of: a) the number of people traversing a beach (physical intensity), b) the types of activities undertaken (e.g. walking or games); c) the frequency and temporal mode at which trampling occurs (e.g. pulsed, dispersed); and d) the time of the year especially with regards to the breeding season of species sensitive to disturbance. Furthermore, the action of trampling is usually not conducted in isolation from other human pressures on beaches, and in many cases it is accompanied by direct disturbance from humans, at times amplified by their companion dogs or vehicles used to travel on beaches [22, 44, 49–51]. Thus, trampling can exert at least three modes of pressure on beach birds: i) declines in food resources (crushing of invertebrates), direct disturbance by humans (investment in escape and alter behaviour), and disturbance impacts from vehicles and dogs [52, 53]. All these factors are amenable to specific experimental tests in the future. Experiments will be particularly useful in situations where certain information is required to inform conservation decisions (e.g. thresholds of impact) or where potential management scenarios are to be assessed (e.g. temporal or spatial restrictions, number of walkers etc.). A major challenge for experiments that seek to quantify the magnitude of human impacts is to ensure that additional environmental harm is kept to a minimum. Here we have shown that trampling effects on beaches can limit food availability for beach-nesting birds, which in turn may reduce the carrying capacity of beaches and lead to increased mortality of nesting individuals and their offspring. It will, therefore, be critical to identify the thresholds of pedestrian trampling beyond which declines in invertebrates become permanent and propagate to detrimental effects on threatened birds more broadly. Supporting Information S1 Table Containing complete taxa x plot data. (PDF) Click here for additional data file. Thank-you to all the field volunteers, Rowan Walker and Aidan Griffiths. This project was conducted under Department of Environment, Water, Land and Planning Permit No. 10007183. We greatly appreciate that Glenn Ehmke made the photo of the plover chick available to us. ==== Refs References 1 Huijbers CM , Schlacher TA , McVeigh RR , Schoeman DS , Olds AD , Brown MB , et al Functional replacement across species pools of vertebrate scavengers separated at a continental scale maintains an ecosystem function . Functional Ecology . 2016 ;30 (6 ):998 –1005 . 10.1111/1365-2435.12577 2 Schlacher TA , Weston MA , Schoeman DS , Olds AD , Huijbers CM , Connolly RM . Golden opportunities: A horizon scan to expand sandy beach ecology . Estuarine, Coastal and Shelf Science . 2015 ;157 (0 ):1 –6 . 10.1016/j.ecss.2015.02.002 . 3 Schlacher TA , Dugan J , Schoeman DS , Lastra M , Jones A , Scapini F , et al Sandy beaches at the brink . Diversity & Distributions . 2007 ;13 :556 –60 . 10.1111/j.1472-4642.2007.00363.x 4 Schlacher TA , Lucrezi S , Connolly RM , Peterson CH , Gilby BL , Maslo B , et al Human threats to sandy beaches: A meta-analysis of ghost crabs illustrates global anthropogenic impacts . Estuarine, Coastal and Shelf Science . 2016 ;169 :56 –73 . 10.1016/j.ecss.2015.11.025 . 5 Reyes-Martínez MJ , Ruíz-Delgado MC , Sánchez-Moyano JE , García-García FJ . Response of intertidal sandy-beach macrofauna to human trampling: An urban vs. natural beach system approach . Marine Environmental Research . 2015 ;103 :36 –45 . 10.1016/j.marenvres.2014.11.005 25460060 6 Lucrezi S , Schlacher TA , Walker SJ . Monitoring human impacts on sandy shore ecosystems: a test of ghost crabs (Ocypode spp.) as biological indicators on an urban beach . Environmental Monitoring and Assessment . 2009 ;152 :413 –24 . 10.1007/s10661-008-0326-2 18563608 7 Moffett MD , McLachlan A , Winter PED , De Ruyck AMC . Impact of trampling on sandy beach macrofauna . Journal of Coastal Conservation . 1998 ;4 (1 ):87 –90 . 8 Schlacher TA , Thompson L . Beach recreation impacts benthic invertebrates on ocean-exposed sandy shores . Biological Conservation . 2012 ;147 :123 –32 . 9 Maslo B , Burger J , Handel SN . Modeling foraging behavior of piping plovers to evaluate habitat restoration success . Journal of Wildlife Management . 2012 ;76 (1 ):181 –8 . 10.1002/jwmg.210 . WOS:000298586200022. 10 Rafael Barboza F , Defeo O . Global diversity patterns in sandy beach macrofauna: A biogeographic analysis . Scientific Reports . 2015 ;5 :Article number 14515. 10.1038/srep14515 11 Schlacher TA , Weston MA , Schoeman DS , Olds AD , Huijbers CM , Connolly RM . Golden opportunities: a horizon scan to expand sandy beach ecology . Estuarine, Coastal and Shelf Science . 2015 ;157 :1 –6 . 10.1016/j.ecss.2015.02.002 12 Lastra M , Schlacher TA , Olabarria C . Niche segregation in sandy beach animals: an analysis with surface-active peracarid crustaceans on the Atlantic coast of Spain . Marine Biology . 2010 ;157 :613 –25 . 10.1007/s00227-009-1346-3 13 Eberhart-Phillips LJ , Colwell MA . Conservation challenges of a sink: The viability of an isolated population of the Snowy Plover . Bird Conservation International . 2014 ;24 (3 ):327 –41 . 10.1017/S0959270913000506 14 Weston MA . The foraging and diet of non-breeding hooded plovers Thinornis rubricollis in relation to habitat type . Journal of the Royal Society of Western Australia . 2007 ;90 (2 ):89 –95 . 15 Schlacher TA , Meager JJ , Nielsen T . Habitat selection in birds feeding on ocean shores: landscape effects are important in the choice of foraging sites by oystercatchers . Marine Ecology . 2014 ;35 (1 ):67 –76 . 10.1111/maec.12055 16 Tulp I , Schekkerman H , Bruinzeel LW , Jukema J , Visser GH , Piersma T . Energetic demands during incubation and chick rearing in a uniparental and a biparental shorebird breeding in the high arctic . Auk . 2009 ;126 (1 ):155 –64 . 10.1525/auk.2009.07181 17 Majka CG , Shaffer F . Beetles (Coleoptera) in the diet of Piping Plovers in the Iles de la Madeleine, Québec, Canada . Wader Study Group Bulletin . 2008 ;115 (2 ):77 –83 . 18 Shaffer F , Laporte P . Diet of piping plovers on the Magdalen Islands, Quebec . The Wilson Bulletin . 1994 :531 –6 . 19 Brindock KM , Colwell MA . Habitat selection by western snowy plovers during the nonbreeding season . Journal of Wildlife Management . 2011 ;75 (4 ):786 –93 . 10.1002/jwmg.106 20 Van Der Merwe D , McLachlan A , De Ruyck AMC . Partitioning of feeding habitats by whitefronted plovers charadrios marginatvs at a dune/beach interface . Ostrich . 1992 ;63 (2–3 ):49 –57 . 10.1080/00306525.1992.9633948 21 Weston MA , Ehmke GC , Maguire GS . Manage one beach or two? Movements and space-use of the threatened hooded plover (Thinornis rubricollis) in south-eastern Australia . Wildlife Research . 2009 ;36 (4 ):289 –98 . 22 Meager JJ , Schlacher TA , Nielsen T . Humans alter habitat selection of birds on ocean-exposed sandy beaches . Diversity & Distributions . 2012 ;18 (3 ):294 –306 . 23 Cuttriss A , Maguire GS , Ehmke G , Weston MA . Breeding habitat selection in an obligate beach bird: A test of the food resource hypothesis . Marine and Freshwater Research . 2015 ;66 (9 ):841 –6 . 10.1071/MF14213 24 Maslo B , Lockwood JL . Evidence-based decisions on the use of predator exclosures in shorebird conservation . Biological Conservation . 2009 ;142 (12 ):3213 –8 . 10.1016/j.biocon.2009.07.034 . ISI:000272100100047. 25 Maguire GS , Duivenvoorden AK , Weston MA , Adams R . Provision of artificial shelter on beaches is associated with improved shorebird fledging success . Bird Conservation International . 2011 ;21 (2 ):172 –85 . 10.1017/s0959270910000420 . WOS:000291054500005. 26 Dowling B , Weston MA . Managing a breeding population of the Hooded Plover Thinornis rubricollis in a high-use recreational environment . Bird Conservation International . 1999 ;9 (4 ):255 –70 . 27 McArdle BH , Anderson MJ . Fitting multivariate models to community data: A comment on distance-based redundancy analysis . Ecology . 2001 ;82 (1 ):290 –7 . 28 Clarke KR , Gorley RN . PRIMER v6: User Manual/Tutorial . Plymouth, UK : PRIMER-E ; 2006 190 p. 29 Symonds MRE , Moussalli A . A brief guide to model selection, multimodel inference and model averaging in behavioural ecology using Akaike’s information criterion . Behavioral Ecology and Sociobiology . 2011 :1 –9 . 30 Burnham KP , Anderson DR , Huyvaert KP . AIC model selection and multimodel inference in behavioral ecology: some background, observations, and comparisons . Behavioral Ecology and Sociobiology . 2011 :1 –13 . 31 Anderson DR . Model based inference in the life sciences: a primer on evidence . New York : Springer Verlag ; 2008 . 32 Burnham KP , Anderson DR . Model Selection and Multi-Model Inference: A Practical Information-Theoretic Approach . 2 ed New York : Springer ; 2002 . 33 Schlacher TA , Schoeman DS , Jones AR , Dugan JE , Hubbard DM , Defeo O , et al Metrics to assess ecological condition, change, and impacts in sandy beach ecosystems . Journal of Environmental Management . 2014 ;144 :322 –35 . 10.1016/j.jenvman.2014.05.036 25014753 34 Veloso VG , Neves G , Lozano M , Perez-hurtado A , Gago CG , Hortas F , et al Responses of talitrid amphipods to a gradient of recreational pressure caused by beach urbanization . Marine Ecology—Evolutionary Perspective . 2008 ;29 (S1 ):126 –33 . 35 Lucrezi S , Schlacher TA , Robinson W . Human disturbance as a cause of bias in ecological indicators for sandy beaches: Experimental evidence for the effects of human trampling on ghost crabs (Ocypode spp.) . Ecological Indicators . 2009 ;9 :913 –21 . 10.1016/j.ecolind.2008.10.013 36 Jaramillo E , Contreras H , Quijon P . Macroinfauna and human disturbance in a sandy beach of South-Central Chile . Revista Chilena de Historia Natural . 1996 ;69 (4 ):655 –63 . WR251-0019 WR251: Document Delivery available. 37 Schlacher TA , Richardson D , McLean I . Impacts of off-road vehicles (ORVs) on macrobenthic assemblages on sandy beaches . Environmental Management . 2008 ;41 :878 –92 . 10.1007/s00267-008-9071-0 18266026 38 Schlacher TA , Thompson LMC , Price S . Vehicles versus conservation of invertebrates on sandy beaches: quantifying direct mortalities inflicted by off-road vehicles (ORVs) on ghost crabs . Marine Ecology . 2007 ;28 :354 –67 . 10.1111/j.1439-0485.2007.00156.x 39 Lucrezi S , Schlacher TA , Robinson W . Can storms and shore armouring exert additive effects on sandy beach habitats and biota? Marine & Freshwater Research . 2010 ;61 :951 –62 . 10.1071/MF09259 40 Cairns WE . Biology and behavior of breeding Piping Plovers . The Wilson Bulletin . 1982 :531 –45 . 41 Fer DL , Fraser JD , Kruse CD . Piping plover chick foraging, growth, and survival in the Great Plains . The Journal of Wildlife Management . 2008 ;72 (3 ):682 –7 . 42 Weston MA , Elgar MA . Parental care in Hooded Plovers (Thinornis rubricollis) . Emu . 2005 ;105 (4 ):283 –92 . 43 Maslo B , Handel SN , Pover T . Restoring beaches for Atlantic Coast Piping Plovers (Charadrius melodus): A classification and regression tree analysis of nest-site selection . Restoration Ecology . 2011 ;19 :194 –203 . 10.1111/j.1526-100X.2010.00709.x . ISI:000287974900007. 44 Schlacher TA , Weston MA , Lynn D , Schoeman DS , Huijbers CM , Olds AD , et al Conservation gone to the dogs: when canids rule the beach in small coastal reserves . Biodiversity and Conservation . 2015 ;24 (3 ):493 –509 . 10.1007/s10531-014-0830-3 45 Weston MA , Schlacher TA , Lynn D . Pro-environmental beach driving is uncommon and ineffective in reducing disturbance to beach-dwelling birds . Environmental Management . 2014 ;53 (5 ):999 –1004 . 10.1007/s00267-014-0256-4 24599507 46 Schlacher TA , Nielsen T , Weston MA . Human recreation alters behaviour profiles of non-breeding birds on open-coast sandy shores . Estuarine, Coastal and Shelf Science . 2013 ;118 :31 –42 . 47 Schlacher TA , Weston MA , Lynn DD , Connolly RM . Setback distances as a conservation tool in wildlife-human interactions: testing their efficacy for birds affected by vehicles on open-coast sandy beaches . PLoS ONE . 2013 ;8 (9 ):e71200 10.1371/journal.pone.0071200 24039711 48 Dugan JE , Hubbard DM , McCrary MD , Pierson MO . The response of macrofauna communities and shorebirds to macrophyte wrack subsidies on exposed sandy beaches of southern California . Estuarine, Coastal and Shelf Science . 2003 ;58 (Suppl S ):25 –40 . 744XZ-0004 744XZ: Document Delivery available. 49 Weston MA , Stankowich T . Dogs as agents of disturbance In: Gompper ME , editor. Free-Ranging Dogs and Wildlife Conservation . Oxford : Oxford University Press ; 2014 p. 94 –116 . 50 Weston MA , Schlacher TA , Lynn D . Pro-environmental beach driving is uncommon and ineffective in reducing disturbance to beach-dwelling birds . Environmental Management . 2014 ;53 :999 –1004 . 10.1007/s00267-014-0256-4 24599507 51 Schlacher TA , Nielsen T , Weston MA . Human recreation alters behaviour profiles of non-breeding birds on open-coast sandy shores . Estuarine, Coastal and Shelf Science . 2013 ;118 :31 –42 . 10.1016/j.ecss.2012.12.016 52 Stillman RA , Goss-Custard JD . Seasonal changes in the response of Oystercatchers Haematopus ostralegus to human disturbance . Journal of Avian Biology . 2002 ;33 (4 ):358 –65 . 53 Stillman RA , West AD , Caldow RWG , Durell SEALV . Predicting the effect of disturbance on coastal birds . Ibis . 2007 ;149 :73 –81 .

PMC005xxxxxx/PMC5001727.txt

==== Front PLoS Comput BiolPLoS Comput. BiolplosploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, CA USA 2756498710.1371/journal.pcbi.1005064PCOMPBIOL-D-15-01248Research ArticleBiology and Life SciencesGeneticsGene ExpressionBiology and Life SciencesComputational BiologyGenome AnalysisGene OntologiesBiology and Life SciencesGeneticsGenomicsGenome AnalysisGene OntologiesPeople and PlacesPopulation GroupingsAge GroupsAdultsBiology and Life SciencesCell BiologyCellular TypesAnimal CellsNeuronsBiology and Life SciencesNeuroscienceCellular NeuroscienceNeuronsBiology and Life SciencesEvolutionary BiologyEvolutionary GeneticsBiology and Life SciencesDevelopmental BiologyOrganism DevelopmentOrganogenesisBrain DevelopmentBiology and Life SciencesNeuroscienceCellular NeuroscienceAxon GuidanceBiology and Life SciencesNeuroscienceDevelopmental NeuroscienceAxon GuidanceBiology and Life SciencesDevelopmental BiologyCell DifferentiationNeuronal DifferentiationOn Expression Patterns and Developmental Origin of Human Brain Regions On Expression Patterns and Developmental Origin of Human Brain RegionsKirsch Lior *Chechik Gal ¤The Gonda Multidisciplinary Brain Research Center, Bar-Ilan University, Ramat Gan, IsraelKreiman Gabriel EditorCHB, Harvard Medical School, UNITED STATESThe authors have declared that no competing interests exist. Conceived and designed the experiments: LK GC. Performed the experiments: LK GC. Analyzed the data: LK GC. Contributed reagents/materials/analysis tools: LK GC. Wrote the paper: LK GC. ¤ Current address: Google research, Mountain View, California, United States of America * E-mail: lior.kirsch.biu@gmail.com26 8 2016 8 2016 12 8 e100506426 7 2015 16 7 2016 © 2016 Kirsch, Chechik2016Kirsch, ChechikThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Anatomical substructures of the human brain have characteristic cell-types, connectivity and local circuitry, which are reflected in area-specific transcriptome signatures, but the principles governing area-specific transcription and their relation to brain development are still being studied. In adult rodents, areal transcriptome patterns agree with the embryonic origin of brain regions, but the processes and genes that preserve an embryonic signature in regional expression profiles were not quantified. Furthermore, it is not clear how embryonic-origin signatures of adult-brain expression interplay with changes in expression patterns during development. Here we first quantify which genes have regional expression-patterns related to the developmental origin of brain regions, using genome-wide mRNA expression from post-mortem adult human brains. We find that almost all human genes (92%) exhibit an expression pattern that agrees with developmental brain-region ontology, but that this agreement changes at multiple phases during development. Agreement is particularly strong in neuron-specific genes, but also in genes that are not spatially correlated with neuron-specific or glia-specific markers. Surprisingly, agreement is also stronger in early-evolved genes. We further find that pairs of similar genes having high agreement to developmental region ontology tend to be more strongly correlated or anti-correlated, and that the strength of spatial correlation changes more strongly in gene pairs with stronger embryonic signatures. These results suggest that transcription regulation of most genes in the adult human brain is spatially tuned in a way that changes through life, but in agreement with development-determined brain regions. Author Summary Genome-wide measurements of gene expression across the human brain can reveal new principles of brain organization and function. To achieve this, we aim to discover which genes are differentially expressed and in what brain regions. We found that almost all genes in the adult human brain bear a developmental ‘footprint’ which determines their areal expression-pattern based on the developmental ontology of brain regions, while at the same time their spatial expression pattern changes during life. Furthermore, pairs of paralog genes and similar genes with stronger embryonic footprint, tend to be more strongly correlated (or anti correlated) suggesting that their expression is more strongly spatially tuned, and this tuning changes in development. http://dx.doi.org/10.13039/501100003977Israel Science Foundation1090/12Chechik Gal Marie Curie reintegration grantPIRG06-GA-2009-256566Chechik Gal GC was supported by the Israeli Science Foundation grant #1090/12, and by a Marie Curie reintegration grant #PIRG06-GA-2009-256566. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction The human brain is organized in a hierarchy of multiple substructures, whose cell composition and circuitry are believed to allow each substructure to carry out its distinct function. While physiological and histological differences and similarities between structures have been intensively studied [1–4], the molecular profiles giving rise to those differences are far from being understood. Specifically, it is not known which principles govern the expression patterns of genes across the adult brain and what determines their spatial organization. Recent high-resolution genome-wide transcriptome profiling studies allow addressing these questions [5,6]. The current paper explores the role of development in determining adult expression patterns. In early development of the vertebrate nervous system, the posterior part of the neural tube develops into the spinal cord, and its anterior part divides into three primary vesicles: the prosencephalon, the mesencephalon and the rombencephalon. The prosencephalon further develops into two secondary vesicles: the telencephalon and the diencephalon. The most posterior vesicle, the rombencephalon, forms two secondary vesicles as well, the metencephalon, and the myelencephalon. These five vesicles are aligned along the rostral-caudal axis of the developing brain and establish the primary organization of the central nervous system (Fig 1A) [7]. 10.1371/journal.pcbi.1005064.g001Fig 1 Brain-region ontology and the BRO-agreement score. (A) Illustration of the ontology region tree showing 16 brain structures studied. The full ontology contains 1534 regions, not shown. (B) A 3D model brain illustrating 16 brain regions using the same colors as in A. The left cortex is not shown in order to expose the inner structures. (C) Hierarchical clustering of 16 human brain structures. Agglomerative linking of regions by their average expression profile yields a tree structure that agrees the with ontology tree. The color above the region name matches the colors in the region ontology tree in Fig 1A. (D) The joint distribution of expression distances and ontology distances across all pairs of tissue samples, as computed for the gene NEUROD1. The two distance measures are strongly correlated (Spearman ρ = 0.65, n = 6.85M, p-value < 10−15), showing that the spatial expression pattern agrees with the ontology. During early development, several gene families exhibit distinct spatial expression patterns [8], including, for example, genes involved in axon guidance and in segmentation and compartmentalization of brain regions [9,10]. Genes including netrins, semaphorins, and ephrins function as molecular signatures, guiding axons to form long-range connections [11–13]. Patterning genes [14], like the Hox gene family, play a crucial role in forming brain regions [15]. Some genes, including Hox genes, were shown to retain unique expression patterns related to tissue specificity across the adult body [16–18]. Many other genes, change their expression patterns during development, both before or after birth [19–21]. However, for most genes that are not directly involved in brain development, it is usually not known which processes determine their spatial expression pattern across the adult brain. Zapala and colleagues [22] studied areal expression across the adult mouse brain by clustering brain regions based on the expression levels of ~2000 genes. They found that the pattern of gene expression in a brain region was correlated with the position of that region along the anterior-posterior axis of the neural tube. Regions with similar expression profiles often shared a common embryonic origin. One possible explanation of this finding would be that brain regions sharing an embryonic origin contain similar cell-types, which in turn express shared gene markers. For example, pyramidal neurons are the main excitatory neurons both in the neocortex and in the hippocampus. The agreement between embryonic origin and expression similarity may therefore be explained by a small number of cell-type specific gene markers. Indeed, French et al. have shown that regions with similar expression profiles tend to have similar neuronal connectivity patterns [23]. Ko et al. have shown that neuron-specific and astrocyte-specific gene markers show distinct patterns of expression across brain regions [24]. Grange et al. have estimated the spatial densities of 64 cell types in the mouse brain and identified genes with a localized pattern of expression [25]. The question remains however, how genome-wide is the agreement of expression-pattern with embryonic origin and how does this agreement develop during the lifetime of an organism. It is also not clear to what extent expression differences are pronounced between sub regions of the major brain structures. Specifically, the cortex is sometimes viewed as largely homogeneous in various properties, including neuronal density, connectivity patterns and distribution of cell-types, properties which were shown to be consistent across cortical sub-regions [4,26]. Also, cortical microcircuits can be induced to perform processing tasks that are naturally performed in other regions [27–30]. These views are in agreement with the results of Ko et al. [24], where clustering brain voxels based on their expression profile showed that the cortex is significantly different from other structures, but is largely homogeneous by itself. However, the level of homogeneity of cortex expression patterns has not been quantified directly so far. To quantify the relation between expression patterns across the adult human brain and the embryonic origin of the corresponding brain regions on a gene-by-gene basis, we analyze two genome-wide mRNA expression datasets. For each gene separately, we computed an index that measures how its expression pattern in the adult agrees with a brain-region ontology based on embryonic brain development. Surprisingly, we found that almost all genes exhibit a spatial expression pattern that significantly agrees with the brain-region ontology. This effect is particularly strong in neuron-specific genes as expected, but also in many genes that serve more generic functions. This suggests that the brain tunes the areal expression pattern of genes in a way that strongly depends on embryonic development, and this holds even for genes that participate in brain-wide functions. Furthermore, pairs of genes sharing related functions tend to be spatially more anti-correlated if their expression pattern agrees with the brain-region ontology. Results To characterize distinct gene expression patterns across the adult human brain, we analyzed genome-wide expression measurements from two sources. First, a set of 3702 samples from 6 adult post-mortem human brains [5], and second, a set of 491 samples from 20 adult post-mortem human brains [6]. See Methods for details on both datasets. In the results below, we refer to these datasets as ABA6-2013 and Kang-2011. Spatial pattern of expression and agreement with brain-region ontology We evaluated how expression of each individual gene across the brain agrees with an ontology of brain regions provided by the Allen Institute (http://human.brain-map.org) [31]. This ontology is coarsely based on brain development, covering both developing and adult human brain structures. The fine structure of the ontology, including cortical parcellation, is based on classical cytoarchitecture. Instead of analyzing expression variability across regions using a ‘flat’ representation of regions, the ontology allows to take into account the ‘structured’ similarities among regions stemming from the shared embryonic origin of regions. We used the full tree ontology which contained 1534 brain regions. Brain samples from the ABA6-2013 dataset were associated with 414 ontology regions, and those from the Kang-2011 dataset were associated with 16 regions of the ontology (see Methods S2 and S3 Tables). Fig 1A depicts the region-ontology tree at a coarse resolution for visualization purposes. Nodes in the tree are colored blue-to-red, roughly corresponding to position of regions on the anterior-posterior axis. Fig 1B depicts the same regions on a 3D model of a human brain using the corresponding colors. In the mouse brain, regions that share similar expression patterns group in a way that matches the region ontology [22,32]. To test if these result are reproduced in human, we clustered brain regions based on the full-genome expression profile (see Methods). The resulting clustering agrees with the findings in the mouse. Fig 1C shows the hierarchical clustering of 16 brain regions from ABA6-2013, where brain regions with a common developmental origin, share similar expression patterns. However, this clustering depends on the joint expression patterns of all genes, and the question remains: which genes and processes contribute to this effect? To quantify how the expression pattern of each individual gene agrees with the region ontology, we defined an index, which we call Brain-Region Ontology agreement score (BRO-agreement score), calculated as follows. For a given gene, we consider all pairs of tissue-samples, and for each pair we computed two measures of distances. The first, expression distance, is the absolute difference of expression values in the two samples. The second, ontology distance, is the distance between the corresponding regions in the ontology tree. The BRO score for each gene is defined as the spearman correlation between the two distances computed for all tissue pairs. It provides a measure of the agreement between expression difference and ontology distance (for more details see Methods). We also tested a second index based on triplet ranking which yielded similar results (see Methods). To illustrate the use of the BRO score, consider the expression pattern of the gene NEUROD1, a transcription factor involved in regulation of brain development. Fig 1D depicts the joint distribution of the ontology distance and the expression distance computed for NEUROD1 expression measured at all tissue pairs. The two distance measures are strongly correlated, with a BRO score of 0.65, suggesting that the variability in the expression of NEUROD1 across the adult brain is largely explained by the position of the region in the development region ontology. BRO_score(gene i):=spearman corr(a,b)∈all sample pairs(dtree(a,b), dexpression(ai, bi)) We computed the BRO index for every gene in the ABA6-2013 dataset based on a fine-resolution ontology of 414 brain regions (see Methods). To assess significance, the BRO score of each gene was compared to a randomized BRO score obtained by permuting the expression profile of the gene across regions. We find that 92% of genes significantly agree with the brain-region ontology more than random (FDR–corrected, q-value < 0.01). Using the triplets score (see Methods), 95% of the genes in this dataset were significant. This surprisingly-high fraction suggests that most genes have distinct areal expression patterns across the adult brain, and that these patterns are largely determined by the embryonic origin of the brain regions they are expressed in. To test reproducibility and robustness of these results, we compare the BRO index obtained for every gene in the two microarray datasets (see Methods), covering a total of 26 postmortem brains. Fig 2A shows the joint distribution of BRO-agreement scores (light dots) computed in ABA6-2013 (abscissa, 414 regions) and Kang-2011 (ordinate, 16 regions). It also shows the distribution of the baseline random distribution that was generated using permutation test (Fig 2A, dark dots, see Methods). We further tested that the effect was robust across subjects (see Methods and supplemental S7 Fig) and compared the results obtained with the BRO index to those obtained with ANOVA (see Methods). 10.1371/journal.pcbi.1005064.g002Fig 2 Distribution of BRO-agreement scores of individual genes. (A) Heatmap showing the joint distribution of BRO-agreement scores of all genes in 16 regions of ABA6-2013 (absica) and Kang-2011 (ordinate). Colors correspond to the density. (B) A scatter plot showing BRO-agreement scores for the two datasets in A. Each light-grey dot corresponds to a single gene (a total of 17K genes). Dark-grey dots correspond to permuted data (see Methods). The BRO scores are significantly correlated across the two datasets (Spearman, ρ = 0.53, n = 16947, p-value<10−16). (C) Marginal distribution of BRO scores in the ABA6-2013 dataset. BRO scores for most genes are significantly greater than randomized scores. (D) Same as C, for the Kang-2011 data. (E) BRO-agreement scores traced through life based on the full developmental dataset in Kang-2011. Samples are aggregated based on the developmental stages defined in [6]. Numbers above the line denote the number of subjects in each age group. In the Kang-2011 dataset, Fig 2A and 2D, 66% of the genes have significant BRO scores (same as with the triplets method). The BRO scores for the two datasets are significantly correlated across the two datasets (Spearman ρ = 0.53, p-value < 10-16, n = 16947, Fig 2A). The differences between the two datasets in the fraction of genes that reach the significance threshold is due to the smaller number of tissue samples in Kang-2011 and due to the limited coverage of non-cortical regions, since only 144 samples out of 491 (29%) were from non-cortical regions in Kang-2011, compared to 2187 out of 3702 (59%) in ABA6-2013. See full details in supplementary S1 Table. We provide BRO scores for all genes as a supplementary data file (S1 Data). We also quantified BRO scores in mouse, but due to multiple differences between the datasets, direct comparison of the results is hard (S1 Text). Variability across regions and samples may be due to tissue-wide effects such as fluctuations in sampling, biases in cell density or variability in cell-type proportions, and these may be correlated with the developmental ontology. To test for a possible effect of cell density fluctuations across samples, we repeated the experiment after scaling each sample by its mean expression across genes. Fluctuations in the mean expression across samples were small (coefficient of variation, CV = std/mean = ~2.5%) and did not have a significant effect on BRO scores (Spearman ρ = 0.997 comparing with and without scaling). Similar results were obtained when normalizing to a set of highly-stable genes (see Methods) (CV = ~3%). This type of normalization is effective when expression of the gene is highly stable and the relation between cell density and number of transcript is largely linear. As a second control, we aimed to test if BRO scores simply reflect the proportions of glia-neuron mixture, which varies across the brain. Since estimating cell-type proportions is challenging [25], and not yet available in human, we tested the relation between BRO-agreement scores and correlation with known cell-type specific markers, by quantifying how well the spatial variability of a gene can be explained by known neuronal and glial markers (see Methods). For neuronal markers, the median explained variance was 0.14 (median absolute deviation of 0.12, maximum explained variance = 0.26), and similar results were obtained with glial markers. One interpretation that is consistent with these finding, is that the mixture proportions of cell types in a sample, based on the markers available to us today, has a limited explaining power of spatial variability. It should be noted however, that cell type specific markers have both a limited sensitivity and limited specificity, and do not reflect perfectly cell type proportions. These results suggest that BRO-agreement is not a mere reflection of neuron-to-glia mixture proportions. The transcriptome of the cerebellum is known to differ substantially from that of the rest of the brain [5,21,33]. This is apparent in Fig 1C, where the cerebellum is well separated from other regions. To test how strongly the cerebellum contributes to the high BRO agreement scores, we recomputed the scores while excluding cerebellar samples. Even with cerebellar samples excluded, 90% of genes are BRO-significant. An intuition for the robustness of the BRO can be gained from a principle component analysis (PCA) in a subsequent section. The non-cerebellar structures are aligned along an anterior-posterior dimension when projected onto the first two principal components. The order along this axis agrees with their place of origin in the neural tube and is captured by the BRO-score. The BRO-agreement score reflects how strongly adult expression pattern of a gene may be driven by the ontology, which was coarsely based on brain development. It is natural to test how BRO-agreement scores change during life. We therefore computed the per-gene BRO-agreement scores in subjects of multiple ages based on the data of [6]. Samples were grouped as in [6], and covered postnatal and embryonic age groups starting at 13 post-conception week, all having a common set of brain regions. When considering ages from late embryonic development to adulthood, the mean BRO score follows an “hourglass” pattern, with BRO scores being lowest around birth (Fig 2E). This developmental pattern is in agreement with previous studies which analyzed aerial variability in the mouse [21] and human [19] brain. Based on these previous studies, the elevated BRO scores in embryonic development are due to higher areal variability in neural and brain development functions, while the postnatal rise in BRO scores is due to elevated variability in signaling and plasticity functions. The life-long BRO-agreement profile has two major differences from previous studies, namely, the BRO scores of both very early and very late stages are significantly lower, deviating significantly from the hourglass pattern. These differences are captured here because the scores are computed separately for very early embryonic and very late-life stages. Functional characterization of genes and their BRO-agreement scores What genes and functions achieve the highest BRO scores? To answer this question, we first tested functional enrichment of gene ontology (GO) categories [34]. We used a threshold-independent approach based on ranking genes by their BRO (mHG [35]). The top enriched biological processes (S4 Table) are all brain related, and mainly belong to two families of functions: cell-to-cell signaling like synaptic transmission, and development-related categories like neuron differentiation and neurogenesis. The first family, genes in cell-signaling categories, included both genes from generic signal transduction pathways and receptors of more specific neuromodulator systems. For instance, the two genes with highest BRO score in synaptic transmission were RASGRF2 –which coordinates activation of MAPK signaling, and CAMK2A—which is involved in calcium signaling). The 3rd and 4th genes in that category were genes coding for serotonin receptors HTR2A and HTR4. Finally, some of the top ranked genes are related to brain-related disorders: G-protein signaling regulator RGS4, serotonin receptor HTR2A and the postsynaptic protein encoder–NRGN are related to schizophrenia. HTR2A, CHH9 are related to autistic disorders. To understand better which genes achieve the highest BRO scores, we further studied gene families and functions that are of particular interest: genes expressed in specific cell types, and genes involved in regulation of brain development. First, it has been suggested that genes that are expressed distinctively in specific cell types, contribute significantly to expression differences among brain regions [24]. Specifically, the combined expression of neuronal gene markers was shown to correspond to the major subdivisions of the brain [24]. Cahoy et al [36] identified genes that are enriched in specific cell types in the mouse brain and can be used as cell-type specific markers. Operating under the assumption that the human orthologs of those markers have preserved spatial expression patterns, their BRO-scores provide a way to test quantitatively how strongly neuron-specific genes agree with the brain region ontology. Fig 3A depicts three subsets of cell-specific gene markers, all of which are in particularly strong agreement with the tree structure: markers for neurons, astrocytes and oligodendrocytes. Cell-type specific genes have higher BRO scores than other genes on average (Fig 3A bottom). Interestingly, neuron-specific gene markers agree more strongly with the region-ontology than the average gene (p-value < 10−70,Wilcoxon comparing the median of the distributions of BRO scores of neuron-specific genes, median = 0.33 and the general population of genes, median = 0.11) while oligodendrocytes-specific and astrocyte-specific markers are far less so (oligodendrocytes median = 0.19, Wilcoxon, p-value = 10−5, astrocytes median = 0.16, Wilcoxon, p-value = 10−3). These results are in agreement with Ko et al. [24,37,38], which showed that the combined expression pattern of genes that are cell specific agrees with the region-ontology. The analysis above extends their results by showing that the region-ontology agreement occurs at the level of individual genes and is prevalent across cell-specific markers (neurons: 270/271 which are 99% of the neuronal markers are BRO significant; astrocytes: 151/160 which are 94% of the astrocytes markers; oligodendrocytes: 103/106 which are 97% of the oligodendrocytes markers).Using the same set of markers, Tan et al. showed that genetic markers for neurons and oligodendrocyte are on the opposite ends of the first principal component [37]. This means that while neurons and oligodendrocyte are similar in that they both agree with the developmental ontology, they also show a very distinct pattern of spatial expression. 10.1371/journal.pcbi.1005064.g003Fig 3 The distribution of BRO-agreement scores on different subsets of genes. The two top panels show a scatter plot of BRO scores in ABA6-2013 and Kang-2011. The corresponding lower two panels show the (marginal) distributions in the ABA6-2013 dataset. (A) Cell-type specific genes have higher agreement scores than all genes (Wilcoxonon tail test; neurons median = 0.33: p-value < 10−70 oligoodendrocytes median = 0.19: p-value = 10−5, astrocytes median = 0.16: p-value = 10−3). (B) Axon guidance genes receive higher scores than general genes (Wilcoxon median = 0.21; p-value = 10−7). Hox genes are less in agreement with region-ontology than the full set of genes. 21 Hox genes are BRO significant (67%) (compared to the randomized scores, with alpha = 0.01). PAX2, PAX3 and PAX6 obtain high BRO scores. Fig 3B depicts the distribution of BRO scores for three gene families involved in regional specificity during brain development: axon guidance genes, Hox genes, and Pax genes. Genes involved in early brain developmental have been shown to have regional expression patterns in the adult [39,40][23]. Here, we find that genes involved in axon guidance have very high BRO scores (Higher than the average gene, Wilcoxon, median = 0.21 p-value = 10−7) (Fig 3B bottom). This suggests that beyond their embryonic role, genes involved in axon guidance may assume other functional roles in the adult brain. Second, Hox genes play a major role in anterior-posterior patterning across the body and across the brain during development and largely retain these patterns in the adult body [16]. Their role in the adult brain is less clear. Here we find that many Hox genes have BRO scores above the random set, but on average, their scores are lower than the average gene (Wilcoxon, median = 0.03 p-value = 10−9). This suggests that unlike other gene groups discussed above, Hox genes are less involved in regional patterns in the adult brain. These view is also supported by Takahashi et al. which observed that while Hox genes show an expression gradient through the entire adult body, only one third of Hox genes are differentially expressed in brain-specific tissues [16]. Finally, we examined Pax genes. These genes are involved in early regionalization of the embryo brain and were suggested to play a role in differentiation and maintenance of specific subsets of cells in the adult brain [41,42]. It has been shown before that genes important for brain developmental have regional expression patterns in the adult [39,40], including genes involved in brain connectivity [23]. Here we find that three Pax genes, PAX2, PAX3 and PAX6, obtain significantly large BRO scores (Fig 3B). Interestingly, PAX6 is a major determinant of regionalization in the mammalian brain [9,10,42]. It was shown to be essential to cortex development, to mark cortex regionalization and to regulate radial migration of neuronal precursors [39,43,44]. The differential areal expression pattern of PAX6 in the adult raises the hypothesis that PAX6 continues to play a region-specific role in the adult brain. Evolutionary gene age and BRO-agreement The above results suggest that spatial regionalization of human brain expression is present both in brain-specific functions and also in more generic processes that can be found in simpler organisms. Importantly, spatial regionalization of the nervous system is not unique to mammals or vertebrae, and some of the mechanisms controlling spatial patterning are shared across evolutionary-remote species [45]. For instance, Hox genes, whose expression exhibit anterior-posterior gradients in mammals, also hold spatial information in species that diverged from the human lineage early in evolution [46–48]. The natural question therefore arises: how is brain regionalization of a gene related to the evolutionary age of that gene? For instance, one may hypothesize that genes with high BRO agreement would be genes those that evolved recently, in organisms having a nervous system similar to the mammalian brain. To test this hypothesis, we compare the BRO index with an index quantifying the evolutionary age genes [49]. Surprisingly, we found that evolutionary-older genes have on average higher BRO scores than evolutionary-recent genes (Fig 4) (Cellular organisms: median = 0.129; Primates: median = 0.068, Wilcoxon p-value = 10−6). These older genes are also active in signaling pathways and other basic functions in the cell (S4 Table). The top BRO-scoring genes have orthologs across a wide variety of species, and participate in functions that are not specific to neural processes. Presumably, these genes were conserved as the result of a pressure to preserve these basic functions. For example, the gene ENC1 encodes an actin-binding protein involved in regulation of neuronal process formation and in differentiation of neural crest cells. As another example, CAMK2A is involved in calcium signaling as part of the NMDAR signaling complex. At the same time, CAMK2A has an early evolutionary origin and has orthologs even in rice. On the other range of the evolutionary timeline, genes associated with speciation of primates obtain lower BRO scores on average. These results suggest that genes with strong BRO scores and spatial patterns are not necessarily specific to neural processes, but rather that the brain spatially tunes the expression of genes involved in fundamental molecular functions. With that said, it is also possible that these newer genes exhibit more refined differences across brain regions, but that these changes are not captured by the current coarse-scale analysis (also compare with [50]). 10.1371/journal.pcbi.1005064.g004Fig 4 Evolutionary-older genes have on average higher BRO-scores. (A) The median BRO score as a function of the evolutionary age of genes. Older genes receive on average higher BRO scores than evolutionary recent genes. (B) Focus on the distribution of BRO scores for the oldest gene group and of the most recent (primates) gene group. Genes in the cellular organisim group have a median BRO-score of 0.129, while genes in the primate group have a median BRO-score of 0.068. The two distributions are signifatcly different.(Wilcoxonon two-tail test: p-value = 10−6). Source of spatial variability in expression Expression variability has many contributing factors, including subject-to-subject variability, regional variability and experimental noise. The above results suggest that the variability between brain regions is significant for most genes. But, how large is regional variability compared to other sources of expression variability? To answer this question, we used principal component analysis (PCA) to extract the main axes of variability in the data (see Methods). Interestingly, the PCA of the human expression was also analyzed previously by Tan et al. [31]. Tan et al. used PCA to embed genes in a low dimension space that preserved much of the gene-to-gene variability. In that space, they found that neurons and oligodendrocytes are on the opposite end of the first principal component. Here we address the complementary analysis, looking for the dimensions that preserve the sample-to-sample variability. Fig 5A shows the projection of brain samples onto the 1st and 2nd principal components, which together account for 34% of the variance (S3 Fig). Samples are colored by the brain region from which they were taken. Brain regions are well separated in this projection, in a way that matches the anterior-posterior axis and the BRO. The isolated cluster of samples on the left belongs to the cerebellum, which is well known to exhibit a unique molecular and cellular organization [21,51,52]. This analysis shows that the BRO is a major determinant of variability in human brain transcriptome. 10.1371/journal.pcbi.1005064.g005Fig 5 Projection of the samples from the human6 dataset on the 1st and 2nd principal componenets with two coloring schemes. (A) The samples are colored according to the position of the corresponding embryonic region, using the same color scheme as in Fig 1A. (B) The samples are colored according to one of the six donors. As a comparison, Fig 5B shows the same projection on the two top PCs, but this time the samples are colored by the subject from which each sample was taken. Expression differences between people are pronounced mostly in frontal regions (top right samples), but are dramatically weaker than the differences between brain regions. Subject-to-subject differences are more pronounced when projecting on the 3rd and 4th principal components (S2 Fig). To quantify the relative contribution of subject identity and region of origins to expression variability, we computed the fraction of variance explained by these two features. For every gene, we examined it expression across samples separately and computed the fraction of explained-variance (See Methods) (S8A Fig). The subject-identity explains 0.13 (+/- 0.12) of a gene’s expression-variance, while the region explains 0.28 (+/- 0.21) of a gene’s expression-variance. Together, both sources explain nearly half of the sample variability (median at 0.43 +/- 0.18). Region and subject identity explain “different” component of the variance: the fraction of variance explained by region is inversely correlated with the variance explained by subject id. (Spearman ρ = -0.51 S8B Fig). Using the same dataset, Hawrylycz et al. recently took the complementary track and searched for stable expression-patterns across subjects [53]. They showed that genes with conserved patterning across subjects display strong relationships to anatomical structure, functional connectivity and other features of the human brain. BRO agreement and spatial variability in paralogs and functionally-related genes With many genes exhibiting spatial expression that matches the developmental origin of brain regions, the question remains if and how expression variability is used by the brain to tune the functional properties of cells and circuits. One particularly interesting aspect of such tuning is how the brain controls the expression of similar genes, including paralog and other functionally-related gene pairs. In many cases, the brain is known to switch from expressing one paralog variant to another variant. Such switches have been studied mostly in the context of development and synaptic pathways, including the widely studied switch in NMDA receptors from subunit NR2A to NR2B [54–57]. These developmental switches can be traced to occur within a brain region, and in some cases well after birth [54,56]. Here we study spatial switching in pairs of genes, where genes coding for different protein variants are expressed in different brain regions. We set to study the relation between spatial expression switching and developmental origin of regions, using the per-gene BRO-agreement score. To study fine spatial tuning, we aimed to focus on pairs of genes that share similar functions. To collect such gene pairs we used two approaches. First, we used a set of paralog genes defined by Ensembl (denoted ensemble-based paralogs). Second, to further focus on genes with putative similar function, we collected pairs of genes that share the same functional role in cellular pathways, as captured by KEGG. We also required that these gene pairs have a significant sequence similarity and denote this set Kegg-based pairs (see Methods). Both sets were restricted to brain-related synaptic pathways. We first compared the distribution of spatial correlation strengths of similar gene pairs, quantified by log(p-values) on ABA6-2013 data. We found that both the KEGG-based gene pairs and the Ensembl-based paralogs are significantly more spatially anticorrelated than random gene pairs (Fig 6A). Furthermore, KEGG-based gene pairs are more anticorrelated than Ensembl paralogs in these pathways (Fig 6A). The spatial correlations of KEGG-based pairs are fairly consistent when compared to the correlation using the same gene pairs in adult brains from the Kang-2011 dataset, considering they were measured by different labs and in different brain regions (Fig 6B). 10.1371/journal.pcbi.1005064.g006Fig 6 (A) Distribution of gene pairs with anti-correlated spatial expression. Kegg-based gene pairs include 1496 pairs with (1) sequence similarity > 30%, and (2) sharing a sub-component in one of 17 KEGG synaptic pathways (see Methods). Ensembl-based paralogs include 3503 pairs of paralogs (as defined by Ensembl) where both genes in a pair are included in one of same 17 pathways (see Methods). Baseline corresponds to the distribution expected at random. (B) Consistency of spatial correlations across two datasets ABA2013 and Kang-2011. The spatial correlations of paralog pairs across the two dataset, show a significant agreement (Spearman ρ = 0.48, p-value < 10−78). Each point correspond to the median correlation across adult subjects, in one gene pair (total of 1496 pairs). (C-F) Examples of development of spatial correlations in the Serotonin system. (C) The pair of genes coding for Serotonin receptors HTR2A and HTR1F exhibit a continuous rise in spatial correlation, riding from slightly negative in early embryonic development to strong positive correlation. (D) The pair of genes coding for Serotonin receptors HTR5A and HTR2C show a sharp transition from positive to negative spatial correlation in early development, which is then preserved through life. (E) The paralogs CACNA1A and CACNA1D exhibit a rise in spatial correaltion (F) The paralogs HTR7 and HTR5A show a continuous change in spatial correation, from positive corealation during embryonic development to negative one at adulthood. Next, we compared the strength of spatial correlation of each pair of genes with their combined BRO-agreement scores, and found a strongly significant correlation between the two (KEGG-based set, Spearman with log(p-value), ρ = 0.36 p-value < 10-42, n = 1496). More surprisingly, when considering anti-correlated gene pairs, pairs with high BRO-scores tend to be more strongly anti-correlated (Spearman rho = -0.27 p-value<10−9 n = 1496). This effect is not simply due to some pairs having more variable expression across the brain, since the dependency on BRO is significantly stronger (p-value <10−42) than the dependency on spatial variability (p-value < 0.01, quantified using the standard deviation across samples). One interpretation of these findings is that the brain tunes the expression of pairs of functionally-related pairs of genes, such that they are expressed differently in brain regions, and that this tuning is in strong agreement with the developmental origin. Together with the BRO results, these findings suggest that correlated spatial expression may be formed early in development. To test this hypothesis we computed the spatial correlations for each subject in the Kang-2011 data, which allows tracing how spatial correlations develop with age. We then searched for KEGG-based pairs whose spatial correlations follow a trends, and found that 17% of the pairs exhibit a significant trend (257/1496, FDR corrected p-value<0.01, F-test from fitting a linear regression model, as compared with the constant model). Far fewer pairs of Ensembl paralogs exhibit a significant trend (6.8% of the pairs 240/3503). Fig 6C shows the top pair in the KEGG-based pairs set (p-value of trend < 10−4). It corresponds to a pair of Serotonin receptor genes, HTR2A and HTR1F from two different receptors (5-HT1, 5-HT2). Interestingly, their spatial correlation is negative prenatally and is around zero around birth. However, it continuously grows throughout life, reaching high positive correlation at adulthood. This pattern is interesting for several reasons. First, the gradual increase in correlations throughout life is not likely to be caused by changes in cell proportions, since there is a significant change in correlation between childhood and adulthood. Second, other genes in the Serotonin system exhibit different patterns. Fig 6D shows a pair of genes coding for Serotonin receptors HTR5A and HTR2C. Here the early embryonic positive correlation is replaced by a negative correlation around birth, which remains quite stable during life. Fig 6E and 6F show similar patterns in two pairs of Ensembl paralogs, CACNA1A vs CACNA1D two genes coding for Calcium channels, and HTR7 vs. HTR5A, two serotonin receptors. HTR7 is known to be involved in both early and post-natal development [58]. One possible interpretation of the prevalence of high BRO-agreement scores is that the expression patterns of many are determined early in development, and are preserved through life and in the adult brain. Alternatively, it is also possible that gene-expression changes in a dynamic way through life, but keep following patterns that agree with the embryonic origin of regions. To test these two hypotheses, we quantified the relation between the strength of expression changes of pairs through life, and the BRO scores of the gene pair. We find that the two are positively correlated (ρ = 0.14, p-value<10–7, S9A Fig, for KEGG-based pairs, and, ρ = 0.17, p-value<10–21, S9B Fig for Ensembl-based paralogs), namely, pairs of genes with higher BRO scores actually tend to exhibit more changes in their spatial correlations, consistent with the second hypothesis. These results are consistent with the view that spatial expression patterns in the adult are not a mere reflection of the brain structure as determined in early development, but are tuned to use genes coding for different protein variants in a differential way across the brain. Discussion To characterize the areal patterns of gene expression in the human brain, we analyzed two datasets of mRNA expression from post-mortem adult donors. For each gene, we computed an index that measures how its expression pattern agrees with a hierarchical ontology of brain-regions, based on their developmental origin. We find that 92% of human genes exhibit an expression pattern that significantly agrees with the known brain-region ontology. The fact that such a large fraction of the human genome is differentially expressed across brain regions suggests that control of expression in the brain is largely region-specific. When focusing on genes that are expressed specifically in neurons, glia and oligodendrocytes, we find that cell-type specific genes tend to strongly agree with the tree-structured ontology. This suggests that not only do these markers differ between regions, as suggested by Ko et al. [24], but that they also follow a specific pattern of expression which corresponds to the embryonic origin of the region and to a larger extent than the average gene. Interestingly, significant BRO scores are not limited to neurons, which are often known to differ across brain regions, but are also observed in glia-specific genes, which are often viewed as performing brain-wide and generic functions. Having adult expression patterns that strongly agree with the developmental brain region ontology could have various interpretations. First, adult spatial expression patterns could be determined by the embryonic origin of each region, for example because brain regions differ by their cell-type profiles, or due to the expression of region-specific markers. Alternatively, adult expression may reflect delicate tuning of expression where different brain regions utilize different protein variants, optimized for the function of each brain region. We find evidence that support the second alternative. First, gene with high BRO-scores tend to change their expression more during development. Second, pairs of functionally-related genes (participating in a similar role in synaptic pathways) have stronger spatial anti-correlation than paralogs in those pathways. Finally, in those pairs of functionally related genes, pairs with higher BRO scores tend to have stronger spatial anti-correlation. The approach we presented has various limitations. Transcriptome data measured from brain tissues involves a mixture of various cell types whose proportions and conditions are not known. Developing demixing approach to infer proportions from the mixture [25] is an important challenge, can be based on single-cell transcriptomics (as in Darmanis et al 2015), and is likely to significantly change our understanding of brain transcriptome. Genes involved in patterning and axon guidance clearly exhibit regional patterns during early development [59]. The above results show that their expression continues to be governed by the region ontology in the adult brain, long after their developmental role has been completed. As one specific example, consider a gene from the top BRO-scorers in the ABA6-2013 dataset, FEZF2 (forebrain embryonic zinc finger protein 2), a transcription repressor involved in specification of subcerebral projection neurons [60,61]. FEZF2 is believed to play a role in layer and neuronal patterning of subcortical projections and axonal fasciculation and was shown to regulate axon targeting of layer 5 subcortical projection neurons, where axons of FEZF2 deficient mice failed to reach their targets [62]. In the adult human brain, our results show that FEZF2 retains strong areal differences in adulthood, and is strongly expressed in the cortex, less so in the midbrain and the least in the hindbrain (S6 Fig). Indeed, the mouse variant of FEZF2 is known to be expressed in adult projection neurons [62]. Importantly, these results suggest that FEZF2 assumes another functional role in the adult cortex. Indeed, it has been shown that projecting neurons in the mouse motor cortex expressing Fezf2 have distinct physiological characteristics [63]. The abundance of genes that retain their areal differential expression in adulthood suggests that this may be the general case, and many genes that play a role in brain development later assume new roles that affect the function of the adult brain. The fraction of genes having distinct areal expression pattern has been previously estimated using a different method (ANOVA). In the Kang-2011 data, it was found to be on the order of hundreds of genes in the adult human brain (Pletikos et al. [19], Fig 2B). In the ABA6-2013 data 84% of the genes were found to have differential expression across brain regions [5]. ANOVA estimates are sensitive to differences in the mean expression of regions, regardless of the region ontology, and could capture genes whose expression pattern in some brain regions is different from others. As such, they are more sensitive to genes that are uniquely expressed in one or few region. The BRO-score can therefore be viewed as a complementing measure, which is sensitive to global areal-differential expression that is consistent with the brain region ontology. The fact that the expression of most genes in the adult brain is governed by earlier development stages suggest that many studies which deal with regional differential expression should be carefully interpreted. For example, combining samples taken from ontology-distant regions would lead to large expression variance, reflecting the developmental origin of the structures tested. Furthermore, areal differential expression should be measured compared to a baseline expression profile that takes in to account the region ontology. The results in this paper suggest that spatial expression patterns in the adult human brain are controlled in a way that follows the embryonic origin of regions, but at the same time that spatial patterns of related genes may change during development in a correlated way. It remains to be discovered which transcription control mechanisms maintain these distinct areal expression patterns. Methods Gene expression measurements We analyzed gene expression data from two sources. First, a set of 3702 microarrays provided by the Allen Human Brain Dataset (ABA6-2013) from human.brain-map.org [5]. We mapped 58692 microarray probes to genes based on mapping provided by the Allen Institute. When multiple probes were available for a gene, we selected the probe that was most consistent across the 6 human subjects; suggested by human.brain-map.org. Specifically, when we analyzed genes with multiple probes, we first computed the expression correlation across regions of each probe, then averaged the correlation scores across all pairs of subjects and chose the probe that was most correlative. Overall, we analyzed 20773 transcripts. The number of samples per donor ranged from 363 to 946, for a total of 3702 tissue samples. The second dataset was a set of 1340 microarray samples collected by Kang and colleagues from 57 postmortem brains containing expression values for 17565 genes [6]. We refer to this dataset as Kang-2011. We limited the analysis to donors that are older than 12, yielding a total of 20 donors and 491 tissue samples. The Kang dataset was also used for the analysis of BRO-agreement over development (Fig 2E). For this analysis we also used the pre-natal subjects and early-childhood subjects. The subject ages range from 13 post conception weeks to 82 years. We grouped the subjects into 12 age-groups following the original Kang paper, and computed BRO scores per age group. Brain region ontology We used the brain region ontology hierarchy provided by the Allen institute human.brain-map.org containing 1534 regions organized in a hierarchical manner. From the full set of regions we used two ontologies: A fine region-ontology with 414 regions which had measurements that were associated with them, and a coarse region-ontology with 16 brain regions. The list of 16 gross regions is given in supplemental S2 Table. The coarse part of the ontology (upper part in Fig 1A) was largely based on brain development while the fine parcellation of regions was based more on cytoarchitecture. The results we report are based on the coarse 16 region-ontology. We report below results for both coarse and fine grained ontologies. Measurements from the Kang-2011 dataset were obtained from 16 regions. We mapped those regions to 16 regions of the Allen ontology, and the mapping is given in supplemental S3 Table. BRO-agreement scores The BRO-agreement score was computed separately for each gene as follows. For each pair of samples (a,b), we define their tree similarity as the distance (number of edges) between the regions in the ontology hierarchy tree dtree(a,b). We define their expression similarity as the absolute difference between the expression values of the two samples for the current gene (i)—dexpression(ai,bi). We computed the two distances over all pairs of tissue samples, and computed the Spearman correlation between the two as the BRO score. BRO_score(gene i)≔spearman corr(a,b)∈all sample pairs( dtree(a,b),  dexpression(ai,bi) ) To generate random scores, we calculate Bro-agreement scores of permutated vectors. Genes with a BRO-score above the top 1% of the permuted scores were considered significant. We also tested a second ontology-agreement score based on triplet ranking. We randomly selected 106 sample triplets (a, b, c) and computed the fraction of times that a triplet is ranked with the same ordering in both the tree and the expression distance measures: BRO_triplet(gene i)≔#(dtree(a,b)<dtree(a,c)) &(dexpression(a,b)<dexpression(a,c)))#(dtree(a,b)<dtree(a,c)) This score gave similar results which are not shown here. To handle biases that could arise from different scales in the samples we also checked a normalized version of the Bro-agreement. In this normalized version we first normalized each sample to zero mean and unit variance and then computed the BRO-agreement. The results were robust to this change and we choose to present the un-normalized BRO-agreement scores. To handle biases that could arise from the number of regions in the ontology, we used two different granularities of the ontology tree. The first uses 16 gross regions and the second uses the entire tree (414 regions). Combining scores from multiple subjects We tested two ways to combine expression measures from multiple subjects into a single BRO score. First, we simply aggregated all samples of all subjects from a given region, and computed the BRO agreement score. Second, in the ABA6-2013 dataset, the number of samples per subject is large enough, such that a BRO score can be computed separately for each subject. We then computed a global BRO score of a gene as the average over the 6 individual-subject BRO scores. 94% of the genes were significant compared to random, according to the first score, and 92% according to the second method. For consistency with the Kang-2011 data, the figures use the first method. We report the more conservative estimate of significant genes as 92%. We computed BRO-scores which uses the fine region-ontology. The upper branches in this ontology are more developmental oriented and the lower branches are more cytoarchitecture-driven. Using this ontology 89% of the genes are BRO significant. The BRO scores of the fine region-ontology and of the coarse 16 region-ontology are very similar with a spearman correlation of 0.98. To test if these results are sensitive to the number of region available in ontology tree, we repeated the analysis of the ABA6-2013 dataset at a coarser resolution of 16 regions. Using this coarse resolution, 95% of the genes were BRO significant. To test if the number of BRO significant genes is sensitive to the number of available samples, we randomly subsampled subsets of size 500 samples, which decreased the fraction of BRO significant genes to 51% ± 2% (supplementary S1 Table). Robustness across subjects The percent of BRO-significant was tested for robustness using the ABA6-2013 dataset, where hundreds of samples are available for each subject. For each gene, we calculated its BRO-score but this time for each subject separately. The fraction of genes with a significant BRO score (p-value < 0.01) is stable across the individual subjects, yielding 89%, 90%, 76%, 91%, 83% and 86% (supplementary S7 Fig). The fraction of significant genes at the group level is slightly higher; suggesting that the groups score manages to remove some of the inter-subject noise. ANOVA analysis Expression variability of a single gene across regions is sometimes captured by comparing the mean expression level in each region using ANOVA [5,6]. This approach would find a gene as significant even if it is differentially expressed in a single region, or if it is expressed in a set of regions regardless of their position in the ontology tree. Hence in principle, the BRO agreement index poses a stronger requirement of agreement with the ontology. When computing ANOVA across 414 regions of the ABA6-2013 dataset, and using the average scores for each of the 6 subjects, 86% of the genes (17895 out of 20773) were significantly differentially expressed across regions (FDR-corrected q-value < 0.01), 83% of these genes (17266 out of 20773) were also BRO-significant. 8% of the genes (1719 out of 20773) had BRO-significant scores but not ANOVA-significant, most likely because the BRO index combined weak affects across multiple nodes of the tree. For each gene, we computed one-way ANOVA on samples expression levels. The p-values reported are under the null hypothesis that samples are drawn from regions which have the same mean expression. We computed ANOVA separately for each human subject and report the average ANOVA score across subjects for each gene. Similar to the BRO scores we performed ANOVA on the fine grained ontology and on the coarse ontology. We then corrected the p-values for multiple comparisons using FDR. Cell type-specific markers For the genome wide analysis of BRO scores We used the human orthologs of the set of genes characterized by Cahoy et al. [36], who used microarrays to profile expression patterns in purified populations of neurons, astrocytes and oligodendrocytes. For testing how spatial variability could be explained by cell-type specific markers, we used a set of known markers collected from various sources including [36]. For neuronal markers we used EMX1, MAP2, GRIA2, DLG4, DLG3, NRGN, STMN2, SYT1, CELF4, CELF5, CELF6. For glia markers we used GFAP, MBP, SLC1A2, SLC1A3, DLG4, DLG3, MAP2. To compute the dependence of spatial variability on those markers we fitted a quadratic function for each of the genes separately using a least square loss, and computed the explained variance R2, compared to a constant model. Sets of gene pairs We studied two sets of gene pairs: Ensembl paralogs and KEGG-based gene pairs. For the first set, we used the paralogs available from Ensembl (ensembl.org, May 2016), and limited to gene pairs that had an Entrez id and were included in synaptic and brain related pathways as described by KEGG. Specifically, these included 17 pathways with KEGG accession numbers 04020, 04724, 04725, 04726, 04727, 04728, 04730, 05010, 05012, 05014, 05016, 05030, 05031, 05032, 05033, 05034, 04080. Second, for KEGG-based gene pairs, we created a set of gene pairs designed to capture functionally-related genes, by collecting gene pairs that reside within the same functional element in KEGG pathway repository. These KEGG elements group together proteins with common functionally and interaction partners. We found these KEGG elements to be usually more functionally-coherent than protein families, and at the same time less specific than protein sub-families. We further required that pairs have sequence similarity above 30% (See also [64]). Trends in spatial correlation To find pairs whose spatial correlation has a trend, we fitted a linear regression model with least square loss with age as the predicting variable and spatial correlation as the predicted variable. Significance of the trend was measured based on the F statistic of the explained variance and was FDR corrected for multiple hypotheses. Hox, Pax and axon guidance genes We used the set of genes that belong to the human axon guidance pathway. The set was manually curated by KEGG (www.genome.jp/kegg). The Hox genes used are: HOXA1, HOXA10, HOXA11, HOXA2, HOXA3, HOXA4, HOXA5, HOXA6, HOXA7, HOXA9, HOXB1, HOXB13, HOXB3, HOXB4, HOXB5, HOXB6, HOXB9, HOXC10, HOXC12, HOXC13, HOXC4, HOXC5, HOXC8, HOXC9, HOXD1, HOXD12, HOXD13, HOXD3, HOXD4, HOXD8, HOXD8, and HOXD9. The Pax genes used are: PAX1, PAX2, PAX3, PAX4, PAX5, PAX6, PAX7 and PAX8. Highly stable gene A set of 11 genes collected by Eisenberg et al. [65] and available at (http://www.tau.ac.il/~elieis/HKG/). Hierarchical clustering analysis We used agglomerative hierarchical with average linkage and Euclidean distance over 3702 samples from ABA6-2013 obtained from six subjects. Samples from the same brain regions were first averaged to create a single profile for each region. Principal component analysis We used all 3702 samples from ABA6-2013 to compute the covariance matrix of gene expression levels, and then computed the top principal component of the expression covariance matrix. Explained sample-variance analysis For each gene, we evaluated how much of its expression variance (over samples) can be explained using two sources of information: The region the sample was taken from and the identity of the subject the sample was extracted from. The explained variance was computed for each gene by fitting linear model using each of these sources of information, and using both of them together. Gene enrichment analysis For robustness, we combined two BRO-score of each gene by multiplying the BRO-score computed with the Kang-2011 data with that computed with the ABA-2013 data. We performed the ranked based enrichment analysis using Gorilla (http://cbl-gorilla.cs.technion.ac.il/). Gene age-index We used the gene age-index published by Domazet-Lošo and Tautz [66]. Supporting Information S1 Table Sensitivity of the number of BRO significant genes in the ABA6-2013. Sensitivity to the number of regions in the ontology tree depth was assessed by using two ontologies with different resolutions: A fine-resolution ontology that contained 414 regions and a coarse-resolution ontology of 16 gross regions. Sensitivity to the number of available samples was assessed by computing the BRO score using random subsets of 500 samples. (XLS) Click here for additional data file. S2 Table Names of brain regions and their corresponding symbols ABA ontology [31]. (XLS) Click here for additional data file. S3 Table Region notations. Mapping the notation of regions in Kang-2011 and in the Allen Institute region ontology ABA6-2013 [5,6]. (XLS) Click here for additional data file. S4 Table Top 10 ranking of GO biological processes enriched in genes with high BRO scores. Enrichment was computed using mGH [35] testing for enriched biological processes using the full ranked list of BRO-score genes. All top processes are brain-related, with high enrichment for cell signaling and neural development. (XLS) Click here for additional data file. S1 Fig Correlation between tissues. Pearson correlation between expression vectors of all pairs of tissue samples. Samples are ordered first by gross region than by donor. Samples from regions that are close in the developmental region ontology are highly correlated in their expression profile. Within each region, samples are also correlated with other samples from the same donor. (TIF) Click here for additional data file. S2 Fig Projection of samples from ABA6-2013 dataset onto the 3rd and 4th principal componenets as in Fig 5, with two coloring schemes. (A) Each point is a tissue sample. Samples are colored based on the position of the corresponding embryonic region. (B) Colors correpsond to donor identity. A significant fraction of the sample variance across the 3rd and 4th principal components is explained by subject-to-subject variablity. (TIF) Click here for additional data file. S3 Fig Fraction of variance explained by the top principal components, as in S2 Fig. The first two principal components caputre 34% of the variance. Adding the 3rd and 4th principal components explain more than 51% of the sample-to-sample variance. (TIF) Click here for additional data file. S4 Fig Distribution of cortex-BRO-agreement scores in ABA6-2013 and Kang-2011. Color scheme and x-axis scale matche those of Fig 2. (A) A scatter plot showing BRO-agreement scores for the two datasets. Each light-grey dots corresponds to a single genes (a total of 17K genes). Dark-grey dots correspond to permuted data (see Methods). (B) Marginal distribution of BRO scores in the ABA6-2013 dataset. In the ABA6-2013 dataset, 11% of the genes (2207 out of 20773) are BRO-significant in the cortex. BRO score was also computed separately for each subject, using the ABA6-2013 dataset. With these per-subject scores, the number of BRO-significant genes varied considerably across the six subjects (30%, 33%, 5%, 37%, 16% and 18%), and the correlation of BRO scores between subjects is on average lower (mean Spearman Correlation of cortex BRO scores of a pair of subjects is 0.23 ± 0.13, compared with 0.76 ± 0.08 for the whole brain). (C) Marginal distribution of BRO scores in the Kang-2011 dataset. The large fraction of BRO-significant genes observed in ABA6-2013 was not found in the Kang-2011 dataset, where the two distributions largely overlap. (TIF) Click here for additional data file. S5 Fig The distribution of cortical BRO-agreement scores on various subsets of genes. Color scheme and x-axis scale match those of Fig 3. (A, B) Neurons and astrocytes receive significantly higher agreement scores than all genes (Wilcoxon one tail test; neurons: p-value = 10-9, astrocytes: p-value = 10−22). Oligoodendrocyte genes are in less agreement with region-ontology than the full set of genes. Comparing these BRO scores to a randomized scores we find that 45% of the neuronal markers are cortex-BRO significant, 69% of the astrocytes markers are cortex-BRO significant and that 7% of the oligodendrocytes markers are cortex-BRO significant;). (C, D) Axon-guidance genes receive higher scores than genes on average (Wilcoxon one tail test, p-value = 10−3). Hox genes are less in agreement with region-ontology than the full set of genes (still significantly larger than random). PAX2 and PAX6 obtain high BRO scores. (TIF) Click here for additional data file. S6 Fig FEZF2 (ZNF312) expression pattern corresponds with the brain region ontology. FEZF2 shows a clear transtion of its expression levels. The samples from the cortex show high experssion values where the samples midbrain has less and the samples of the hindbrain has the least expression. (A) The mean expression levels of the FEZF2 within different region in the human brain (the color scheme as consitant with that of Fig 1A. (B) We embed the samples in a 3D space using its MRI standartize location where the color of each sample shows the expression level of FEZF2. The scatter shows a transiation from high expression levels in the cortex to lower expression levels in the inner brain structures and to the hindbrain. (TIF) Click here for additional data file. S7 Fig The distribution of BRO-scores computed separately for each subject in ABA6-2013. The percent of BRO-significant genes (p-value < 0.01) is stable when computed for each subject separately: 89%, 90%, 76%, 91%, 83% and 86%). (TIF) Click here for additional data file. S8 Fig Region information explains more of the sample variance than subject identity. The joint information of subject id and region id explains almost half of the sample variance. For each gene, we represented the identity information as 1-hot-vectors, and computed the explained sample-variance by fitting a linear model (A) Distributions of explained variance across genes, as explained by region, subject or both. (B) The joint distribution of both the explained variance from region and from subject identity. (TIF) Click here for additional data file. S9 Fig BRO score is positively correlated with pairs of genes which show a significant trend in development. Trend significance of change in spatial correlation through life was quantified using the standard F-test comparing residual of a linear regression model with a constant model. Pair BRO score is the minimum BRO scores of the two genes in the pair. (A) We computed the developmental trend for each pair of genes from the KEGG-based set. We found that BRO-agreement scores are (weakly but significantly) positively correlated with having a significant trend. Each dot corresponds to one brain related sequence-similar pair. 17% of pairs (257/1496) had FDR-corrected significant (q<0.01) linear trend of their spatial correlation, as illustrated in Fig 6. (B) Similarly, we found that the developmental trend of paralog pairs which are brain related is also positively correlated with the BRO score. 6.8% of the brain related paralog-pairs show a significant trend (240/3503). (TIF) Click here for additional data file. S10 Fig (A) 2D scatter of BRO scores for genes with a matching homolog in mouse (Zapala et al.[22]) and human (ABA2013) and a random baseline (B) 2D heat map of the joint distribution of BRO-scores for mouse and human. (C) The distribution of BRO-scores in mouse and a random baseline, showing that the fraction of BRO significant genes is smaller in the mouse dataset (30%) than in humans. (TIF) Click here for additional data file. S11 Fig (A) The Allen ontology used for the human analysis. (B) The Allen ontology used for the mouse analysis. Both follow ontologies first devide the brain into the 5 embrionic vesiceles and then go into more detailed regionalization. The leaf regions are not the same since the data was gathered in different experiments each with a unique focus. (TIF) Click here for additional data file. S1 Text Supplemental analysis text. BRO in the human cortex and a comparsion between the BRO in human and in mouse. (DOCX) Click here for additional data file. S1 Data A list of per gene BRO score. The table containt: Gene symbo, Gene Entrez-ID, BRO score that is the product of the BRO of ABA6-2013 and Kang-2011, BRO score computed from the ABA6-2013 dataset, BRO score computed from the Kang-2011 dataset. A mark per gene which indicates it involvment in one of these 11 classes: cell-cell signaling, synaptic transmission, neuron differentiation, neuron projection development, generation of neurons, axon development, neuron development, schizophrenia, autistic disorder, seizures,epilepsy, substance related disorders. (CSV) Click here for additional data file. The authors are grateful to the Allen Institute for Brain Science for making their data available to the scientific community and for helping us with any questions. They are also grateful to Noa Liscovitch for valuable discussions and insights. ==== Refs References 1 Amunts K , Schleicher A , Bürgel U , Mohlberg H , Uylings HB , Zilles K . Broca’s region revisited: cytoarchitecture and intersubject variability . J Comp Neurol . 1999 ;412 : 319 –41 . Available: http://www.ncbi.nlm.nih.gov/pubmed/10441759 10441759 2 Brodmann K . Vergleichende lokalisationslehre der grobhirnrinde . Barth , Leipzig 1909 ; 3 Amunts K , Kedo O , Kindler M , Pieperhoff P , Mohlberg H , Shah NJ , et al Cytoarchitectonic mapping of the human amygdala, hippocampal region and entorhinal cortex: intersubject variability and probability maps . Anat Embryol (Berl) . 2005 ;210 : 343 –52 . 10.1007/s00429-005-0025-5 16208455 4 Braitenberg V , Schüz A . Cortex: Statistics and Geometry of Neuronal Connectivity [Internet] . Berlin, Heidelberg : Springer Berlin Heidelberg ; 1998 10.1007/978-3-662-03733-1 5 Hawrylycz MJ , Lein ES , Guillozet-Bongaarts AL , Shen EH , Ng L , Miller JA , et al An anatomically comprehensive atlas of the adult human brain transcriptome . Nature. Nature Publishing Group, a division of Macmillan Publishers Limited. All Rights Reserved .; 2012 ;489 : 391 –9 . 10.1038/nature11405 6 Kang HJ , Kawasawa YI , Cheng F , Zhu Y , Xu X , Li M , et al Spatio-temporal transcriptome of the human brain . Nature. Nature Publishing Group, a division of Macmillan Publishers Limited. All Rights Reserved .; 2011 ;478 : 483 –9 . 10.1038/nature10523 7 Stiles J , Jernigan TL . The basics of brain development . Neuropsychol Rev . 2010 ;20 : 327 –48 . 10.1007/s11065-010-9148-4 21042938 8 Johnson MB , Kawasawa YI , Mason CE , Krsnik Z , Coppola G , Bogdanović D , et al Functional and evolutionary insights into human brain development through global transcriptome analysis . Neuron . 2009 ;62 : 494 –509 . 10.1016/j.neuron.2009.03.027 19477152 9 Martínez S . The isthmic organizer and brain regionalization . Int J Dev Biol . 2001 ;45 : 367 –71 . Available: http://www.ncbi.nlm.nih.gov/pubmed/11291867 11291867 10 Krauss S , Johansen T , Korzh V , Fjose A . Expression pattern of zebrafish pax genes suggests a role in early brain regionalization . Nature . 1991 ;353 : 267 –70 . 10.1038/353267a0 1680220 11 Dickson BJ . Molecular mechanisms of axon guidance . Science (80-). American Association for the Advancement of Science ; 2002 ;298 : 1959 –64 . 10.1126/science.1072165 12 Barallobre MJ , Pascual M , Del Río JA , Soriano E . The Netrin family of guidance factors: emphasis on Netrin-1 signalling . Brain Res Brain Res Rev . 2005 ;49 : 22 –47 . 10.1016/j.brainresrev.2004.11.003 15960985 13 Sansom SN , Livesey FJ . Gradients in the brain: the control of the development of form and function in the cerebral cortex . Cold Spring Harb Perspect Biol . 2009 ;1 : a002519 10.1101/cshperspect.a002519 20066088 14 O’Leary DDM , Chou S-J , Sahara S . Area patterning of the mammalian cortex . Neuron . 2007 ;56 : 252 –69 . 10.1016/j.neuron.2007.10.010 17964244 15 Alexander T , Nolte C , Krumlauf R . Hox genes and segmentation of the hindbrain and axial skeleton . Annu Rev Cell Dev Biol. Annual Reviews ; 2009 ;25 : 431 –56 . 10.1146/annurev.cellbio.042308.113423 16 Takahashi Y , Hamada J , Murakawa K , Takada M , Tada M , Nogami I , et al Expression profiles of 39 HOX genes in normal human adult organs and anaplastic thyroid cancer cell lines by quantitative real-time RT-PCR system . Exp Cell Res . 2004 ;293 : 144 –153 . Available: http://www.ncbi.nlm.nih.gov/pubmed/14729064 14729064 17 Neville SE , Baigent SM , Bicknell AB , Lowry PJ , Gladwell RT . Hox gene expression in adult tissues with particular reference to the adrenal gland . Endocr Res . 2002 ;28 : 669 –673 . 12530681 18 Morgan R . Hox genes: a continuation of embryonic patterning? Trends Genet . 2006 ;22 : 67 –9 . 10.1016/j.tig.2005.11.004 16325300 19 Pletikos M , Sousa AMM , Sedmak G , Meyer KA , Zhu Y , Cheng F , et al Temporal specification and bilaterality of human neocortical topographic gene expression . Neuron . 2014 ;81 : 321 –32 . 10.1016/j.neuron.2013.11.018 24373884 20 Colantuoni C , Lipska BK , Ye T , Hyde TM , Tao R , Leek JT , et al Temporal dynamics and genetic control of transcription in the human prefrontal cortex . Nature. Nature Publishing Group ; 2011 ;478 : 519 –523 . 21 Liscovitch N , Chechik G . Specialization of gene expression during mouse brain development . Sporns O , editor. PLoS Comput Biol. Public Library of Science ; 2013 ;9 : e1003185 10.1371/journal.pcbi.1003185 22 Zapala MA , Hovatta I , Ellison JA , Wodicka L , Del Rio JA , Tennant R , et al Adult mouse brain gene expression patterns bear an embryologic imprint . Proc Natl Acad Sci U S A . 2005 ;102 : 10357 –62 . 10.1073/pnas.0503357102 16002470 23 French L , Pavlidis P . Relationships between gene expression and brain wiring in the adult rodent brain . PLoS Comput Biol . 2011 ;7 : e1001049 10.1371/journal.pcbi.1001049 21253556 24 Ko Y , Ament S a , Eddy J a , Caballero J , Earls JC , Hood L , et al Cell type-specific genes show striking and distinct patterns of spatial expression in the mouse brain . Proc Natl Acad Sci U S A . 2013 ;110 : 3095 –100 . 10.1073/pnas.1222897110 23386717 25 Grange P , Bohland JW , Okaty BW , Sugino K , Bokil H , Nelson SB , et al Cell-type-based model explaining coexpression patterns of genes in the brain . Proc Natl Acad Sci U S A . 2014 ;111 : 5397 –402 . 10.1073/pnas.1312098111 24706869 26 Schüz A , Palm G . Density of neurons and synapses in the cerebral cortex of the mouse . J Comp Neurol . 1989 ;286 : 442 –55 . 10.1002/cne.902860404 2778101 27 Roe AW , Pallas SL , Kwon YH , Sur M . Visual projections routed to the auditory pathway in ferrets: receptive fields of visual neurons in primary auditory cortex . J Neurosci . 1992 ;12 : 3651 –64 . Available: http://www.ncbi.nlm.nih.gov/pubmed/1527604 1527604 28 Sur M , Garraghty PE , Roe AW . Experimentally induced visual projections into auditory thalamus and cortex . Science . 1988 ;242 : 1437 –41 . Available: http://www.ncbi.nlm.nih.gov/pubmed/2462279 2462279 29 Rauschecker JP . Compensatory plasticity and sensory substitution in the cerebral cortex . Trends Neurosci . 1995 ;18 : 36 –43 . 10.1016/0166-2236(95)93948-W 7535489 30 Amedi A , Raz N , Pianka P , Malach R , Zohary E . Early “visual” cortex activation correlates with superior verbal memory performance in the blind . Nat Neurosci . 2003 ;6 : 758 –66 . 10.1038/nn1072 12808458 31 The Allen Institute. Ontology and nomenclature in the Allen human brain atlas [Internet]. Available: http://help.brain-map.org/download/attachments/2818165/HBA_Ontology-and-Nomenclature.pdf?version=1&modificationDate=1382051847989 32 Bohland JW , Bokil H , Pathak SD , Lee C-K , Ng L , Lau C , et al Clustering of spatial gene expression patterns in the mouse brain and comparison with classical neuroanatomy . Methods . 2010 ;50 : 105 –12 . 10.1016/j.ymeth.2009.09.001 19733241 33 Ng L , Pathak SD , Kuan C , Lau C , Dong H , Sodt A , et al Neuroinformatics for genome-wide 3D gene expression mapping in the mouse brain . IEEEACM Trans Comput Biol Bioinforma IEEE ACM. Published by the IEEE CS, CI, and EMB Societies & the ACM ; 2007 ;4 : 382 –393 . 10.1109/tcbb.2007.1035 34 Consortium GO . The Gene Ontology (GO) project in 2006 . Database . 2006 ;34 : 322 –326 . 10.1093/nar/gkj021 35 Eden E , Navon R , Steinfeld I , Lipson D , Yakhini Z . GOrilla: a tool for discovery and visualization of enriched GO terms in ranked gene lists . BMC Bioinformatics . 2009 ;10 : 48 10.1186/1471-2105-10-48 19192299 36 Cahoy JD , Emery B , Kaushal A , Foo LC , Zamanian JL , Christopherson KS , et al A transcriptome database for astrocytes, neurons, and oligodendrocytes: a new resource for understanding brain development and function . J Neurosci . 2008 ;28 : 264 –78 . 10.1523/JNEUROSCI.4178-07.2008 18171944 37 Tan PPC , French L , Pavlidis P . Neuron-Enriched Gene Expression Patterns are Regionally Anti-Correlated with Oligodendrocyte-Enriched Patterns in the Adult Mouse and Human Brain . Front Neurosci . 2013 ;7 : 5 10.3389/fnins.2013.00005 23440889 38 Grange P , Bohland JW , Hawrylycz M , Mitra PP . Brain Gene Expression Analysis: a MATLAB toolbox for the analysis of brain-wide gene-expression data . 2012 ; 59 Available: http://arxiv.org/abs/1211.6177 39 Muzio L . Emx1, Emx2 and Pax6 in Specification, Regionalization and Arealization of the Cerebral Cortex . Cereb Cortex . 2003 ;13 : 641 –647 . 10.1093/cercor/13.6.641 12764040 40 Marín F , Nieto MA . The expression of Scratch genes in the developing and adult brain . Dev Dyn . 2006 ;235 : 2586 –91 . 10.1002/dvdy.20869 16773660 41 Bishop KM , Rubenstein JLR , O’Leary DDM . Distinct actions of Emx1, Emx2, and Pax6 in regulating the specification of areas in the developing neocortex . J Neurosci . 2002 ;22 : 7627 –38 . Available: http://www.ncbi.nlm.nih.gov/pubmed/12196586 12196586 42 Stoykova A , Gruss P . Roles of Pax-genes in developing and adult brain as suggested by expression patterns . J Neurosci . 1994 ;14 : 1395 –1412 . Available: http://www.jneurosci.org/content/14/3/1395.short 8126546 43 Talamillo A , Quinn JC , Collinson JM , Caric D , Price DJ , West JD , et al Pax6 regulates regional development and neuronal migration in the cerebral cortex . Dev Biol . 2003 ;255 : 151 –163 . 10.1016/S0012-1606(02)00046-5 12618140 44 Manuel M , Price DJ . Role of Pax6 in forebrain regionalization . Brain Res Bull . 2005 ;66 : 387 –93 . 10.1016/j.brainresbull.2005.02.006 16144620 45 Robertshaw E , Kiecker C . Phylogenetic Origins of Brain Organisers . Scientifica . 2012 pp. 1 –14 . 10.6064/2012/475017 46 Ferrier DE , Holland PW . Ancient origin of the Hox gene cluster . Nat Rev Genet . 2001 ;2 : 33 –8 . 10.1038/35047605 11253066 47 Krumlauf R , Marshall H , Studer M , Nonchev S , Sham MH , Lumsden A . Hox homeobox genes and regionalisation of the nervous system . J Neurobiol . 1993 ;24 : 1328 –40 . 10.1002/neu.480241006 7901322 48 Finnerty JR , Pang K , Burton P , Paulson D , Martindale MQ . Origins of bilateral symmetry: Hox and dpp expression in a sea anemone . Science . 2004 ;304 : 1335 –7 . 10.1126/science.1091946 15131263 49 Domazet-Loso T , Brajković J , Tautz D . A phylostratigraphy approach to uncover the genomic history of major adaptations in metazoan lineages . Trends Genet . 2007 ;23 : 533 –9 . 10.1016/j.tig.2007.08.014 18029048 50 Zhang YE , Landback P , Vibranovski MD , Long M . Accelerated recruitment of new brain development genes into the human genome . PLoS Biol . 2011 ;9 : e1001179 10.1371/journal.pbio.1001179 22028629 51 Kirsch L , Liscovitch N , Chechik G . Localizing Genes to Cerebellar Layers by Classifying ISH Images . Ohler U , editor. PLoS Comput Biol . 2012 ;8 : e1002790 10.1371/journal.pcbi.1002790 23284274 52 Schilling K , Oberdick J , Rossi F , Baader SL . Besides Purkinje cells and granule neurons: an appraisal of the cell biology of the interneurons of the cerebellar cortex . Histochem Cell Biol . 2008 ;130 : 601 –15 . 10.1007/s00418-008-0483-y 18677503 53 Hawrylycz M , Miller JA , Menon V , Feng D , Dolbeare T , Guillozet-Bongaarts AL , et al Canonical genetic signatures of the adult human brain . Nat Neurosci. Nature Publishing Group, a division of Macmillan Publishers Limited. All Rights Reserved .; 2015 ;18 : 1832 –1844 . 10.1038/nn.4171 54 Paoletti P , Bellone C , Zhou Q . NMDA receptor subunit diversity: impact on receptor properties, synaptic plasticity and disease . Nat Rev Neurosci. Nature Publishing Group, a division of Macmillan Publishers Limited. All Rights Reserved .; 2013 ;14 : 383 –400 . 10.1038/nrn3504 55 Monyer H , Burnashev N , Laurie DJ , Sakmann B , Seeburg PH . Developmental and regional expression in the rat brain and functional properties of four NMDA receptors . Neuron . 1994 ;12 : 529 –40 . Available: http://www.ncbi.nlm.nih.gov/pubmed/7512349 7512349 56 Monyer H , Sprengel R , Schoepfer R , Herb A , Higuchi M , Lomeli H , et al Heteromeric NMDA receptors: molecular and functional distinction of subtypes . Science . 1992 ;256 : 1217 –21 . Available: http://www.ncbi.nlm.nih.gov/pubmed/1350383 1350383 57 Flint AC , Maisch US , Weishaupt JH , Kriegstein AR , Monyer H . NR2A Subunit Expression Shortens NMDA Receptor Synaptic Currents in Developing Neocortex . J Neurosci . 1997 ;17 : 2469 –2476 . Available: http://www.jneurosci.org/content/17/7/2469.long 9065507 58 Volpicelli F , Speranza L , di Porzio U , Crispino M , Perrone-Capano C . The serotonin receptor 7 and the structural plasticity of brain circuits . Front Behav Neurosci. Frontiers Media SA ; 2014 ;8 : 318 10.3389/fnbeh.2014.00318 59 Tessier-Lavigne M , Goodman CS , Goodman CS , Shaft CJ , Katz LC , Shatz CJ , et al The molecular biology of axon guidance . Science. American Association for the Advancement of Science ; 1996 ;274 : 1123 –33 . 10.1126/science.274.5290.1123 60 Chen J-G , Rasin M-R , Kwan KY , Sestan N . Zfp312 is required for subcortical axonal projections and dendritic morphology of deep-layer pyramidal neurons of the cerebral cortex . Proc Natl Acad Sci U S A . 2005 ;102 : 17792 –7 . 10.1073/pnas.0509032102 16314561 61 Molyneaux BJ , Arlotta P , Hirata T , Hibi M , Macklis JD . Fezl is required for the birth and specification of corticospinal motor neurons . Neuron . 2005 ;47 : 817 –31 . 10.1016/j.neuron.2005.08.030 16157277 62 Chen B , Schaevitz LR , McConnell SK . Fezl regulates the differentiation and axon targeting of layer 5 subcortical projection neurons in cerebral cortex . Proc Natl Acad Sci U S A . 2005 ;102 : 17184 –9 . 10.1073/pnas.0508732102 16284245 63 Tantirigama MLS , Oswald MJ , Duynstee C , Hughes SM , Empson RM . Expression of the Developmental Transcription Factor Fezf2 Identifies a Distinct Subpopulation of Layer 5 Intratelencephalic-Projection Neurons in Mature Mouse Motor Cortex . J Neurosci . 2014 ;34 : 4303 –4308 . 10.1523/JNEUROSCI.3111-13.2014 24647950 64 Bar-Shira O , Maor R , Chechik G , Aoki-Kinoshita K , Kanehisa M , Kanehisa M , et al Gene Expression Switching of Receptor Subunits in Human Brain Development . Reimers M , editor. PLOS Comput Biol. Public Library of Science ; 2015 ;11 : e1004559 10.1371/journal.pcbi.1004559 65 Eisenberg E , Levanon EY . Human housekeeping genes, revisited . Trends Genet. Elsevier ; 2013 ;29 : 569 –74 . 10.1016/j.tig.2013.05.010 66 Domazet-Loso T , Tautz D . Phylostratigraphic tracking of cancer genes suggests a link to the emergence of multicellularity in metazoa . BMC Biol . 2010 ;8 : 66 10.1186/1741-7007-8-66 20492640

PMC005xxxxxx/PMC5001728.txt

==== Front PLoS OnePLoS ONEplosplosonePLoS ONE1932-6203Public Library of Science San Francisco, CA USA 2756485110.1371/journal.pone.0161875PONE-D-16-13509Research ArticleBiology and Life SciencesImmunologyImmune System ProteinsImmune ReceptorsToll-like ReceptorsMedicine and Health SciencesImmunologyImmune System ProteinsImmune ReceptorsToll-like ReceptorsBiology and Life SciencesBiochemistryProteinsImmune System ProteinsImmune ReceptorsToll-like ReceptorsBiology and Life SciencesCell BiologySignal TransductionImmune ReceptorsToll-like ReceptorsBiology and life sciencesCell biologySignal transductionCoreceptorsCD coreceptorsBiology and life sciencesCell biologySignal transductionCell signalingSignaling cascadesMAPK signaling cascadesBiology and Life SciencesCell BiologyCellular TypesAnimal CellsConnective Tissue CellsFibroblastsBiology and Life SciencesAnatomyBiological TissueConnective TissueConnective Tissue CellsFibroblastsMedicine and Health SciencesAnatomyBiological TissueConnective TissueConnective Tissue CellsFibroblastsBiology and Life SciencesCell BiologyCellular Structures and OrganellesExtracellular MatrixBiology and Life SciencesImmunologyImmune ResponseInflammationMedicine and Health SciencesImmunologyImmune ResponseInflammationMedicine and Health SciencesDiagnostic MedicineSigns and SymptomsInflammationMedicine and Health SciencesPathology and Laboratory MedicineSigns and SymptomsInflammationBiology and Life SciencesGeneticsGene ExpressionBiology and Life SciencesMolecular BiologyMolecular Biology TechniquesArtificial Gene Amplification and ExtensionPolymerase Chain ReactionReverse Transcriptase-Polymerase Chain ReactionResearch and Analysis MethodsMolecular Biology TechniquesArtificial Gene Amplification and ExtensionPolymerase Chain ReactionReverse Transcriptase-Polymerase Chain ReactionHyaluronan Oligosaccharides Induce MMP-1 and -3 via Transcriptional Activation of NF-κB and p38 MAPK in Rheumatoid Synovial Fibroblasts Hyaluronan Oligosaccharides in Rheumatoid Synovial FibroblastsHanabayashi Masahiro Takahashi Nobunori *Sobue Yasumori Hirabara Shinya Ishiguro Naoki Kojima Toshihisa Department of Orthopedic Surgery and Rheumatology, Nagoya University Graduate School of Medicine, Nagoya, JapanKaramanos Nikos K EditorUniversity of Patras, GREECECompeting Interests: The authors have declared that no competing interests exist. Conceptualization: MH NT YS SH NI TK. Formal analysis: MH. Funding acquisition: NT NI TK. Investigation: MH YS. Methodology: MH NT TK. Resources: MH NT NI TK. Supervision: NT TK. Validation: MH NT YS SH NI TK. Visualization: MH. Writing – original draft: MH NT. Writing – review & editing: MH NT. * E-mail: nobunori@med.nagoya-u.ac.jp26 8 2016 2016 11 8 e01618754 4 2016 12 8 2016 © 2016 Hanabayashi et al2016Hanabayashi et alThis is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.Objective To explore the effect of hyaluronan oligosaccharides (HAoligos) on interactions between HA and its principal receptor, CD44, in rheumatoid synovial fibroblasts (RSFs) and matrix metalloproteinase (MMP) production. Methods RSFs were isolated from rheumatoid synovial tissue. HA distribution was visualized by immunocytochemistry. MMP-1 and MMP-3 induction was analyzed by real-time RT-PCR and immunoblotting. The interaction between HAoligos and their MMP-producing receptors was tested by blocking with anti-CD44 and anti-Toll-like receptor 4 (TLR-4). Phosphorylation of nuclear factor κB (NF-κB) and mitogen-activated protein kinase (MAPK) was analyzed by immunoblotting. Results Endogenous HA decreased after treatment with HAoligos, while MMP-1 and MMP-3 expression increased in a dose-dependent manner. Pretreatment with anti-CD44 or anti-TLR-4 antibody significantly reduced the effect of HAoligos on MMP-1 and MMP-3 mRNA expression. NF-κB and p38 MAPK phosphorylation was enhanced by HAoligos pretreated with anti-TLR-4, and HAoligo-induced MMP production was blocked with an inhibitor of NF-κB and p38 MAPK pathways. Conclusions Disruptive changes in CD44-HA interactions by HAoligos enhanced MMP-1 and MMP-3 production via activation of NF-κB and p38 MAPK signaling pathways in RSFs. JSPS KAKENHI Grant Number 25462366This work was supported by a grant from the Japan Society for the Promotion of Science (KAKENHI, #25462366). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data AvailabilityAll relevant data are within the paper and its Supporting Information files.Data Availability All relevant data are within the paper and its Supporting Information files. ==== Body Introduction Rheumatoid arthritis (RA) is a systemic inflammatory disease characterized by joint destruction induced by hyperplasia and chronic inflammation of synovial membranes. Activated fibroblast-like synoviocytes in the lining layer of the synovium contribute significantly to cartilage degradation [1, 2]. Rheumatoid synovial fibroblasts (RSFs) in particular up-regulate the expression of matrix metalloproteinases (MMPs), which are key enzymes that degrade cartilaginous and bone matrices [3]. MMP-1 and MMP-3 are the main MMPs produced by fibroblasts and macrophage-like cells in the synovium, with significantly higher levels found in the synovial fluid of patients with RA compared to patients with osteoarthritis [4, 5]. These MMPs not only degrade collagens, proteoglycans, and other extracellular matrix (ECM) macromolecules in cartilage, but also activate other MMPs [6]. The actions of MMP-1 and MMP-3 lead to destruction of articular cartilage and subchondral bone, resulting in joint deformity and severe pain for patients with RA. It is therefore vital to elucidate the mechanisms of MMP-induced joint destruction and develop targeted treatment plans. Many connective tissue cells have a large hyaluronan (HA) and proteoglycan-rich pericellular matrix that is tethered to the cell surface via interactions with primary HA receptor CD44 [7–9]. HA is a high-molecular weight polysaccharide consisting of repeating disaccharide glucuronic acid and N-acetylglucosamine units [10]; it is also an abundant component of ECMs and plays a role in regulating cell behavior during embryonic development, healing processes, inflammation, and tumor development [11]. A previous study found that decreasing extracellular HA in chondrocytes induced stimulation of catabolic genes involved in matrix degradation, such as MMPs [12]. However, the effects in RSFs are still uncertain, despite RSFs being major regulators of local inflammation in RA joints. In one study, the concentration of HA in synovial fluid of patients with RA was significantly less compared to healthy individuals [13]. In other studies, lower concentrations of HA were found in the synovia of patients with RA [14, 15]. These observations raise the possibility that decreased pericellular HA levels contribute to MMP expression in RSFs. HA oligosaccharides (HAoligos), of the size of a HA hexasaccharide—decasaccharide, can compete with the binding of high-molecular weight HA (HMW-HA) to CD44 to displace it and the proteoglycan-rich pericellular matrix from the cell surface [7, 8, 12, 16–22]. Loss of the cell-associated pericellular matrix and uncoupling of HA from CD44 results in signal transduction events with outcomes based on cell type [8, 12, 16, 18–21]. The current CD44-HA model clusters CD44 receptors in the plasma membrane using multivalent interactions with HMW-HA; the cluster is reversed with the addition of excess HAoligos that displace endogenous HA and generate monovalent interactions between itself and CD44 [11, 16, 21]. CD44 is a transmembrane receptor with a short intracellular tail domain but no inherent kinase activity [23]; signaling events induced by the unclustering of CD44 by HAoligos are thus likely to involve the activation of CD44-associated proteins, such as kinases and cytoskeletal elements [24]. Although in vivo fragmentation of HA has yet to be elucidated, activated fibroblasts secrete hyaluronidase to degrade HMW-HA into fragments during inflammation [25, 26]. This degradation can be augmented by reactive oxygen species produced at inflammation sites [27, 28]. In this study, HAoligos were used as a molecular tool to mimic decreasing HA and disrupted CD44-HA interactions. We investigated the effect of HAoligo treatment on MMP production in RSFs and explored the CD44-dependent signaling pathway responsible for MMP production by HAoligos. Materials and Methods Reagents Mouse anti-human CD44 (clone BU52) monoclonal antibody was purchased from Ancell (Bayport, MN, USA). Mouse anti-human Toll-like receptor 4 (TLR-4; clone HTA125) monoclonal antibody was purchased from Abcam (Cambridge, UK). For immunoblot analysis, mouse anti-human MMP-1 (clone 41-1E5) and MMP-3 (clone 55-2A4) monoclonal antibodies were purchased from Daiichi Fine Chemical (Toyama, Japan). Rabbit anti-human nuclear factor κB (NF-κB) p65 (#3987), phospho-NF-κB p65 (#3033), p38 mitogen-activated protein (p38 MAPK; #9212), phospho-p38 MAPK (#4511), Erk1/2 (#4695), phospho-Erk1/2 (#4370), c-Jun N-terminal kinase (JNK; #9252), phospho-JNK (#9251), and β-actin antibodies (#4970) were purchased from Cell Signaling Technology (Beverly, MA, USA). Horseradish peroxidase (HRP)-conjugated goat anti-mouse (#7076) and anti-rabbit (#7074) IgG were also obtained from Cell Signaling Technology. Dulbecco’s modified Eagle’s medium (DMEM) and trypsin ethylenediaminetetraacetic acid (EDTA) were obtained from Sigma-Aldrich Co. (St. Louis, MO, USA). Fetal bovine serum (FBS) was purchased from PAA Laboratories GmbH (Pasching, Austria). Cell culture Synovial tissues were obtained from six patients with RA at the site of total knee replacement surgery that fulfilled the revised criteria of the American College of Rheumatology (1987)[29]. Written informed consent was obtained from each patient prior to the study, and experiments were performed in accordance with a study protocol approved by the Ethics committee of Nagoya University. The tissue was minced into small pieces and digested with 4 mg/ml collagenase (Sigma-Aldrich) in DMEM for two hours at 37°C. After removing tissue debris by filtering through a 70-μm cell strainer (BD Biosciences, Franklin Lakes, NJ, USA), cells were centrifuged for five minutes at 1,500 rpm. The cell pellet was resuspended in DMEM containing 100 U/ml penicillin, 100 μg/ml streptomycin, 0.25 μg/ml amphotericin B (Invitrogen, Carlsbad, CA, USA), and 10% heat-inactivated FBS, plated on 75 cm2 tissue culture flasks (TPP Techno Plastic Products AG, Schaffhausen, Switzerland), and cultured at 37°C in a humidified 5% CO2 atmosphere. At confluence, cells were passaged 1:3 by treatment with EDTA. Cells with 3–5 passages were considered to be RSFs. Before experiments, RSFs were seeded into six-well culture plates (TPP) at a density of 4.0 × 105 cells/well. In some experiments, RSFs were seeded into 22.1 cm2 dishes (TPP) at a density of 1.0 × 106 cells/dish. Preparation of hyaluronan oligosaccharides HAoligos were refined as previously described [30]. Briefly, HMW-HA from human umbilical cord (Sigma-Aldrich) was dissolved in 0.1 M sodium acetate buffer (pH 5.0) containing 0.15 M NaCl and treated with bovine testicular hyaluronidase (320 units/mg, type I-S; Sigma-Aldrich) for 16 hours at 37°C. Duration of incubation and temperature were selected to best generate a predominant proportion of HA hexasaccharides [20]. At the end of the reaction, hyaluronidase was heat-inactivated, precipitated with 10% trichloroacetic acid, and pelleted by centrifugation at 2,000 rpm for 10 minutes. The oligosaccharide-containing supernatant was dialyzed in 1,000 molecular weight cut-off (MWCO) dialysis tubing (Spectrum Medical Industries, Houston, TX, USA) with four changes of H2O over 48 hours, lyophilized, and redissolved in sterile phosphate buffer saline (PBS). Hyaluronan staining with hyaluronic acid binding protein HA distribution in RSFs after treatment with HAoligos was visualized by staining with a biotinylated HA binding protein (b-HABP; Seikagaku Biobusiness Co., Tokyo, Japan) that has a high affinity for the decasaccharide unit of HA. RSFs were seeded into chamber slides (BD Biosciences) for 48 hours and treated with or without 250 μg/ml HAoligos for one or three hours. Cells were then fixed with 4% paraformaldehyde, buffered with PBS at room temperature for one hour, treated with 0.3% H2O2 in PBS for 30 minutes at room temperature to block internal peroxidase activity, and incubated with 1% bovine serum albumin in PBS for one hour at room temperature. Cells were incubated with 2.0 μg/ml b-HABP for one hour at room temperature, and bound b-HABP was detected by the addition of streptavidin-peroxidase reagents and diaminobenzidine (DAB)-containing substrate solution (Nichirei Biosciences Inc., Tokyo, Japan). As a negative control, cells were pretreated with 100 units/ml bovine testicular hyaluronidase (Sigma-Aldrich) for three hours before incubation with b-HABP. Hyaluronan oligosaccharide treatment Forty-eight hours after seeding, RSFs were washed with PBS 2 times and brought to serum-free conditions overnight prior to beginning the experiment. To investigate MMP-1 and MMP-3 mRNA expression and protein secretion, RSFs were treated with various concentrations of HAoligos (0–500 μg/ml) in serum-free DMEM for 12 or 24 hours. In some experiments, cells were pretreated with 5 μg/ml anti-CD44 or 5 μg/ml anti-TLR-4 antibody for one hour before HAoligo treatment. To investigate intracellular signaling pathways, RSFs were pretreated with anti-TLR-4 antibody followed by 250 μg/ml HAoligos for 0–120 min. For signal inhibition assays, RSFs were pretreated with inhibitors for NF-κB (Helenalin; Merck KGaA, Darmstadt, Germany) or p38 MAPK (SB203580; Merck KGaA) for one hour, followed by the addition of 250 μg/ml HAoligos for 24 hours in the presence or absence of each inhibitor. To examine the effects of HMW-HA on HAoligo-induced MMP mRNA expression, RSFs were pretreated with 250 μg/ml HAoligos for 12 hours and followed with 250 μg/ml HAoligos or 1 mg/ml HMW-HA (600–1,200 kDa; Seikagaku Co., Tokyo, Japan) for 12 hours. Real-time RT-PCR Total RNA was isolated from RSF cultures using the illustra triplePrep kit (GE Healthcare UK Ltd, Buckinghamshire, UK), and reverse transcription was performed with the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, CA, USA) according to manufacturer protocols. Equal amounts of each reverse transcription product were amplified by real-time RT-PCR using the LightCycler FastStart DNA Master SYBR Green I kit (Roche, Mannheim, Germany). Thermal cycling and fluorescence detection were performed using the LightCycler 480 System (Roche, USA). MMP-1 and MMP-3 mRNA expression levels were normalized to glyceraldehyde phosphate dehydrogenase (GAPDH) as fold change compared to control at each time point. MMP-1, MMP-3, and GAPDH primer pairs are as follows: MMP-1, forward 5’-CTGGGAGCAAACACATCTGA-3’, reverse 5’-CTGGTTGAAAAGCATGAGCA-3’; Gen Bank ID: NM_002421. MMP-3, forward 5’-GAAGACTTTCCAGGGATTGACT-3’, reverse 5’-GTGCCTTCTACTACTCTTTCAAC -3’; Gen Bank ID: NM_002422. GAPDH, forward 5’-TGAACGGGAAGCTCACTGG-3’, reverse 5’-TCCACCACCCTGTTGCTGTA-3’; Gen Bank ID: NM_008084. All primers were designed by Nihon Gene Research Laboratories (Miyagi, Japan). DNA assay After conditioned media was collected, genomic DNA was isolated from RSF cultures using the illustra triplePrep kit (GE Healthcare). DNA content was determined by BioPhotometer plus (Eppendorf, Hamburg, Germany). Western blot analysis To investigate MMP-1 and MMP-3 secreted into conditioned media, medium samples were collected in the presence of a protease inhibitor cocktail (Thermo Fisher Scientific Inc., Waltham, MA, USA), according to manufacturer protocols. Samples were centrifuged at 14,000 g for 10 minutes at 4°C. After centrifugation, supernatants were mixed with Lane Marker Reducing Sample buffer (Thermo Fisher Scientific) and heated at 80°C for 10 minutes. Samples were then separated by electrophoresis on 10% SDS-polyacrylamide gel and electroblotting onto polyvinylidene fluoride membranes using a Mini Trans-Blot apparatus (BIO-RAD, Hercules, CA, USA). The sample amount loaded was determined by the DNA content of RSFs in each well. After blocking in 0.5% skim milk for one hour, membranes were incubated overnight at 4°C with primary antibodies against MMP-1 or MMP-3. After washing with PBS containing 0.1% Tween 20, membranes were incubated for 30 minutes with HRP–conjugated secondary antibodies. Detection was performed using chemiluminescence with West Pico Chemiluminescent Substrate (Thermo Fisher Scientific). To investigate intracellular signaling pathways, total protein was extracted from RSFs treated 0–120 minutes with or without HAoligos using Cell Lysis Buffer (Cell Signaling Technology) containing protease and phosphatase inhibitor cocktails (Thermo Fisher Scientific). Protein content was determined using protein assay reagent (BIO-RAD) and BSA as a standard. In some cases, blots were treated with striping buffer and re-probed using another primary antibody. Band intensities were captured with a digital image scanner and quantified using densitometry software (CS Analyzer 3.0; ATTO, Tokyo, Japan). Casein zymography To investigate the activity of MMPs secreted into conditioned media, media samples were collected, centrifuged at 14,000 g for 10 minutes at 4°C, and the resultant supernatants were incubated at 37°C with 2.5 mM p-aminophenylmercuric acetate for four hours to activate MMPs. Samples were mixed with Tris-Glycine SDS Sample Buffer (Thermo Fisher Scientific) and separated by electrophoresis on 12% Zymogram (Casein) Protein Gels (Thermo Fisher Scientific). The amount of sample loaded was determined based on the DNA content of RSFs in each well. After electrophoresis, gels were incubated in Zymogram Renaturing Buffer (Thermo Fisher Scientific) for 30 minutes at room temperature with gentle agitation and subsequently in Zymogram Developing Buffer (Thermo Fisher Scientific) for 30 minutes. Gels were then incubated in fresh Zymogram Developing Buffer at 37°C overnight, stained with Coomassie brilliant blue overnight, and bleached with 45% methanol / 10% acetic acid. Statistical analysis Values are expressed as mean ± SD. Statistical significance was analyzed by Student’s t-test. P values less than 0.05 were considered significant. All analyses were conducted with SPSS for Windows ver. 19 (SPSS, Chicago, IL). Results HAoligo treatment of RSFs reduces endogenous HA To assess whether HAoligos can displace HMW-HA from the cell surface, RSFs were cultured with or without HAoligos for three hours. DAB staining for HA (brown) was abundant on the cell surface or in the cytoplasm of untreated RSFs (Fig 1A). In contrast, stainable HA decreased after treatment with HAoligos (Fig 1B) or hyaluronidase (Fig 1C). After a one hour incubation with HAoligos, slight displacement of HMW-HA was observed (S1 Fig). These results suggest that HAoligos can displace HMW-HA from the cell surface. 10.1371/journal.pone.0161875.g001Fig 1 Effect of hyaluronan oligosaccharides (HAoligos) on endogenous HA deposition in rheumatoid synovial fibroblasts (RSFs). HA accumulation in RSFs after treatment with or without HAoligos was visualized by diaminobenzidine (DAB) staining using a biotinylated HA binding protein. A, Control medium. B, 250 mg/ml HAoligos. C, 400 units/ml bovine testicular hyaluronidase. Original magnification × 400, scale bar: 200 μm. Effect of HAoligos on MMP-1 and MMP-3 mRNA expression and protein secretion To explore the mechanism of MMP-1 and MMP-3 induction, cultured RSFs were treated with various concentrations of HAoligos for 12 or 24 hours. As shown in Fig 2A, induction of MMP-1 mRNA by HAoligos was first observed at 50 μg/ml and increased in a dose-dependent manner, with maximum enhancement at 250 μg/ml. Consistent with the mRNA findings, MMP-1 protein secreted into culture media also increased with HAoligo treatment (Fig 2B). MMP-3 mRNA expression and protein secretion were similar to that observed with MMP-1. Moreover, MMPs secreted into conditioned media had caseinolytic activity (Fig 2C). These results indicate that HAoligos affect MMP-1 and MMP-3 induction at both the transcript and protein level. 10.1371/journal.pone.0161875.g002Fig 2 Expression levels of matrix metalloproteinase (MMP) mRNA and protein secretion induced by HAoligo treatment. RSFs were treated with various concentrations of HAoligos (0–500 μg/ml). A, MMP-1 and MMP-3 mRNA expression levels after 12 or 24 hours of HAoligo treatment as detected by quantitative real-time RT-PCR. Results shown are fold change in mRNA copy number. Values are mean ± S.D. of three independent experiments analyzed in triplicate, and are expressed as n-fold increase compared to untreated control. *p<0.05 and **p<0.01 vs. control. B, MMP-1 and MMP-3 protein secretion in conditioned media after 24 hours of HAoligo treatment as evaluated by immunoblot analysis. The amount of sample separated by SDS-PAGE before transfer and immunostaining was determined by the DNA content of RSFs. Results shown are representative of three experiments with similar results. C, The activity of MMPs secreted into conditioned media was evaluated by casein zymography. RSFs were stimulated with 0, 100, or 250 μg/ml HAoligos for 24 hours. The image was digitally inverted to generate dark bands on a light background. Involvement of TLR-4 and CD44 in HAoligo-induced MMP mRNA expression Pattern recognition receptors such as TLR-4 are involved in HAoligo-induced cell signaling [31–36], raising the possibility that they play a role in HAoligo-mediated MMP production in RSFs. The effect of HAoligos on MMP mRNA expression via interaction with CD44 and TLR-4 was thus evaluated using neutralizing antibodies against each receptor. As shown in Fig 3, MMP-1 and MMP-3 mRNA expression induced by 250 μg/ml HAoligo treatment significantly decreased with a 5 μg/ml CD44 neutralizing antibody. Similar results were obtained in RSFs pretreated with a 5 μg/ml TLR-4 neutralizing antibody. Furthermore, no cell toxicity was observed for the indicated concentrations of antibodies (data not shown). These findings show that HAoligos interact with CD44 as well as TLR-4 to induce MMP mRNA expression. 10.1371/journal.pone.0161875.g003Fig 3 Expression levels of MMP mRNA induced by HAoligo treatment with or without CD44 and Toll-like receptor 4 (TLR-4) neutralizing antibodies. RSFs were pre-treated for one hour with or without 5μg/ml anti-CD44 and 5μg/ml TLR-4 antibodies and followed by HAoligo treatment (250 μg/ml) for 24 hours. MMP-1 and MMP-3 expression levels were determined by quantitative real-time RT-PCR. Values are mean ± S.D of three independent experiments analyzed in triplicate, and are expressed as n-fold increase compared to untreated control. **p<0.01 vs. control. TLR-4-independent activation of NF-κB and p38 MAPK by HAoligos Activation of NF-κB, p38 MAPK, ERK, and JNK by HAoligo treatment was examined in RSFs. After stimulation with 250 μg/ml HAoligos, the cell lysate was analyzed by immunoblot. HAoligos enhanced the phosphorylation of NF-κB and p38 MAPK, while JNK and Erk phosphorylation levels were equivalent to control samples (S5 Fig). To assess the involvement of TLR-4 pathways, RSFs were also pretreated with a TLR4 neutralizing antibody. HAoligo treatment of these cells increased the phosphorylation of NF-κB and p38 MAPK, while ERK and JNK phosphorylation levels were equivalent to control samples (Fig 4A and 4B). Total protein levels were unchanged during the experiment. These results indicate that HAoligo treatment resulted in activation of NF-κB and p38 MAPK through a TLR-4 independent pathway. 10.1371/journal.pone.0161875.g004Fig 4 Intracellular signaling cascades stimulated by HAoligos with pre- and co-incubation with a TLR4 neutralizing antibody. RSFs were pre-treated for one hour with an anti-TLR4 antibody, followed by HAoligo treatment (250 μg/ml). Phosphate buffered saline was substituted as a control. A, Phospho-nuclear factor κB (NF-κB), NF-κB, phospho-p38 mitogen-activated protein kinase (MAPK), p38 MAPK, phospho-Erk, Erk, phospho-c-Jun N-terminal kinase (JNK), and JNK, as evaluated by immunoblot analysis of RSFs derived from three patients with RA. Results shown are representative of three experiments with similar results. B, Densitometry of phospho-NF-κB/NF-κB, phospho-p38 MAPK/p38 MAPK, phospho-ERK/ERK and phospho-JNK/JNK. Values are mean ± S.D. of three independent experiments and expressed as n-fold increase compared to untreated control at time 0. *p<0.05 and **p<0.01 vs. control. Requirement of NF-κB and p38 MAPK for MMP production in HAoligo-stimulated RSFs To investigate whether NF-κB and p38 MAPK were responsible for the production of MMP-1 and MMP-3 stimulated by HAoligos in RSFs, selective protein kinase inhibitors were used. As shown in Fig 5, pre- and co-incubation with inhibitors for NF-κB (helenalin) or p38 MAPK (SB203580) resulted in significant decreases in MMP-1 and MMP-3 production stimulated by HAoligos. These findings suggest that MMP-1 and MMP-3 production induced by HAoligos involves NF-κB and p38 MAPK. 10.1371/journal.pone.0161875.g005Fig 5 Effects of chemical inhibitors on MMP production induced by HAoligo treatment. After pretreatment with helenalin or SB203580 for one hour at the indicated concentrations, RSFs were stimulated with 250 μg/ml HAoligos for 24 hours in the presence or absence of each inhibitor. The conditioned media was analyzed by immunoblot using antibodies against MMP-1 and MMP-3. The amount of sample separated by SDS-PAGE before transfer and immunostaining was determined by the DNA content of RSFs in each well. RSFs derived from three patients with RA were used. In donor A, SB203580 lanes were not adjacent to helenalin lanes. Thus, for clarity, SB203580 lanes were cut out and displayed next to helenalin lanes. Effect of HMW-HA on MMP-1 and MMP-3 mRNA expression induced by HAoligo treatment To determine whether HMW-HA added to cells pretreated with HAoligos could reduce MMP-1and MMP-3 up-regulation, MMP mRNA expression was analyzed. As shown in Fig 6, HAoligo-induced MMP-1 and MMP-3 mRNA expression was significantly attenuated following the addition of HMW-HA. These results suggest that HMW-HA reduces MMP-1 and MMP-3 expression induced by HAoligos 10.1371/journal.pone.0161875.g006Fig 6 Effect of HMW-HA on MMP mRNA expression induced by HAoligo treatment. RSFs were treated with HAoligos (250 μg/ml) for 12 hours. In next 12 hours, RSFs were treated with HAoligos (250 μg/ml) or HMW-HA (1mg/ml). MMP-1 and MMP-3 expression levels were determined by quantitative real-time RT-PCR. Values are presented as mean ± S.D of two independent experiments analyzed in triplicate, and are expressed as n-fold increase relative to untreated control. **p<0.01 vs. HAoligos. Discussion In the present study, low-molecular mass HAoligos were used to disrupt CD44-HA interactions. We demonstrated that these small oligosaccharides displace HMW-HA from the cell surface (Fig 1) and influence the up-regulation of MMP-1 and MMP-3 at both the transcript and protein level, thereby leading to matrix degradation (Fig 2). Previous studies demonstrated that HAoligos increased MMPs in articular chondrocytes [12, 30, 37]; however, fibroblast-like synoviocytes are more critical for understanding the pathology of RA joint destruction rather than chondrocytes in cartilage tissue. Synoviocytes are the major cellular constituent of synovium, the principal tissue involved in the regulation of inflamed joints, and the primary reason HAoligo treatment in RSFs was explored. The receptor involved in HA-mediated MMP induction remains somewhat ambiguous, as HA can associate with several cell surface molecules. Although CD44 is the best known HA receptor [9], HAoligos can also interact with TLR-4 to promote cytokine and MMP up-regulation [31–36]. TLRs are pattern recognition receptors found in a variety of cells and tissues; they can detect invading microbes as well as sense tissue damage and, in response, induce expression of pro-inflammatory genes [38, 39]. To determine the HA receptor involved in MMP up-regulation, we performed an inhibition assay using neutralizing antibodies to CD44 and TLR-4. Our data showed that although TLR-4 is involved in HAoligo-induced MMP-1 and MMP-3 mRNA expression in RSFs, CD44 mediates HAoligo-initiated cell signaling (Fig 3). NF-κB and MAPK families play a critical role in the induction of MMPs [40]. Previous studies with chondrocytes reported that NF-κB and p38 MAPK are involved in MMP-3 and MMP-13 production through CD44-HAoligos interactions [12, 30]. Another study determined that p38 MAPK plays a role in signal transduction for MMP-1 induced by HAoligos in periodontal ligament cells [41]. We found that CD44-HAoligo interaction increased the activation of NF-κB and p38 MAPK, which led to MMP-1 and MMP-3 secretion in RSFs (Figs 4 and 5). A possible explanation for the different signaling pathways responsible for MMP production may be that it is dependent on cell type. Recent clinical trials have clearly shown that biologics targeting inflammatory cytokines can achieve good clinical results in the treatment of RA patients; however, complete suppression of joint destruction has yet to be achieved. Joint destruction pathways other than cytokines are believed to contribute to RA pathogenesis independent of biologic treatment strategies. Innate immunity is one possible pathway, and our findings suggest that decreasing HA in RA joints may also lead to destroyed joints. HMW-HA is a biodegradable, biocompatible, nontoxic, non-immunogenic, and non-inflammatory linear polysaccharide. Given its natural presence in the body, HA has advantages over synthesized molecules in terms of biophysiological effects and possible side effects that make it an attractive drug. Recent studies have shown that HA plays an important role in joint lubrication, protects articular cartilage from damage, and acts as a biological inhibitor of inflammation and degradation of joints by down-regulating gene expression of osteoarthritis-associated enzymes, such as MMP-3 in fibroblast-like synoviocytes [42], and inhibiting MMP-1 and MMP-3 production by RSFs [43, 44]. Restoration of HA to normal levels in RA synovial fluid could therefore decrease elevated MMP-1 and MMP-3 levels in RA joints. Although some studies have reported the clinical efficacy of HA intra-articular injection in RA patients [45, 46], only a few have reported that HA has the potential to prevent joint destruction. In the present study, HMW-HA attenuated the HAoligo-induced upregulation of MMP-1 and MMP-3 mRNA (Fig 6). Our findings provide new data to support the efficacy of HA treatment in RA. There are some limitations to this study. First, the prepared HAoligos may have been contaminated, e.g., by hyaluronidase. However, we performed preliminary experiments to rule out contamination. For instance, we found that the HA-free solution did not affect the expression levels of MMP-1 and -3, suggesting that the HA-free solution was free of contamination and contained only HMW-HA (S2 Fig). Furthermore, HMW-HA was able to rescue mRNA levels of MMP-1 and -3 induced by HAoligos. These findings collectively suggest that the HMW-HA used was essentially free of contamination. Second, it is possible that other untested HA receptors, such as the receptor for hyaluronan-mediated motility (RHAMM), may have caused the observed effects. Further studies will be needed to determine whether signaling through other receptors are affected by treatment with HAoligos. Third, other effects of HAoligos on RSFs were not analyzed. This study demonstrated that HAoligos increased the expression of matrix degradation-related factors in RSFs. In other cell types, such as articular chondrocytes, HAoligos reportedly induce both catabolic and anabolic factors simultaneously [30]. Thus, further studies regarding other effects of HAoligos in RSFs are warranted. In conclusion, we confirmed that disruptive changes in CD44-HA interactions by HAoligos enhance MMP-1 and MMP-3 production via activation of NF-κB and p38 MAPK signaling pathways in RSFs. This study illustrates that loss of the pericellular matrix leads to the activation of a catabolic cascade by RSFs, and the joint destruction pathway observed could provide new evidence for the biological effects of intra-articular supplementation of HMW-HA in inflammatory RA joints. Supporting Information S1 Fig HA distribution in rheumatoid synovial fibroblasts (RSFs) after treatment with hyaluronan oligosaccharides (HAoligos). HA accumulation in RSFs after treatment with or without HAoligos for one hour was visualized by diaminobenzidine (DAB) staining using a biotinylated HA binding protein. Compared to results of longer incubation (three hours), slighter displacement of HMW-HA was observed after incubation with HAoligos for one hour. A, Control medium. B, 250 mg/ml HAoligos. Original magnification × 400, scale bar: 200 μm. (TIF) Click here for additional data file. S2 Fig Effect of hyaluronan-free solution on MMP mRNA expression. Hyaluronan (HA)-free solution was prepared in the same manner as the preparation of hyaluronan oligosaccharides, but without HMW-HA, and dialyzed with PBS in the final dialysis. HAoligos in PBS were also prepared using conventional methods and dialyzed with PBS in the final dialysis. The concentration of HAoligos in PBS was calculated using the weight after lyophilization. RSFs were treated with the same amount of HA-free solution as the amount of HAoligos in PBS. HAoligos in PBS significantly affected MMP mRNA expression, whereas the HA-free solution did not. *p<0.05 and **p<0.01 vs. HAoligos in PBS. (TIF) Click here for additional data file. S3 Fig Effects of isotype-matched control IgG on MMP mRNA expression induced by HAoligos. RSFs were pre-treated for one hour with or without antibodies (5 μg/ml), and followed by treatment with HAoligos (250 μg/ml) for 24 hours. MMP mRNA expression induced by HAoligos significantly decreased when RSFs were pre-treated with CD44 or TLR-4 neutralizing antibodies. In contrast, isotype-matched control IgG of anti-CD44 or -TLR4 antibodies did not reduce HAoligo-induced MMP mRNA expression. IgG1: isotype-matched control IgG of anti-CD44 antibody (Ancell). IgG2a: isotype-matched control IgG of anti-CD44 TLR4 antibody (Abcam). *p<0.05 and **p<0.01 vs. control IgG. (TIF) Click here for additional data file. S4 Fig Suppressive effect of BU52 and BU75 anti-CD44 antibodies on MMP mRNA expression induced by HAoligos. RSFs were pre-treated for one hour with or without antibodies (5 μg/ml), and followed by treatment with HAoligos (250 μg/ml) for 24 hours. The suppressive effect of the BU52 antibody on MMP mRNA expression induced by HAoligos was comparable to that of the BU75 antibody (Ancell). **p<0.01 vs HAoligos. (TIF) Click here for additional data file. S5 Fig Intracellular signaling pathways stimulated by HAoligos. Total protein was extracted from RSFs treated for 0–120 minutes with or without HAoligos (250 μg/ml). Phosphate buffered saline was used as the control. Levels of phospho-NF-κB, NF-κB, phospho-p38 MAPK, p38 MAPK, phospho-Erk, Erk, phospho- JNK, and JNK were evaluated by immunoblot analysis. HAoligos enhanced the phosphorylation of NF-κB and p38 MAPK, while JNK and Erk phosphorylation levels were equivalent to control samples. (TIF) Click here for additional data file. ==== Refs References 1 Pap T , van der Laan WH , Aupperle KR , Gay RE , Verheijen JH , Firestein GS , et al Modulation of fibroblast-mediated cartilage degradation by articular chondrocytes in rheumatoid arthritis . Arthritis and rheumatism . 2000 ;43 (11 ):2531 –6 . Epub 2000/11/18. 10.1002/1529-0131(200011)43:11<2531::AID-ANR21>3.0.CO;2-V .11083277 2 Firestein GS . Evolving concepts of rheumatoid arthritis . Nature . 2003 ;423 (6937 ):356 –61 . Epub 2003/05/16. 10.1038/nature01661 .12748655 3 Mengshol JA , Mix KS , Brinckerhoff CE . Matrix metalloproteinases as therapeutic targets in arthritic diseases: bull's-eye or missing the mark? Arthritis and rheumatism . 2002 ;46 (1 ):13 –20 . Epub 2002/01/31. .11817584 4 Konttinen YT , Ainola M , Valleala H , Ma J , Ida H , Mandelin J , et al Analysis of 16 different matrix metalloproteinases (MMP-1 to MMP-20) in the synovial membrane: different profiles in trauma and rheumatoid arthritis . Annals of the rheumatic diseases . 1999 ;58 (11 ):691 –7 . Epub 1999/10/26. 10531073 5 Yoshihara Y , Nakamura H , Obata K , Yamada H , Hayakawa T , Fujikawa K , et al Matrix metalloproteinases and tissue inhibitors of metalloproteinases in synovial fluids from patients with rheumatoid arthritis or osteoarthritis . Annals of the rheumatic diseases . 2000 ;59 (6 ):455 –61 . Epub 2000/06/02. 10834863 6 Chakraborti S , Mandal M , Das S , Mandal A , Chakraborti T . Regulation of matrix metalloproteinases: an overview . Molecular and cellular biochemistry . 2003 ;253 (1–2 ):269 –85 . Epub 2003/11/19. .14619979 7 Knudson W , Aguiar DJ , Hua Q , Knudson CB . CD44-anchored hyaluronan-rich pericellular matrices: an ultrastructural and biochemical analysis . Experimental cell research . 1996 ;228 (2 ):216 –28 . Epub 1996/11/01. 10.1006/excr.1996.0320 .8912714 8 Evanko SP , Angello JC , Wight TN . Formation of hyaluronan- and versican-rich pericellular matrix is required for proliferation and migration of vascular smooth muscle cells . Arteriosclerosis, thrombosis, and vascular biology . 1999 ;19 (4 ):1004 –13 . Epub 1999/04/09. .10195929 9 Aruffo A , Stamenkovic I , Melnick M , Underhill CB , Seed B . CD44 is the principal cell surface receptor for hyaluronate . Cell . 1990 ;61 (7 ):1303 –13 . Epub 1990/06/29. .1694723 10 Laurent TC , Fraser JR . Hyaluronan . FASEB journal: official publication of the Federation of American Societies for Experimental Biology . 1992 ;6 (7 ):2397 –404 . Epub 1992/04/01. .1563592 11 Toole BP . Hyaluronan: from extracellular glue to pericellular cue . Nature reviews Cancer . 2004 ;4 (7 ):528 –39 . Epub 2004/07/02. 10.1038/nrc1391 .15229478 12 Ohno S , Im HJ , Knudson CB , Knudson W . Hyaluronan oligosaccharides induce matrix metalloproteinase 13 via transcriptional activation of NFkappaB and p38 MAP kinase in articular chondrocytes . The Journal of biological chemistry . 2006 ;281 (26 ):17952 –60 . Epub 2006/05/02. 10.1074/jbc.M602750200 16648633 13 Dahl LB , Dahl IM , Engstrom-Laurent A , Granath K . Concentration and molecular weight of sodium hyaluronate in synovial fluid from patients with rheumatoid arthritis and other arthropathies . Annals of the rheumatic diseases . 1985 ;44 (12 ):817 –22 . Epub 1985/12/01. 4083937 14 Pitsillides AA , Worrall JG , Wilkinson LS , Bayliss MT , Edwards JC . Hyaluronan concentration in non-inflamed and rheumatoid synovium . British journal of rheumatology . 1994 ;33 (1 ):5 –10 . Epub 1994/01/01. .8162458 15 Uzuki H , Watanabe T , Katsura Y , Sawai T . Quantitative histochemical study of hyaluronic acid binding protein and the activity of uridine diphosphoglucose dehydrogenase in the synovium of patients with rheumatoid arthritis . Analytical and quantitative cytology and histology / the International Academy of Cytology [and] American Society of Cytology . 1999 ;21 (1 ):75 –80 . Epub 1999/03/09. .10068780 16 Ghatak S , Misra S , Toole BP . Hyaluronan oligosaccharides inhibit anchorage-independent growth of tumor cells by suppressing the phosphoinositide 3-kinase/Akt cell survival pathway . The Journal of biological chemistry . 2002 ;277 (41 ):38013 –20 . Epub 2002/07/30. 10.1074/jbc.M202404200 .12145277 17 Knudson CB . Hyaluronan receptor-directed assembly of chondrocyte pericellular matrix . The Journal of cell biology . 1993 ;120 (3 ):825 –34 . Epub 1993/02/01. 7678838 18 Zoltan-Jones A , Huang L , Ghatak S , Toole BP . Elevated hyaluronan production induces mesenchymal and transformed properties in epithelial cells . The Journal of biological chemistry . 2003 ;278 (46 ):45801 –10 . Epub 2003/09/05. 10.1074/jbc.M308168200 .12954618 19 Ward JA , Huang L , Guo H , Ghatak S , Toole BP . Perturbation of hyaluronan interactions inhibits malignant properties of glioma cells . The American journal of pathology . 2003 ;162 (5 ):1403 –9 . Epub 2003/04/23. 10.1016/S0002-9440(10)64273-3 12707023 20 Knudson W , Casey B , Nishida Y , Eger W , Kuettner KE , Knudson CB . Hyaluronan oligosaccharides perturb cartilage matrix homeostasis and induce chondrocytic chondrolysis . Arthritis and rheumatism . 2000 ;43 (5 ):1165 –74 . Epub 2000/05/19. 10.1002/1529-0131(200005)43:5<1165::AID-ANR27>3.0.CO;2-H .10817571 21 Knudson CB , Knudson W . Hyaluronan and CD44: modulators of chondrocyte metabolism . Clinical orthopaedics and related research . 2004 ;(427 Suppl ):S152 –62 . Epub 2004/10/14. .15480059 22 Lesley J , Hascall VC , Tammi M , Hyman R . Hyaluronan binding by cell surface CD44 . The Journal of biological chemistry . 2000 ;275 (35 ):26967 –75 . Epub 2000/06/29. 10.1074/jbc.M002527200 .10871609 23 Ponta H , Sherman L , Herrlich PA . CD44: from adhesion molecules to signalling regulators . Nature reviews Molecular cell biology . 2003 ;4 (1 ):33 –45 . Epub 2003/01/04. 10.1038/nrm1004 .12511867 24 Girish KS , Kemparaju K , Nagaraju S , Vishwanath BS . Hyaluronidase inhibitors: a biological and therapeutic perspective . Current medicinal chemistry . 2009 ;16 (18 ):2261 –88 . Epub 2009/06/13. .19519390 25 Weigel PH , Fuller GM , LeBoeuf RD . A model for the role of hyaluronic acid and fibrin in the early events during the inflammatory response and wound healing . Journal of theoretical biology . 1986 ;119 (2 ):219 –34 . Epub 1986/03/21. .3736072 26 Sampson PM , Rochester CL , Freundlich B , Elias JA . Cytokine regulation of human lung fibroblast hyaluronan (hyaluronic acid) production. Evidence for cytokine-regulated hyaluronan (hyaluronic acid) degradation and human lung fibroblast-derived hyaluronidase . The Journal of clinical investigation . 1992 ;90 (4 ):1492 –503 . Epub 1992/10/01. 10.1172/JCI116017 1401082 27 Moseley R , Waddington RJ , Embery G . Degradation of glycosaminoglycans by reactive oxygen species derived from stimulated polymorphonuclear leukocytes . Biochimica et biophysica acta . 1997 ;1362 (2–3 ):221 –31 . Epub 1998/04/16. .9540853 28 Agren UM , Tammi RH , Tammi MI . Reactive oxygen species contribute to epidermal hyaluronan catabolism in human skin organ culture . Free radical biology & medicine . 1997 ;23 (7 ):996 –1001 . Epub 1997/01/01. .9358242 29 Arnett FC , Edworthy SM , Bloch DA , McShane DJ , Fries JF , Cooper NS , et al The American Rheumatism Association 1987 revised criteria for the classification of rheumatoid arthritis . Arthritis and rheumatism . 1988 ;31 (3 ):315 –24 . Epub 1988/03/01. .3358796 30 Schmitz I , Ariyoshi W , Takahashi N , Knudson CB , Knudson W . Hyaluronan oligosaccharide treatment of chondrocytes stimulates expression of both HAS-2 and MMP-3, but by different signaling pathways . Osteoarthritis and cartilage / OARS, Osteoarthritis Research Society . 2010 ;18 (3 ):447 –54 . Epub 2009/10/31. 10.1016/j.joca.2009.10.007 19874928 31 Taylor KR , Trowbridge JM , Rudisill JA , Termeer CC , Simon JC , Gallo RL . Hyaluronan fragments stimulate endothelial recognition of injury through TLR4 . The Journal of biological chemistry . 2004 ;279 (17 ):17079 –84 . Epub 2004/02/07. 10.1074/jbc.M310859200 .14764599 32 Termeer C , Benedix F , Sleeman J , Fieber C , Voith U , Ahrens T , et al Oligosaccharides of Hyaluronan activate dendritic cells via toll-like receptor 4 . The Journal of experimental medicine . 2002 ;195 (1 ):99 –111 . Epub 2002/01/10. 11781369 33 Jiang D , Liang J , Noble PW . Hyaluronan in tissue injury and repair . Annual review of cell and developmental biology . 2007 ;23 :435 –61 . Epub 2007/05/18. 10.1146/annurev.cellbio.23.090506.123337 .17506690 34 Fieber C , Baumann P , Vallon R , Termeer C , Simon JC , Hofmann M , et al Hyaluronan-oligosaccharide-induced transcription of metalloproteases . Journal of cell science . 2004 ;117 (Pt 2 ):359 –67 . Epub 2003/12/06. 10.1242/jcs.00831 .14657275 35 Voelcker V , Gebhardt C , Averbeck M , Saalbach A , Wolf V , Weih F , et al Hyaluronan fragments induce cytokine and metalloprotease upregulation in human melanoma cells in part by signalling via TLR4 . Experimental dermatology . 2008 ;17 (2 ):100 –7 . Epub 2007/11/23. 10.1111/j.1600-0625.2007.00638.x .18031543 36 Campo GM , Avenoso A , Campo S , D'Ascola A , Nastasi G , Calatroni A . Small hyaluronan oligosaccharides induce inflammation by engaging both toll-like-4 and CD44 receptors in human chondrocytes . Biochemical pharmacology . 2010 ;80 (4 ):480 –90 . Epub 2010/05/04. 10.1016/j.bcp.2010.04.024 .20435021 37 Ohno S , Im HJ , Knudson CB , Knudson W . Hyaluronan oligosaccharide-induced activation of transcription factors in bovine articular chondrocytes . Arthritis and rheumatism . 2005 ;52 (3 ):800 –9 . Epub 2005/03/08. 10.1002/art.20937 15751080 38 Akira S , Uematsu S , Takeuchi O . Pathogen recognition and innate immunity . Cell . 2006 ;124 (4 ):783 –801 . Epub 2006/02/25. 10.1016/j.cell.2006.02.015 .16497588 39 Monaco C , Terrando N , Midwood KS . Toll-like receptor signaling: common pathways that drive cardiovascular disease and rheumatoid arthritis . Arthritis care & research . 2011 ;63 (4 ):500 –11 . Epub 2011/04/01. 10.1002/acr.20382 .21452263 40 Huber LC , Distler O , Tarner I , Gay RE , Gay S , Pap T . Synovial fibroblasts: key players in rheumatoid arthritis . Rheumatology . 2006 ;45 (6 ):669 –75 . Epub 2006/03/29. 10.1093/rheumatology/kel065 .16567358 41 Nakatani Y , Tanimoto K , Tanaka N , Tanne Y , Kamiya T , Kunimatsu R , et al Effects of hyaluronan oligosaccharide on the expression of MMP-1 in periodontal ligament cells . Archives of oral biology . 2009 ;54 (8 ):757 –63 . Epub 2009/06/16. 10.1016/j.archoralbio.2009.05.005 .19523610 42 Wang CT , Lin YT , Chiang BL , Lin YH , Hou SM . High molecular weight hyaluronic acid down-regulates the gene expression of osteoarthritis-associated cytokines and enzymes in fibroblast-like synoviocytes from patients with early osteoarthritis . Osteoarthritis and cartilage / OARS, Osteoarthritis Research Society . 2006 ;14 (12 ):1237 –47 . Epub 2006/06/30. 10.1016/j.joca.2006.05.009 .16806998 43 Shimizu M , Yasuda T , Nakagawa T , Yamashita E , Julovi SM , Hiramitsu T , et al Hyaluronan inhibits matrix metalloproteinase-1 production by rheumatoid synovial fibroblasts stimulated by proinflammatory cytokines . The Journal of rheumatology . 2003 ;30 (6 ):1164 –72 . Epub 2003/06/05. .12784385 44 Hiramitsu T , Yasuda T , Ito H , Shimizu M , Julovi SM , Kakinuma T , et al Intercellular adhesion molecule-1 mediates the inhibitory effects of hyaluronan on interleukin-1beta-induced matrix metalloproteinase production in rheumatoid synovial fibroblasts via down-regulation of NF-kappaB and p38 . Rheumatology . 2006 ;45 (7 ):824 –32 . Epub 2006/02/02. 10.1093/rheumatology/kel026 .16449361 45 Matsuno H , Yudoh K , Kondo M , Goto M , Kimura T . Biochemical effect of intra-articular injections of high molecular weight hyaluronate in rheumatoid arthritis patients . Inflammation research: official journal of the European Histamine Research Society [et al] . 1999 ;48 (3 ):154 –9 . Epub 1999/04/29. .10219658 46 Saito S , Momohara S , Taniguchi A , Yamanaka H . The intra-articular efficacy of hyaluronate injections in the treatment of rheumatoid arthritis . Modern rheumatology / the Japan Rheumatism Association . 2009 ;19 (6 ):643 –51 . Epub 2009/08/04. 10.1007/s10165-009-0207-8 .19649562

PMC005xxxxxx/PMC5001786.txt

==== Front Biochim Biophys ActaBiochim. Biophys. ActaBiochimica et Biophysica Acta0006-3002Elsevier Pub. Co S0005-2728(16)30589-810.1016/j.bbabio.2016.07.007ReviewMetabolite transport and associated sugar signalling systems underpinning source/sink interactions Griffiths Cara A. aPaul Matthew J. aFoyer Christine H. c.foyer@leeds.ac.ukb⁎a Plant Biology and Crop Science, Rothamsted Research, Harpenden, Hertfordshire AL5 2JQ, UKb Centre for Plant Sciences, School of Biology, Faculty of Biological Sciences, University of Leeds, Leeds LS2 9JT, UK⁎ Corresponding author. c.foyer@leeds.ac.uk1 10 2016 10 2016 1857 10 1715 1725 11 1 2016 6 6 2016 23 7 2016 © 2016 The Authors2016This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).Metabolite transport between organelles, cells and source and sink tissues not only enables pathway co-ordination but it also facilitates whole plant communication, particularly in the transmission of information concerning resource availability. Carbon assimilation is co-ordinated with nitrogen assimilation to ensure that the building blocks of biomass production, amino acids and carbon skeletons, are available at the required amounts and stoichiometry, with associated transport processes making certain that these essential resources are transported from their sites of synthesis to those of utilisation. Of the many possible posttranslational mechanisms that might participate in efficient co-ordination of metabolism and transport only reversible thiol-disulphide exchange mechanisms have been described in detail. Sucrose and trehalose metabolism are intertwined in the signalling hub that ensures appropriate resource allocation to drive growth and development under optimal and stress conditions, with trehalose-6-phosphate acting as an important signal for sucrose availability. The formidable suite of plant metabolite transporters provides enormous flexibility and adaptability in inter-pathway coordination and source-sink interactions. Focussing on the carbon metabolism network, we highlight the functions of different transporter families, and the important of thioredoxins in the metabolic dialogue between source and sink tissues. In addition, we address how these systems can be tailored for crop improvement. Graphical abstract Metabolite transport not only allows the flawless coordination of pathways in different organelles and between cells, but it is also balances energy provision and utilisation throughout the whole plant. Plant metabolite transport displays enormous plasticity and flexible regulation. Focussing particularly on carbon transporters, this review discusses transport functions in source and sinks organs, with sucrose and trehalose in source-sink communication that encompasses sophisticated perception and signalling systems.Image 1 Highlights • Metabolite transport facilitates inter-pathway regulation, cell to cell communication and co-ordination of source and sink processes. • Highlighting the carbon metabolism network, this review focuses on the role of transporters, carbon flow, and signalling.- The integration of sucrose and trehalose pathways as nodes in the regulation of plant growth and signalling is approached. Keywords Phloem loadingRedox regulationSource-sink interactionsSucrose transportersSugar signallingTrehalose ==== Body 1 Introduction Plant metabolism is driven by the energy-transducing reactions of the chloroplasts and mitochondria, which use ATP, reducing power {NAD(P)H)} and associated metabolites as the major currency of energy exchange. Plant cells synthesize all of the metabolic building blocks for growth, biomass production and defence such as sugars and carbohydrate polymers, lipids, amino acids and secondary metabolites such as alkaloids and terpenoids. Metabolite fluxes through parallel pathways often occur simultaneously in differing cellular compartments [1]. Moreover, the metabolic requirements of different developmental and defence processes change dynamically with time and according to prevailing environmental conditions, requiring overlapping layers of short and long-term regulation. Dynamic regulation of photosynthetic and respiratory metabolism involving extensive metabolite exchange provides tight but flexible delivery of the correct building blocks in appropriate amounts at the right time and place. While our understanding of the co-ordination of the pathways of primary carbon and nitrogen assimilation has greatly increased in recent decades, relatively little is known about regulation of the essential transport systems between compartments, cells and different plant organs [2], [3], [4]. Metabolites such as sugars and amino acids occupy central positions in the coordination of processes in different cellular compartments, facilitating multiple points of reciprocal control between pathways [5], [6], [7]. In addition, metabolites, such as sucrose and nitrate act as signals regulating gene expression to optimise pathway fluxes according to prevailing environmental conditions. In this review we discuss the importance of transporters in the metabolic coordination network within cells, between cells and between organs, with a particular focus on carbon metabolism and sugar transport and signalling, and the main pathways that interact with carbon metabolism to ensure appropriate provision of resources between source and sink organs. 2 Carbon/nitrogen interactions The efficient operation of carbon metabolism leading to carbon gain intrinsically depends on the successful uptake of other nutrients particularly nitrogen and phosphorus. Increasing focus is being placed on this interdependence because of the uncertainties that persist regarding how plant yields will be influenced by climate change and the increasing availability of carbon dioxide in the atmosphere. On one hand, studies in FACE (Free-Air CO2 Enrichment) systems have demonstrated that nitrogen use efficiency will increase as atmospheric carbon dioxide becomes more available [8]. Conversely, the concomitant inhibition of photorespiration will have an adverse effect on primary nitrogen assimilation because of limitations on redox cycling [9]. In addition, for future agriculture to be sustainable it is crucial that further yield gains are achieved with current and preferably low levels of soil fertilization [10]. Nitrogen use efficiency (NUE) is a highly complex trait involving N uptake efficiency (NUpE) and N assimilation efficiency (NUtE). NUpE is influenced by root architecture and the activities of large families of NO3− and NH4+ transporters [10]. Several families of nitrate transporters (NRT1, NRT2 and CLC) mediate the uptake and transport of nitrate in plants [10]. In general, NRT2 transporters have a high affinity for nitrate, while most of the NRT1 family have a low affinity for nitrate. Perhaps the best characterised example is NRT1.1, which is a dual affinity transporter. It is switched from low- to high-affinity transport forms by phosphorylation of Thr101 [10]. The post-translational modification of NRT1.1 enhances the affinity of the protein for nitrate, while high nitrate also acts as a transcriptional suppressor of NRT1.1expression. NRT1.1 is an important nitrate-sensing component that regulates lateral root development [11] by facilitating auxin transport [12]. This transporter also participates in the control of the expression of genes such as the high-affinity nitrate transporter, NRT2.1, whose expression is also regulated by nitrate availability [13], [14], [15]. Nitrate reduction in the cytosol is catalysed by NADH-dependent nitrate reductases (NR). While the activity of this enzyme is regulated in response to environmental and metabolic triggers as well as protein phosphorylation, the rate of nitrate assimilation can be limited by the availability of NADH [2], [3], [16], [17]. Metabolite transport between the chloroplasts and cytosol is important in boosting the cytosolic NADH pool, involving dicarboxylates transport and shuttle systems for malate and oxaloacetate. The 2-oxoglutarate/malate transporter, AtpOMT1 plays an important role in this process, functioning both as an oxaloacetate/malate transporter in the malate valve pathway and as a 2-oxoglutarate/malate transporter mediating the transfer of carbon skeletons [18], [19]. Nitrite generated by the action of NR is transported into the chloroplasts where it is reduced by nitrite reductase (NiR) to ammonium (NH4+), which is assimilated into amino acids by the glutamine synthetase/glutamine-2-oxoglutarate aminotransferase (GS-GOGAT) pathway. Thereafter, a raft of aminotransferases and other enzymes catalyse transfer of the amino group to form other amino acids. Provision of the 2-oxoglutarate required for ammonia assimilation requires partial operation of the TCA cycle in the mitochondria [3]. 3 Pathway co-ordination of by reversible thiol-disulphide exchange The extensive cycling of reducing equivalents facilitated by metabolite transport [9], [20], [21] also facilitates pathway co-ordinates through influences on post-transcriptional protein modifications (PTM) that provide dynamic and reversible protein processing to modulate enzyme activity, binding properties and function. Many types of PTM (over 450) have been identified to date. Many proteins involved in carbon and nitrogen metabolism are subject to PTMs such as protein acetylation, succinylation, malonylation, butyrylation, and propionylation. For example, the large subunit of ribulose-1, 5-bisphosphate carboxylase (RuBisCO) is extensively succinylated and acetylated. Deacetylation of the RuBisCO protein has been shown to activate the enzyme [22]. However, in most cases the functional significance of these PTMs has yet to be resolved. In contrast, the functions of protein phosphorylation and thiol-disulphide exchange processing (Fig. 1) have been extensively characterised. For example, the light and thiol-dependent activation of photosynthetic CO2 fixation pathway enzymes requires thioredoxins (TRX), which are small proteins with disulphide reductase activities [23]. TRX reductases in the stroma reduce TRXs using either reduced ferredoxin or NADPH produced by the photosynthetic electron transport chain [24]. Conversely, TCA cycle enzymes such as succinate dehydrogenase and fumarase are reductively inactivated by TRX [25], [26], [27]. In this way, the TRX systems in the plastids, mitochondria and cytosol link photosynthetic and respiratory electron transport activities to functional changes in enzyme activities [23], [28]. The photosynthetic electron transport chain provides the reductive “push” that keeps the stromal TRXs reduced in the light, ensuring that the activation states of the thiol modulated enzymes involved in CO2 fixation are matched to rate of production of reduced ferredoxin and NADPH [24], [29], [30], [31]. When this push is removed in the dark, the TRXs revert to their oxidized forms, which in turn allow oxidative inactivation of the thiol-modulated enzymes. In addition, chloroplast TRXs also regulate malate and oxaloacetate transport [32] and starch metabolism through effects on adenosine diphosphate (ADP)-glucose pyrophosphorylase (AGPase). The small subunit of this key enzyme of starch biosynthesis is regulated by redox-dependent dimerization in response to sugar availability [33], [34], [35]. Appropriate resource allocation between different plant organs might also be achieved by post-translational modification of sucrose transporters. Sucrose export from leaves occurs both in the light and dark to ensure the continuous and stable provision of adequate carbon resources to drive plant growth and development. The transport of sucrose is mediated by membrane-localised sucrose transporters, whose properties will be discussed later in detail. The activities of these transporters are regulated by changes in cellular redox status and by protein-protein interactions [36]. For example, SUT1 interacts with a small cysteine-rich cell wall protein in potato called SN1 [37]. The SN1 protein belongs to the Snakin/gibberellic acid stimulated (GAS) family in Arabidopsis [38], [39]. Silencing of SN1 leads to perturbations in cellular redox metabolism, particularly antioxidant activities [39]. SUT1 is able to form dimers with a protein disulphide isomerase in a redox-dependent interaction and is hence able to interact with other proteins that are involved in metabolism or secretion [40]. However, no increases in the transport activity of StSUT1 have been shown to be dependent on this dimerization [41]. 4 Transport of sugars between the chloroplasts and cytosol Triose phosphate/phosphate, glucose, maltose transporters are important mediators of carbon transfer between the plastids and cytosol [7]. Carbon assimilated during photosynthesis is either transported out of the chloroplast or used for starch biosynthesis in the stroma. Carbon is exported from the chloroplasts in the light as triose phosphate or 3-phosphoglyceric acid. This occurs across the triose-phosphate/phosphate translocator (TPT) in strict stoichiometric exchange for inorganic phosphate (Pi). TPT belongs to the plastidial phosphate translocator family, which has three other members: the pentose phosphate/phosphate translocator (XPT), phosphoenolpyruvate (PEP)/phosphate translocator (PPT) and the glucose 6-phosphate/phosphate translocator (GPT). TPT is relatively abundant in the chloroplast inner membrane, constituting 10–12% of the protein content [42]. Triose-phosphates transported by TPT are utilised for sucrose synthesis in the cytosol or for the generation of organic acids through the anaplerotic pathway. Inorganic phosphate produced during sucrose synthesis is transported back into the chloroplast via the TPT to be used in ATP synthesis [43], [44], [45]. Therefore, the re-cycling of phosphate maintains photosynthetic electron transport and the pentose phosphate pathways [46]. While transgenic potato plants that have reduced TPT levels or where TPT is knocked out completely show no phenotype, the chloroplasts have a reduced capacity to import phosphate by up to 30%, together with a 40–60% reduction in maximal photosynthesis [47]. The TPT is therefore central to the regulation of carbon partitioning between starch and sucrose [48], [49]. Transgenic tobacco plants that over-express TPT did not show a marked phenotype or any marked effects on amino acid production [50]. However, they incorporated more CO2 into sucrose and had higher leaf starch/sucrose ratios than the wildtype [50]. Transgenic Arabidopsis lines over-expressing TPT and also fructose-1, 6-bisphosphatase had higher photosynthetic carbon assimilation rates. In addition, these lines also accumulated more sucrose, glucose and fructose than controls and showed enhanced growth compared to the wild type [51]. Another type of phosphate transporter (PPT) facilitates the transport of three-carbon compounds containing a phosphate group at the C2 position, such as phosphoenolpyruvate (PEP) and 2-phosphoglyceratefor phosphate across the inner plastid envelope membrane in heterotrophic tissues [52]. PPT transporters are essential for the transport of PEP into plastids, which with the exception of lipid-storage tissues, cannot produce PEP [53]. Of the two genes encoding PPTs in Arabidopsis, AtPPT2 is expressed in leaves and AtPPT1 transcripts are localised only in the vasculature. Analysis of Atppt1 mutants demonstrated that this transporter is essential for plant development and metabolism [54]. In addition, AtPPT1 may be involved in signal transduction through its association with phenylpropanoid metabolism [54]. PPT is responsible for the supply of PEP to the shikimate pathway in the chloroplast stroma [55], [56]. The cue1 mutant, which is defective in PPT and shows altered chloroplast development, has dark-green veins and light-green interveinal regions [57]. While this phenotype can be explained at least in part by a restriction in PEP supply to the shikimate pathway [57], the cue1 mutant phenotype could not be rescued by the over-expression of a cauliflower PPT [58]. Grafting experiments involving cue1 mutants suggested that leaf and root PPTs may fulfil different roles, with the leaf chloroplast PPT acting as a PEP importer while the root plastid PPT acts as an overflow valve [59]. While PPTs are present in both photosynthetic and heterotrophic tissues, another group of phosphate transporters, the glucose-6-phosphate/phosphate transporters (GPT), are found only in heterotrophic tissues [42]. The GPT group mediates the transport of glucose-6-phosphate, triose phosphates and phosphates into plastids [60]. The Arabidopsis genome contains two GPT genes, AtGPT1 and AtGPT2 [54]. GPT1 is expressed in a similar manner to genes involved in starch metabolism [61]. The Arabidopsis Atgpt1 mutants have severe defects in pollen development [62]. GPT2 plays a crucial role in the partitioning of glucose 6-phosphate between the plastids and cytosol during the transition from heterotrophic to autotrophic growth [63]. Over expression of GPT2 led to an inhibition of carbon metabolism [64] and there was an early onset of senescence [65] although photosynthesis showed similar responses to increasing light intensities. The xyulose-5-phosphate transporters (XPT) catalyse the exchange of xyulose-5-phosphate and to a lesser extent ribulose-5-phosphate for Pi [53]. These transporters link the pentose phosphate pathways in the plastids and cytosol, transporting metabolites for use in the Benson-Calvin cycle or other metabolic pathways [54], [66]. 5 Transporters for the storage of carbohydrates in the vacuole Just as triose phosphates are the main end-products of photosynthesis that are exported by chloroplasts, sucrose is the major export carbohydrate of leaves [67] and also an important signalling metabolite conveying information on resource availability throughout the plant [68], [69]. Sucrose synthesised in the cytosol is transported from the leaves to the phloem. Starch is synthesised and stored in chloroplasts during the light period and it is broken down to glucose and maltose at night and shipped to the cytosol for the synthesis of sucrose [70]. The dynamic synthesis and breakdown of starch, together with sucrose synthesis and export, ensures that the non-photosynthetic tissues receive a constant supply of carbohydrate to drive metabolism and growth in the light and in the dark. The key enzymes of the sucrose biosynthesis pathway, sucrose phosphate synthase (SPS) and sucrose phosphate phosphatase (SPP) appear to form a complex in the cytosol, allowing interactions between the proteins that not only influence the soluble carbohydrate pools but also modifies carbon partitioning to starch [71]. It is interesting to note that SPS has an N-terminal glycosyltransferase and a C-terminal phosphatase domain. This two domain structure bears a remarkable resemblance to that of the enzymes that catalyse the synthesis of another disaccharide, trehalose. Like sucrose, trehalose synthesis involves a two-step process, catalysed by trehalose-6-phosphate synthase (TPS) and trehalose 6-phosphate phosphatase (TPP). Trehalose-6-phosphate (T6P) is formed by TPS and is then dephosphorylated to trehalose by TPP, a reaction sequence that is relevant to the ways in which plants sense sucrose, as discussed later. In addition to carbohydrate storage in leaves as starch, sucrose, hexoses and raffinose can also be stored in the vacuoles [71], which may aid the stabilisation of the organelles during adverse conditions, as well as into the guard cells in a light-dependent manner to influence stomatal aperture [72]. Sugars stored in the vacuole during the day are generally released and exported during the night [73]. Like excessive starch accumulation in chloroplasts, high leaf sucrose concentrations can inhibit photosynthesis [74]. In the case of sucrose, inhibition of photosynthesis is linked to sugar-mediated repression of the expression of photosynthetic genes [75]. Although the sucrose stored in leaf vacuoles represents only are small amount of the sucrose transported to the phloem, vacuolar sucrose levels can exert a strong effect on photosynthesis and stress tolerance. Moreover, vacuolar sucrose storage is particularly important in supporting the heterotrophic growth of reproductive tissues and developing embryos. The transport of sucrose within leaf mesophyll cells is driven by the concentration gradient between in the cytosol and vacuole [76]. Members of sucrose transporter (SUT) family transport sucrose across the tonoplast membranes. The tonoplast sucrose transporters in barley (HvSUT2) and Arabidopsis (AtSUT4) are able to transport sucrose when expressed in yeast [77], [78]. However, HvSUT2 may function to support the starch accumulation required for the embryo during grain development [77]. The Arabidopsis tonoplast monosaccharide transporter (TMT) family consists of three members, which are expressed in a tissue- and cell-specific manner. TMT1 is expressed in pollen, flowers and young developing tissues and TMT2 is expressed in young roots, floral tissues and mature leaves. In contrast, TMT3 transcripts are low in abundance, and are found only in seedlings, mature leaves and stamens. The expression of TMT1 and TMT2 is increased in response to drought and cold treatments, suggesting a role for these transporters in the response to environmental stresses [79]. A homologue of TMT was also found in the tonoplast fraction of barley mesophyll cells [74], suggesting that TMT may also have a role in sugar transport even under optimal growth conditions. More recently, the sugar beet transporter BvTST2.1 was found to contribute to vacuolar sucrose uptake. This transporter operates as a proton antiporter, and has a high similarity to Arabidopsis TMT family amino acid sequences. This prompted the name change of the TMT family of transporters, to the TST family (tonoplast sucrose transporters) [80]. Glucose transporter 1 (AtVGT1) is localised on the tonoplast membrane in Arabidopsis. The Atvgt1 mutants showed delayed flowering, together with a reduced capacity for viable seed production [81]. These data demonstrate that the sugars stored in the vacuole and their transport are essential for plant metabolism and related source/sink interactions. 6 Phloem loading The distribution of resources between plant organs relies on the plant vasculature system comprising of xylem and phloem. The xylem is responsible for the transport of water and minerals collected from roots up to the shoots. The phloem is responsible for the transport of nitrogen- and carbon-containing compounds from source tissues such as leaves, to sink tissues. The phloem tissue consists of two cell types, sieve element (SE), responsible for nutrient conduction, and companion cell (CC), which provides metabolic support for the SE [82]. Together with the xylem, the phloem is an essential vascular network, providing carbon and nutrients to areas of the plants when needed. The phloem is an important trafficking path in source/sink interactions linking key processes required for sugar signalling that not only controls the flow of sugars to developing organs, but also influences gene expression and hormone signalling throughout the plant [83]. Very few sugars that are synthesised in plants are transported long-distance through the phloem. Sucrose is the preferred transport form of carbon in most species. Thus, sucrose is the most abundant sugar found in the phloem in most plants [84]. The majority of sucrose produced in mesophyll cells is released into the apoplast and loaded into the sieve element (SE)/companion cell (CC) complexes of the phloem [82]. The SS/CC complexes are stacked longitudinally to form sieve tubes, which are able to transport sugars throughout the plant. Furthermore, there are at least three different but overlapping phloem types. These are the collection phloem, release phloem and transport phloem, the latter comprising the majority of phloem tissue. The collection phloem is located in smaller veins of the source leaves and is responsible for sucrose export. Release phloem is found in sink tissues (such as fruit, seeds and tubers) and is responsible for sucrose unloading. The transport phloem moves sucrose (and other solutes) around the plant, relying on osmotic pressures between source and sink to aid direction [83]. To date, three mechanisms for phloem loading have been described. In most species, phloem loading is apoplastic and facilitated by the membrane transporters described above, and transport is driven by the proton motive force [84], [85], [86], [87]. However, symplastic transport predominates, requiring high densities of plasmodesmata to facilitate sucrose movement to the phloem. Loading is thereafter driven by the sucrose concentration gradient between the mesophyll and phloem. This type of symplastic loading requires the maintenance of high sucrose concentrations in the mesophyll to maintain the downhill sucrose gradient [88], [89]. However, in some species, symplastic phloem loading in leaves is not directly apparent, despite high concentrations of sucrose in the mesophyll, no evidence of sucrose accumulation against a concentration gradient was found [90]. More recently a revised model of phloem loading has been described in rice whereby sucrose diffuses passively from the mesophyll through the plasmodesmata to the minor veins. Mesophyll sucrose concentrations are higher than the minor veins maintained by a tonoplast SUT transporter, which therefore acts a valve able to regulate sucrose flux into the phloem [88]. In some symplastic loaders, sucrose diffuses into the CCs of the minor veins from the mesophyll, where it is converted to stachyose and raffinose. These oligosaccharides are unable to diffuse back into the mesophyll and accumulate in the phloem, as a result of polymer-trapping [91]. Sucrose transporters play an integral role in apoplastic phloem loading and unloading, as well as in the exchange of sucrose between the plant and beneficial and/or parasitic symbiotic organisms [92]. Crucially, sucrose transporters play a key role in signal transduction between source and sink tissues [93]. Sucrose transporters belong to the major facilitator family, comprising 9 sucrose transporter genes (SUCs or SUTs) in Arabidopsis. Most SUTS are sucrose/H + symporters. However, other sucrose transporters are classified as sucrose facilitators (SUFs) because they catalyse a pH-independent and energy-dependent bi-directional transport of sucrose [94]. Phylogenetic analysis has been used to classify sucrose transporters into five clades: SUT1 (dicot-specific), SUT2 (found in monocots and dicots), SUT3 (monocot-specific), SUT4 (monocots and dicots) and SUT5 (monocot-specific) [84]. SUT1 clade and SUT3 clade members are typically expressed in the SE or CC or in both cell types. The SUT2 and SUT4 clades, which have a low affinity for sucrose, are generally expressed only in the plasma membrane of SEs. However, some SUT4 members have also been observed in the chloroplast and vacuole [84]. While the SUT5 clade, which was recently separated from the SUT3 clade remains poorly characterised, the gene encoding the rice SUT5 protein was reported to be highly expressed in sink leaves [88], [95]. In apoplastic loading/unloading species, sucrose is transported from the apoplast to the SE/CC by the SUT1/SUT3 (dicot/monocot) H +/sucrose co-transporter [96]. Null mutants of maize sut1 had stunted growth, delayed flowering, and early senescence linked to a lower capacity for sucrose export [97]. Antisense tobacco, tomato and potato with low SUT1 showed similar phenotypes [98], [99], [100]. However, while decreased SUT1 expression in rice resulted in lower grain-filling capacity [101], [102], loss of the transporter did not lead to leaf carbohydrate accumulation or stunted phenotype [103]. In general, loss of SUT1 function results in reduced sucrose transport while increased SUT1 expression results in increased sucrose transport [84]. Taken together, these studies demonstrate the central role of SUT1 in apoplastic loading in autotrophic/source tissues, and unloading in heterotrophic/sink tissues. While SUT1 is a key transporter in phloem loading and unloading, the other SUT proteins are important in cell-type specific sucrose transport particularly during reproductive development. For example, members of the SUT2 family are expressed in the phloem SEs and also in pollen. Loss of SUT2 function in tomato decreased pollen viability, through reduced sucrose uptake [99]. The rice SUT3 is also highly expressed in developing pollen [104]. SUT4 members are localised on the plasma membrane and the tonoplast membranes [105]. The rice SUT4 is expressed during seed germination, as well as in pollen and in anthers at the post heading stage. Arabidopsis SUT5, which is expressed specifically in the endosperm, is considered to be a nutrient carrier of the filial tissues during early seed development [106]. It is important to note that although most sucrose transporters only have a high affinity for sucrose, some sucrose transporters, such as AtSUC2, can also transport other glycosides [107]. Although sucrose is the preferred plant transport form of carbon source, it is likely that sucrose derivatives are transported for use in other metabolic processes, or perhaps are used as signal molecules. While the SUT super-family are major transporters involved in the translocation of sucrose between the apoplast and phloem, other transporters responsible for the movement of sucrose also fulfil important roles. For example, the SWEET transporters have been implicated in the transport of sucrose from the phloem parenchyma to the apoplast [108]. There are 17 SWEET family members in Arabidopsis that fall into four clades. In rice, there are 21 SWEET family members falling into the same four clades. The function of clade II and clade IV members remains to be identified, however members of clade I have been shown to mediate glucose import/export [93] and members of clade III preferentially transport sucrose across the plasma membrane [108]. There is growing evidence that SWEETs are bidirectional, pH-independent, low-affinity sucrose transporters, which operate through a uniporter mechanism of transport [109]. The SWEET transporters are localised in the phloem parenchyma. GFP-tagged SWEET11 from Arabidopsis (AtSWEET11) was localised to the plasma membrane [108]. However, localisation in the phloem parenchyma remains to be confirmed. AtSWEET11 and AtSWEET12 share 88% identity at the amino acid level. There appears to be some functional redundancy in the SWEET protein family as the Atsweet11 and Atsweet12 mutants had no marked phenotype. However, Atsweet11:12 double mutants had slower growth and showed carbohydrate accumulation in the leaves [108]. The phenotypes of the Atsweet11:12 mutants were not as severe the sut1 mutants, indicating that other SWEET proteins may compensate for the loss of SWEET11 and SWEET12 activity [109]. Studies on SWEETs and SUT1 have shown that apoplastic phloem loading occurs in two steps [107], [109]. Firstly, sucrose is exported from the phloem parenchyma to the apoplast by SWEETs. Secondly, sucrose is transported into the cells of the SE/CC complex from the apoplast by SUT1 (Fig. 2). 7 Starch accumulation Excess carbohydrates are often stored as starch in source and sink tissues. Sucrose transported from leaves is used for the synthesis of starch at sites where carbohydrate storage is required, for example in seeds [110]. Starch accumulation in chloroplasts predominates during the day. While starch synthesis and degradation can occur simultaneously, starch degradation and remobilisation occur largely at night [110]. The Arabidopsis starch excess (sex) mutant was originally thought to have reduced expression of a chloroplast hexose transporter [111]. However, it was later found to have reduced expression of the starch granule R1 protein, which controls the phosphate content of starch [112]. Increased starch accumulation can be affected by maltose metabolism and transport. Maltose is accumulated in the chloroplasts during starch breakdown [113]. Maltose excess (mex) mutants, which are defective in the maltose transporter MEX1, were unable to convert starch into sucrose during the night, leading to increased starch accumulation in leaves [114]. Similarly, apple mex1 mutants were unable to degrade starch during the night [115]. MEX1 is expressed in mature apple leaves and also in sink tissues. The function of the MEX1 transporter in sink tissues is unknown but it is likely to be required to support seedling growth, ensuring rapid use of stored carbohydrates to drive growth [115]. In addition to transporters, other proteins have been implicated in starch accumulation. Tie-Dyed1 (Tdy1), which has only been described in grasses, is a transmembrane protein that promotes sucrose loading into the phloem [116]. Maize tdy1 mutants are stunted with chlorotic leaves because of excessive leaf starch accumulation [86], possibly at the site of leaf veins [117], and produce oil droplets in the companion cells [118]. However, these mutants are not defective in phloem unloading, suggesting that Tdy1 functions in carbon partitioning through the promotion of phloem loading [116]. Interestingly, the maize sucrose export defective 1 (sxd1) mutant displays a similar phenotype to tdy1 [117]. In addition, it was found that TDY1 and TDY2, a proposed callose synthase, may interact to promote symplastic transport in the phloem [118]. Since these transporters are specific to grasses, some phloem loading processes may be unique to monocots [116]. Further evidence of unique phloem loading process in monocots was described earlier in rice [88]. There may be differences in the types of sugar transporters used to support grain development in different species. For example, HvSUT1 has been implicated in seed starch accumulation. The expression of HvSUT1 was directly associated with sucrose accumulation in the caryopses, where sucrose influx correlated with increased starch accumulation [77]. Further evidence for a role of SUT1 in grain filling has been obtained in rice [119], wheat [120] and barley [77]. These SUT1 transporters maintain the sugar supply to the developing grain by transporting sugar from the apoplast to the phloem [120]. In contrast, AtSUT5 appears to function in endosperm-specific sucrose transport in Arabidopsis [106]. The SWEET transporters are also important in grain filling in Arabidopsis [109]. The sweet11;12;15 triple mutants showed severe defects in seed development producing lower weight wrinkled seeds, with reduced starch and lipid [109]. 8 Trehalose metabolism and sugar signalling Processes that consume sucrose are sensitive to stress-induced inhibition. For example, sucrose accumulation is observed in plants exposed to low temperatures [121], drought and salt stress [122], [123] and nutrient deficiency [124]. Sucrose is sensed by the plant directly, through the generation of hexoses and through sugar signals such as T6P (trehalose-6-phosphate) which relay the sugar status of the plant into mechanisms that enable adaptation to different environmental conditions (Fig. 2). Hexoses are produced only where sucrose is being metabolised, whereas T6P can be produced as a sugar signal wherever sucrose and the trehalose pathway are present. T6P is produced as an intermediate compound in the trehalose biosynthesis pathway. Briefly, UDP-glucose and glucose 6-phosphate are used to produce T6P catalysed by trehalose phosphate synthase (TPS), T6P is then converted into trehalose by trehalose phosphate phosphatase (TPP; Fig. 3). The flux of carbon into trehalose is four orders of magnitude less than into sucrose. Therefore, T6P synthesis will not cause depletion of UDPG and G6P pools. Whether the levels of UDPG and G6P are important regulators of T6P synthesis is unknown but there is correlation between the abundance of these metabolites and that of T6P. However, T6P is synthesised in actively growing tissues where sucrose is metabolised to yield these substrates [124]. Accumulating evidence shows that T6P levels are most related closely to sucrose pool size [124], [125]. T6P is a universal signal of sucrose concentration in plants [121], [124], [126], [127], [128]. The abundance of T6P does not respond to the levels of sugars such as glucose [121]. Thus, T6P is considered to be a specific signal for sucrose availability. T6P levels are controlled largely by the activity of TPS1. Arabidopsis mutants lacking the trehalose phosphate synthase gene, tps1 are embryo lethal [129]. T6P inhibits Sucrose non-Fermenting Related Kinase 1 (SnRK1). Inhibition of this major signalling component results in metabolic reprogramming at sucrose levels above 3 μmol g− 1 FW in Arabidopsis seedlings [128] i.e. about 3 mM sucrose on a whole tissue basis. This sucrose levels results in the accumulation of T6P to about 1 μM on a whole tissue basis. Hence, physiologically-relevant T6P concentrations are between 1 and 10 μM. SnRK1 is inhibited over this concentration range, for example 50% inhibition is produced at 5.4 μM T6P [128]. Relatively small changes in T6P concentrations within this range may therefore give rise to large changes in SnRK1 activity [130] resulting in the metabolic reprogramming of hundreds of genes involved in growth and defence [131], [132]. In the absence of T6P, SnRK1 regulates the expression of a different subset of genes that are involved in catabolism rather than anabolic processes. This regulation is important for stopping growth when sucrose is in short supply, thus preventing starvation and death. Therefore, the combination of T6P and SnRK1 activity is integral in overall plant growth and development (Fig. 4). Glucose 1-phosphate (G1P) and glucose 6-phosphate (G6P) also inhibit SnRK1, but less potently than T6P (at levels of 480 μM, and > 1 mM respectively). However, when these inhibitors are combined with T6P, a synergistic effect is observed with G1P and an additive with G6P [128]. These effects provide considerable flexibility in the regulation of SnRK1. Strong inhibition of SnRK1 may only occur under sucrose-replete conditions, as proposed by Lunn et al. [125]. However, many environment stresses cause sucrose accumulation. Correlations between T6P, sucrose levels and the expression of SnRK1 marker genes have been observed under different growth conditions, such as cold, nitrogen deficiency and following sucrose feeding [128]. T6P promotes growth only when other conditions are not limiting. Hence, T6P functions as part of an integrated network involving sucrose and hormones that regulate growth [133]. The activation of gene expression in response to sucrose accumulation following exposure to cold stress prepares the plants for rapid growth “in anticipation” of return of warmth [128]. This may be an important adaptive response to unseasonal cold. Large differences in synonymous and non-synonymous TPS and TPP substitutions have been observed in Arabidopsis [134]. These genes have strong cell-specific expression profiles and are induced differentially by environment and also by sugars [133], [134], [135]. The regulated expression of TPSs and TPPs is part of the plant response to fluctuations in sucrose supply. The regulation expression of different gene family members is driven by the requirement to optimise T6P in different cell types for certain environments. 9 Applications of sugar signalling in agriculture Manipulations of T6P levels have provided a new paradigm for crop improvement [122]. For example, a rice TPP linked to a MADS6 promoter was recently expressed in maize. MADS6 is expressed during the flowering period and plays an essential role in endosperm nutrient accumulation in ear nodes, ear vasculature and spikelet tissues [136], [137]. Drought at flowering can have a large effect on crop yields. Hence, manipulations have been sought that maintain the flow of sucrose to developing female reproductive tissues during drought [138], [139], [140]. MADS6-TPP expression decreased T6P levels in female reproductive tissues allowing increased sucrose availability in spikelets leading to improved harvest index. Yields were increased significantly (by up to 123%) by MADS6-TPP expression in rigorous field trials over a number of years, which involved multiple sites with a range of environments including soils with water deficits. The largest yield increases were observed under severe drought, however yield improvements were also seen under non-stressed conditions [122] (Fig. 5). T6P may regulate sugar levels as part of a homeostatic mechanism that ensures that sucrose does not accumulate to excessive levels or falls to very low levels, in a mechanism that is analogous to the control of blood glucose levels in animals by glucagon and insulin [125]. T6P was shown to perturb this homeostatic mechanism and to alter sucrose levels in the MADS6-TPP study [122]. Recent studies have confirmed the central role of T6P in managing whole plant carbon budgets and stress responses [141]. For example, a TPP gene was shown to be essential for anaerobic germination under flooding. In this case metabolism of T6P by TPP was perceived as a starvation signal, which enhanced starch mobilisation to drive the growth of the germinating embryo and elongating coleoptiles. Anaerobic germination tolerance enables uniform germination and seedling establishment under submergence. Hence, rice can be directly seeded rather than the current labour-intensive planting methods. It may be possible to harness the ability of T6P to function as both a starvation and satiety signal by increasing T6P. Decreasing T6P could be advantageous at other times. Driving down T6P levels may be effective in cells involved in sucrose transport as already demonstrated in maize and in germinating rice seeds [122], [141]. In other cell types, for example those that are actively converting sucrose to starch or oils, an increase in T6P levels may be effective in ensuring optimal gene expression for biosynthetic processes. Since large changes in yield can be achieved through one transgene which results in small changes in the target protein and in T6P levels, tailoring T6P metabolism could therefore make an important contribution to the improvement global food security. SnRK1 may not be the only target of T6P. In mature leaves T6P does not inhibit SnRK1 because of different composition of SnRK1 complexes between heterotrophic and autotrophic tissues [128] and yet T6P regulates starch metabolism in rosette leaves through redox regulation of AGPase [140] and significantly through starch breakdown [141]. How T6P mediates these effects is not known. Similarly, T6P is necessary for leaf senescence [142]. Remarkably this may be due to the regulation imparted by T6P early in leaf development possibly through SnRK1. T6P is also part of the network that regulates flowering, providing information to plants concerning sucrose availability for flower development [143]. 10 Concluding remarks Plant metabolite transporters are crucial to inter-pathway regulation, cell to cell communication, and source-sink interactions. The sophisticated vasculature of plants is rich in transporters, which regulate the partitioning of resources and signals between source and sink organs. Carbon metabolism and transport is at the heart of whole plant communication and underpins key agricultural traits such as flowering, productivity and stress tolerance [122], [144]. Sugar signalling even exerts a key influence over processes such as apical dominance [145], which was once thought to be the exclusive domain of phytohormone regulation. The cell autonomous expression of trehalose pathway genes allows the fine tuning of specific responses to sucrose availability. T6P not only fulfils a major role in signalling cellular carbon status, but it is also a determinant of sink strength. It is therefore an attractive target for increasing yield and yield resilience. T6P exerts its effects through SnRK1 [96], which like sucrose is a conserved central regulator of plant metabolism. Similarly to sucrose, trehalose can be transported throughout the vascular system [135], however relatively little is known about plant trehalose and T6P transporters. At an intracellular level T6P is mainly localised in the cytosol, however is also present in the chloroplasts [124]. Again, little is known about T6P transport between cellular compartments. An important consideration when aiming to improve crop yields is whether the yield itself is source or sink limited, which appears to be species-dependent. In potatoes for example, source capacity limits yield by up to 80% under glasshouse conditions [146]. In contrast, the yield of wheat seeds is largely dominated by sink activity [147]. Yield improvements can only come from a much deeper understanding of the equilibrium between source and sink tissues that intrinsically involves assimilate transport considerations at a whole plant level [148]. Future increases in crop yields are likely to come from improvements in transporter functions in sources and sinks. Manipulation of sucrose transporters, for example SUT1 and SWEETs, may have a dramatic effect on sucrose remobilisation and source/sink relationships underpinning plant growth and development. This may result in greater understanding of the source-sink relationship, and in-turn facilitate the development of sustainable, high-yielding crops. Conflict of interest There is no conflict of interest. Transparency document Transparency document. Acknowledgements CHF thanks BBSRC (UK; BB/M009130/1) for financial support. Rothamsted Research receives strategic funding from the Biotechnological and Biological Sciences Research Council of the UK. The Transparency document associated with this article can be found, in the online version. Fig. 1 The central role of thiol-disulphide exchange regulation via thioredoxins in plant biology. Fig. 1Fig. 2 Apoplastic loading of sucrose from the mesophyll to the phloem in leaf tissue. Sucrose produced during photosynthesis is stored in the vacuole and transported into the mesophyll by SUT2/4. Sucrose moves to the phloem parenchyma through plasmodesmata where it is transported to the apoplast by SWEETS. SUT1/3 transport sucrose from the apoplast to the companion cell of the phloem and from the companion cell to the sieve element by plasmodesmata, SUT2/4 and SUT1/3. Arrows represent direction of transport. Fig. 2Fig. 3 Sugar signalling via trehalose 6-phosphate (T6P)/SnRK1 interactions. The sucrose status of plant tissue is relayed through T6P to SnRK1. The activity of SnRK1 which is inhibited by T6P determines gene expression for starvation or satiety responses and maintains sucrose homeostasis. There is strong cell and developmental specificity of expression of genes regulating T6P content. Cell specific changes in T6P may enable modification of plant process and particularly productivity and resilience of crops. Fig. 3Fig. 4 Source-sick and T6P-related processes throughout the life cycle. Changes in T6P accumulation, and source-sink relationships affect cellular processes in seed germination, vegetative and reproductive plant growth. Source tissues generally have a low T6P content, and sink tissues a high T6P content. Arrows indicate direction of sucrose transport. Fig. 4Fig. 5 OsMADS6-TPP maize. Expression of TPP in female reproductive tissues of maize leads to increased yield in drought-stressed conditions. TPP is expressed in maize ear vasculature (a). In wild type plants, drought stress causes yield loss (b), in OsMADS6-TPP maize, yield loss due to drought-stress is decreased, resulting in higher yield in comparison to wild-type during drought-stress (c). Fig. 5 ==== Refs References 1 Heinig U. Gutensohn M. Dudareva N. Aharoni A. The challenges of cellular compartmentalization in plant metabolic engineering Curr. Opin. Biotechnol 24 2013 239 246 23246154 2 Foyer C.H. Noctor G. Photosynthetic nitrogen assimilation: inter-pathway control and signaling Foyer C.H. Noctor G. Photosynthetic Nitrogen Assimilation and Associated Carbon and Respiratory Metabolism 2002 Springer Netherlands Dordrecht 1 22 3 Foyer C.H. Noctor G. Hodges M. Respiration and nitrogen assimilation: targeting mitochondria-associated metabolism as a means to enhance nitrogen use efficiency J. Exp. Bot 62 2011 1467 1482 21282329 4 Noctor G. Novitskaya L. Lea P.J. Foyer C.H. Co-ordination of leaf minor amino acid contents in crop species: significance and interpretation J. Exp. Bot 53 2002 939 945 11912236 5 Lunn J.E. Compartmentation in plant metabolism J. Exp. Bot 58 2007 35 47 17030538 6 Forde B.G. Lea P.J. Glutamate in plants: metabolism, regulation, and signalling J. Exp. Bot 58 2007 2339 2358 17578865 7 Flügge U.-I. Häusler R.E. Ludewig F. Gierth M. The role of transporters in supplying energy to plant plastids J. Exp. Bot 62 2011 2381 2392 21511915 8 Leakey A.D.B. Ainsworth E.A. Bernacchi C.J. Rogers A. Long S.P. Ort D.R. Elevated CO2 effects on plant carbon, nitrogen, and water relations: six important lessons from FACE J. Exp. Bot 60 2009 2859 2876 19401412 9 Foyer C.H. Bloom A.J. Queval G. Noctor G. Photorespiratory metabolism: genes, mutants, energetics, and redox signaling Annu. Rev. Plant Biol 60 2009 455 484 19575589 10 Xu G. Fan X. Miller A.J. Plant nitrogen assimilation and use efficiency Annu. Rev. Plant Biol 63 2012 153 182 22224450 11 Remans T. Nacry P. Pervent M. Filleur S. Diatloff E. Mounier E. Tillard P. Forde B.G. Gojon A. The Arabidopsis NRT1.1 transporter participates in the signaling pathway triggering root colonization of nitrate-rich patches Proc. Natl. Acad. Sci. U. S. A. 103 2006 19206 19211 17148611 12 Krouk G. Lacombe B. Bielach A. Perrine-Walker F. Malinska K. Mounier E. Hoyerova K. Tillard P. Leon S. Ljung K. Zazimalova E. Benkova E. Nacry P. Gojon A. Nitrate-regulated auxin transport by NRT1.1 defines a mechanism for nutrient sensing in plants Dev. Cell 18 2010 927 937 20627075 13 Krouk G. Tillard P. Gojon A. Regulation of the high-affinity NO3 − uptake system by NRT1.1-mediated NO3 − demand signaling in Arabidopsis Plant Physiol 142 2006 1075 1086 16998085 14 Muños S. Cazettes C. Fizames C. Gaymard F. Tillard P. Lepetit M. Lejay L. Gojon A. Transcript profiling in the chl1-5 mutant of Arabidopsis reveals a role of the nitrate transporter NRT1.1 in the regulation of another nitrate transporter, NRT2.1 Plant Cell 16 2004 2433 2447 15319483 15 Wang R. Xing X. Wang Y. Tran A. Crawford N.M. A genetic screen for nitrate regulatory mutants captures the nitrate transporter gene NRT1.1 Plant Physiol 151 2009 472 478 19633234 16 Lewis C.E. Noctor G. Causton D. Foyer C.H. Regulation of assimilate partitioning in leaves Funct. Plant Biol 27 2000 507 519 17 Kaiser W.M. Kandlbinder A. Stoimenova M. Glaab J. Discrepancy between nitrate reduction rates in intact leaves and nitrate reductase activity in leaf extracts: what limits nitrate reduction in situ? Planta 210 2000 801 807 10805452 18 Kinoshita H. Nagasaki J. Yoshikawa N. Yamamoto A. Takito S. Kawasaki M. Sugiyama T. Miyake H. Weber A.P. Taniguchi M. The chloroplastic 2-oxoglutarate/malate transporter has dual function as the malate valve and in carbon/nitrogen metabolism Plant J 65 2011 15 26 21175886 19 Selinski J. Scheibe R. Lack of malate valve capacities lead to improved N-assimilation and growth in transgenic A. thaliana plants Plant Signal. Behav 9 2014 20 Guo S. Schinner K. Sattelmacher B. Hansen U.-P. Different apparent CO2 compensation points in nitrate- and ammonium-grown Phaseolus vulgaris and the relationship to non-photorespiratory CO2 evolution Physiol. Plant 123 2005 288 301 21 Dutilleul C. Lelarge C. Prioul J.L. De Paepe R. Foyer C.H. Noctor G. Mitochondria-driven changes in leaf NAD status exert a crucial influence on the control of nitrate assimilation and the integration of carbon and nitrogen metabolism Plant Physiol 139 2005 64 78 16126851 22 Finkemeier I. Laxa M. Miguet L. Howden A. Sweetlove L. Proteins of diverse function and subcellular location are lysine-acetylated in Arabidopsis Plant Physiol 155 4 2011 1779 1790 21311031 23 Buchanan B.B. Balmer Y. Redox regulation: a broadening horizon Annu. Rev. Plant Biol 56 2005 187 220 15862094 24 Foyer C.H. Noctor G. Stress-triggered redox signalling: what's in pROSpect? Plant Cell Environ 39 2016 951 964 26264148 25 Yoshida K. Noguchi K. Motohashi K. Hisabori T. Systematic exploration of thioredoxin target proteins in plant mitochondria Plant Cell Physiol 54 2013 875 892 23444301 26 Daloso D.M. Muller K. Obata T. Florian A. Tohge T. Bottcher A. Riondet C. Bariat L. Carrari F. Nunes-Nesi A. Buchanan B.B. Reichheld J.P. Araujo W.L. Fernie A.R. Thioredoxin, a master regulator of the tricarboxylic acid cycle in plant mitochondria Proc. Natl. Acad. Sci. U. S. A. 112 2015 E1392 E1400 25646482 27 Balmer Y. Vensel W.H. Tanaka C.K. Hurkman W.J. Gelhaye E. Rouhier N. Jacquot J.P. Manieri W. Schurmann P. Droux M. Buchanan B.B. Thioredoxin links redox to the regulation of fundamental processes of plant mitochondria Proc. Natl. Acad. Sci. U. S. A. 101 2004 2642 2647 14983062 28 Scheibe R. Dietz K.J. Reduction-oxidation network for flexible adjustment of cellular metabolism in photoautotrophic cells Plant Cell Environ 35 2012 202 216 21410714 29 Michelet L. Zaffagnini M. Morisse S. Sparla F. Perez-Perez M.E. Francia F. Danon A. Marchand C.H. Fermani S. Trost P. Lemaire S.D. Redox regulation of the Calvin-Benson cycle: something old, something new Front. Plant Sci 4 2013 470 24324475 30 Geigenberger P. Fernie A.R. Metabolic control of redox and redox control of metabolism in plants Antioxid. Redox Signal 21 2014 1389 1421 24960279 31 Hisabori T. Sunamura E. Kim Y. Konno H. The chloroplast ATP synthase features the characteristic redox regulation machinery Antioxid. Redox Signal 19 2013 1846 1854 23145525 32 Miginiac-Maslow M. Lancelin J.M. Intrasteric inhibition in redox signalling: light activation of NADP-malate dehydrogenase Photosynth. Res 72 2002 1 12 16228530 33 Thormahlen I. Meitzel T. Groysman J. Ochsner A.B. von Roepenack-Lahaye E. Naranjo B. Cejudo F.J. Geigenberger P. Thioredoxin f1 and NADPH-dependent thioredoxin reductase C have overlapping functions in regulating photosynthetic metabolism and plant growth in response to varying light conditions Plant Physiol 169 2015 1766 1786 26338951 34 Hendriks J.H. Kolbe A. Gibon Y. Stitt M. Geigenberger P. ADP-glucose pyrophosphorylase is activated by posttranslational redox-modification in response to light and to sugars in leaves of Arabidopsis and other plant species Plant Physiol 133 2003 838 849 12972664 35 Geigenberger P. Kolbe A. Tiessen A. Redox regulation of carbon storage and partitioning in response to light and sugars J. Exp. Bot 56 2005 1469 1479 15863446 36 Krugel U. Kuhn C. Post-translational regulation of sucrose transporters by direct protein-protein interactions Front. Plant Sci 4 2013 237 23847641 37 Krugel U. He H.X. Gier K. Reins J. Chincinska I. Grimm B. Schulze W.X. Kuhn C. The potato sucrose transporter StSUT1 interacts with a DRM-associated protein disulfide isomerase Mol. Plant 5 2012 43 62 21746698 38 Nahirnak V. Almasia N.I. Fernandez P.V. Hopp H.E. Estevez J.M. Carrari F. Vazquez-Rovere C. Potato snakin-1 gene silencing affects cell division, primary metabolism, and cell wall composition Plant Physiol 158 2012 252 263 22080603 39 Nahirnak V. Almasia N.I. Hopp H.E. Vazquez-Rovere C. Snakin/GASA proteins: involvement in hormone crosstalk and redox homeostasis Plant Signal. Behav 7 2012 1004 1008 22836500 40 Krugel U. Veenhoff L.M. Langbein J. Wiederhold E. Liesche J. Friedrich T. Grimm B. Martinoia E. Poolman B. Kuhn C. Transport and sorting of the Solanum tuberosum sucrose transporter SUT1 is affected by posttranslational modification Plant Cell 20 2008 2497 2513 18790827 41 Krugel U. Veenhoff L.M. Langbein J. Wiederhold E. Liesche J. Friedrich T. Grimm B. Martinoia E. Poolman B. Kuhn C. Transport and sorting of the Solanum tuberosum sucrose transporter SUIT1 is affected by post-translational modification Plant Cell 21 2009 4059 4060 42 Flügge U.-I. Phosphate translocators in plastids Annu. Rev. Plant Biol 50 1999 27 45 43 Robinson S.P. Walker D.A. Distribution of metabolites between chloroplast and cytoplasm during the induction phase of photosynthesis in leaf protoplasts Plant Physiol 65 1980 902 905 16661305 44 Heber U. Walker D.A. The chloroplast envelope — barrier or bridge? Trends Biochem. Sci 4 1979 252 256 45 Signora L. Galtier N. Skøt L. Lucas H. Foyer C.H. Over-expression of sucrose phosphate synthase in Arabidopsis thaliana results in increased foliar sucrose/starch ratios and favours decreased foliar carbohydrate accumulation in plants after prolonged growth with CO2 enrichment J. Exp. Bot 49 1998 669 680 46 Flügge U. Stitt M. Heldt H.W. Light-driven uptake of pyruvate into mesophyll chloroplasts from maize FEBS Lett 183 1985 335 339 47 Riesmeier J.W. Hirner B. Frommer W.B. Potato sucrose transporter expression in minor veins indicates a role in phloem loading Plant Cell Online 5 1993 1591 1598 48 Heineke D. Wildenberger K. Sonnewald U. Willmitzer L. Heldt H.W. Accumulation of hexoses in leaf vacuoles: studies with transgenic tobacco plants expressing yeast-derived invertase in the cytosol, vacuole or apoplasm Planta 194 1994 29 33 49 Baxter C.J. Foyer C.H. Turner J. Rolfe S.A. Quick W.P. Elevated sucrose-phosphate synthase activity in transgenic tobacco sustains photosynthesis in older leaves and alters development J. Exp. Bot 54 2003 1813 1820 12815030 50 Hausler R.E. Schlieben N.H. Nicolay P. Fischer K. Fischer K.L. Flugge U.-I. Control of carbon partitioning and photosynthesis by the triose phosphate/phosphate translocator in transgenic tobacco plants (Nicotiana tabacum L.). I. Comparative physiological analysis of tobacco plants with antisense repression and overexpression of the triose phosphate/phosphate translocator Planta 210 2000 371 382 10750894 51 Cho M.-H. Jang A. Bhoo S. Jeon J.-S. Hahn T.-R. Manipulation of triose phosphate/phosphate translocator and cytosolic fructose-1,6-bisphosphatase, the key components in photosynthetic sucrose synthesis, enhances the source capacity of transgenic Arabidopsis plants Photosynth. Res 111 2012 261 268 22297909 52 Fischer K. Kammerer B. Gutensohn M. Arbinger B. Weber A. Häusler R.E. Flügge U.I. A new class of plastidic phosphate translocators: a putative link between primary and secondary metabolism by the phosphoenolpyruvate/phosphate antiporter Plant Cell Online 9 1997 453 462 53 Eicks M. Maurino V. Knappe S. Flügge U.-I. Fischer K. The plastidic pentose phosphate translocator represents a link between the cytosolic and the plastidic pentose phosphate pathways in plants Plant Physiol 128 2002 512 522 11842155 54 Knappe S. Löttgert T. Schneider A. Voll L. Flügge U.-I. Fischer K. Characterization of two functional phosphoenolpyruvate/phosphate translocator (PPT) genes in Arabidopsis –AtPPT1 may be involved in the provision of signals for correct mesophyll development Plant J 36 2003 411 420 14617097 55 Herrmann K.M. Weaver L.M. The shikimate pathway Annu. Rev. Plant Physiol. Plant Mol. Biol 50 1999 473 503 15012217 56 Streatfield S.J. Weber A. Kinsman E.A. Häusler R.E. Li J. Post-Beittenmiller D. Kaiser W.M. Pyke K.A. Flügge U.-I. Chory J. The phosphoenolpyruvate/phosphate translocator is required for phenolic metabolism, palisade cell development, and plastid-dependent nuclear gene expression Plant Cell Online 11 1999 1609 1621 57 Li H. Culligan K. Dixon R.A. Chory J. CUE1: a mesophyll cell-specific positive regulator of light-controlled gene expression in Arabidopsis Plant Cell Online 7 1995 1599 1610 58 Voll L. Häusler R.E. Hecker R. Weber A. Weissenböck G. Fiene G. Waffenschmidt S. Flügge U.-I. The phenotype of the Arabidopsis cue1 mutant is not simply caused by a general restriction of the shikimate pathway Plant J 36 2003 301 317 14617088 59 Staehr P. Lottgert T. Christmann A. Krueger S. Rosar C. Rolcik J. Novak O. Strnad M. Bell K. Weber A.P. Flugge U.I. Hausler R.E. Reticulate leaves and stunted roots are independent phenotypes pointing at opposite roles of the phosphoenolpyruvate/phosphate translocator defective in cue1 in the plastids of both organs Front. Plant Sci 5 2014 126 24782872 60 Kammerer B. Fischer K. Hilpert B. Schubert S. Gutensohn M. Weber A. Flügge U.-I. Molecular characterization of a carbon transporter in plastids from heterotrophic tissues: the glucose 6-phosphate/phosphate antiporter Plant Cell Online 10 1998 105 117 61 Ruuska S.A. Girke T. Benning C. Ohlrogge J.B. Contrapuntal networks of gene expression during Arabidopsis seed filling Plant Cell 14 2002 1191 1206 12084821 62 Niewiadomski P. Knappe S. Geimer S. Fischer K. Schulz B. Unte U.S. Rosso M.G. Ache P. Flügge U.-I. Schneider A. The Arabidopsis plastidic glucose 6-phosphate/phosphate translocator GPT1 is essential for pollen maturation and embryo sac development Plant Cell Online 17 2005 760 775 63 Dyson B.C. Webster R.E. Johnson G.N. GPT2: a glucose 6-phosphate/phosphate translocator with a novel role in the regulation of sugar signalling during seedling development Ann. Bot 113 2014 643 652 24489010 64 Kunz H.H. Häusler R.E. Fettke J. Herbst K. Niewiadomski P. Gierth M. Bell K. Steup M. Flügge U.I. Schneider A. The role of plastidial glucose-6-phosphate/phosphate translocators in vegetative tissues of Arabidopsis thaliana mutants impaired in starch biosynthesis Plant Biol 12 2010 115 128 20712627 65 Pourtau N. Jennings R. Pelzer E. Pallas J. Wingler A. Effect of sugar-induced senescence on gene expression and implications for the regulation of senescence in Arabidopsis Planta 224 2006 556 568 16514542 66 Flügge U.I. Gao W. Transport of isoprenoid intermediates across chloroplast envelope membranes Plant Biol 7 2005 91 97 15666208 67 Ainsworth E.A. Bush D.R. Carbohydrate export from the leaf: a highly regulated process and target to enhance photosynthesis and productivity Plant Physiol 155 2011 64 69 20971857 68 Wind J. Smeekens S. Hanson J. Sucrose: metabolite and signaling molecule Phytochemistry 71 2010 1610 1614 20696445 69 Zeeman S.C. Kossmann J. Smith A.M. Starch: its metabolism, evolution, and biotechnological modification in plants Annu. Rev. Plant Biol 61 2010 209 234 20192737 70 Maloney V.J. Park J.-Y. Unda F. Mansfield S.D. Sucrose phosphate synthase and sucrose phosphate phosphatase interact in planta and promote plant growth and biomass accumulation J. Exp. Bot 66 14 2015 4383 4394 25873678 71 Schneider T. Keller F. Raffinose in chloroplasts is synthesized in the cytosol and transported across the chloroplast envelope Plant Cell Physiol 50 2009 2174 2182 19880397 72 Kang Y. OUTLAW W.H. Andersen P.C. Fiore G.B. Guard-cell apoplastic sucrose concentration–a link between leaf photosynthesis and stomatal aperture size in the apoplastic phloem loader Vicia faba L Plant Cell Environ 30 2007 551 558 17407533 73 Martinoia E. Kaiser G. Schramm M.J. Heber U. Sugar transport across the plasmalemma and the tonoplast of barley mesophyll protoplasts. Evidence for different transport systems J. Plant Physiol 131 1987 467 478 74 Endler A. Meyer S. Schelbert S. Schneider T. Weschke W. Peters S.W. Keller F. Baginsky S. Martinoia E. Schmidt U.G. Identification of a vacuolar sucrose transporter in barley and Arabidopsis mesophyll cells by a tonoplast proteomic approach Plant Physiol 141 2006 196 207 16581873 75 Paul M.J. Pellny T.K. Carbon metabolite feedback regulation of leaf photosynthesis and development J. Exp. Bot 54 2003 539 547 12508065 76 Neuhaus H.E. Transport of primary metabolites across the plant vacuolar membrane FEBS Lett 581 2007 2223 2226 17307167 77 Weschke W. Panitz R. Sauer N. Wang Q. Neubohn B. Weber H. Wobus U. Sucrose transport into barley seeds: molecular characterization of two transporters and implications for seed development and starch accumulation Plant J 21 2000 455 467 10758497 78 Weise A. Barker L. Kuhn C. Lalonde S. Buschmann H. Frommer W.B. Ward J.M. A new subfamily of sucrose transporters, SUT4, with low affinity/high capacity localized in enucleate sieve elements of plants Plant Cell 12 2000 1345 1355 10948254 79 Wormit A. Trentmann O. Feifer I. Lohr C. Tjaden J. Meyer S. Schmidt U. Martinoia E. Neuhaus H.E. Molecular identification and physiological characterization of a novel monosaccharide transporter from Arabidopsis involved in vacuolar sugar transport Plant Cell 18 2006 3476 3490 17158605 80 Jung B. Ludewig F. Schulz A. Meißner G. Wöstefeld N. Flügge U.-I. Pommerrenig B. Wirsching P. Sauer N. Koch W. Sommer F. Mühlhaus T. Schroda M. Cuin T.A. Graus D. Marten I. Hedrich R. Neuhaus H.E. Identification of the transporter responsible for sucrose accumulation in sugar beet taproots Nat. Plants 1 2015 14001 27246048 81 Aluri S. Büttner M. Identification and functional expression of the Arabidopsis thaliana vacuolar glucose transporter 1 and its role in seed germination and flowering Proc. Natl. Acad. Sci 104 2007 2537 2542 17284600 82 De Schepper V. De Swaef T. Bauweraerts I. Steppe K. Phloem transport: a review of mechanisms and controls J. Exp. Bot 64 2013 4839 4850 24106290 83 Rolland F. Sheen J. Sugar sensing and signalling networks in plants Biochem. Soc. Trans 33 2005 269 271 15667323 84 Kühn C. Grof C.P.L. Sucrose transporters of higher plants Curr. Opin. Plant Biol 13 2010 287 297 85 Lalonde S. Wipf D. Frommer W.B. Transport mechanisms for organic forms of carbon and nitrogen between source and sink Annu. Rev. Plant Biol 55 2004 341 372 15377224 86 Braun D.M. Wang L. Ruan Y.-L. Understanding and manipulating sucrose phloem loading, unloading, metabolism, and signalling to enhance crop yield and food security J. Exp. Bot 65 2014 1713 1735 24347463 87 Sauer N. Molecular physiology of higher plant sucrose transporters FEBS Lett 581 2007 2309 2317 17434165 88 Eom J.-S. Choi S.-B. Ward J.M. Jeon J.-S. The mechanism of phloem loading in rice (Oryza sativa ) Mol. Cell 33 2012 431 438 89 Roberts A.G. Oparka K.J. Plasmodesmata and the control of symplastic transport Plant Cell Environ 26 2003 103 124 90 Turgeon R. Medville R. The absence of phloem loading in willow leaves Proc. Natl. Acad. Sci 95 1998 12055 12060 9751789 91 Haritatos E. Keller F. Turgeon R. Raffinose oligosaccharide concentrations measured in individual cell and tissue types in Cucumis melo L. leaves: implications for phloem loading Planta 198 1996 614 622 92 Chen L.-Q. Hou B.-H. Lalonde S. Takanaga H. Hartung M.L. Qu X.-Q. Guo W.-J. Kim J.-G. Underwood W. Chaudhuri B. Chermak D. Antony G. White F.F. Somerville S.C. Mudgett M.B. Frommer W.B. Sugar transporters for intercellular exchange and nutrition of pathogens Nature 468 2010 527 532 21107422 93 Ayre B.G. Membrane-transport systems for sucrose in relation to whole-plant carbon partitioning Mol. Plant 4 2011 377 394 21502663 94 Zhou Y. Qu H. Dibley K.E. Offler C.E. Patrick J.W. A suite of sucrose transporters expressed in coats of developing legume seeds includes novel pH-independent facilitators Plant J 49 2007 750 764 17253986 95 Aoki N. Hirose T. Scofield G.N. Whitfeld P.R. Furbank R.T. The sucrose transporter gene family in rice Plant Cell Physiol 44 2003 223 232 12668768 96 Zhang W.-H. Zhou Y. Dibley K.E. Tyerman S.D. Furbank R.T. Patrick J.W. Review: nutrient loading of developing seeds Funct. Plant Biol 34 2007 314 331 97 Slewinski T.L. Braun D.M. Current perspectives on the regulation of whole-plant carbohydrate partitioning Plant Sci 178 2010 341 349 98 Kühn C. Quick W. Schulz A. Riesmeier J. Sonnewald U. Frommer W. Companion cell-specific inhibition of the potato sucrose transporter SUT1 Plant Cell Environ 19 1996 1115 1123 99 Hackel A. Schauer N. Carrari F. Fernie A.R. Grimm B. Kuhn C. Sucrose transporter LeSUT1 and LeSUT2 inhibition affects tomato fruit development in different ways Plant J 45 2006 180 192 16367963 100 Riesmeier J.W. Willmitzer L. Frommer W.B. Evidence for an essential role of the sucrose transporter in phloem loading and assimilate partitioning EMBO J 13 1994 1 8306952 101 Ishimaru K. Hirose T. Aoki N. Takahashi S. Ono K. Yamamoto S. Wu J. Saji S. Baba T. Ugaki M. Antisense expression of a rice sucrose transporter OsSUT1 in rice (Oryza sativa L.) Plant Cell Physiol 42 2001 1181 1185 11673635 102 Furbank R.T. Scofield G.N. Hirose T. Wang X.-D. Patrick J.W. Offler C.E. Cellular localisation and function of a sucrose transporter OsSUT1 in developing rice grains Funct. Plant Biol 28 2001 1187 1196 103 Scofield G.N. Hirose T. Gaudron J.A. Furbank R.T. Upadhyaya N.M. Ohsugi R. Antisense suppression of the rice transporter gene, OsSUT1, leads to impaired grain filling and germination but does not affect photosynthesis Funct. Plant Biol 29 2002 815 826 104 Takeda T. Toyofuku K. Matsukura C. Yamaguchi J. Sugar transporters involved in flowering and grain development of rice J. Plant Physiol 158 2001 465 470 105 Chincinska I. Gier K. Krugel U. Liesche J. He H. Grimm B. Harren F.J. Cristescu S.M. Kuhn C. Photoperiodic regulation of the sucrose transporter StSUT4 affects the expression of circadian-regulated genes and ethylene production Front. Plant Sci 4 2013 26 23429841 106 Baud S. Wuillème S. Lemoine R. Kronenberger J. Caboche M. Lepiniec L. Rochat C. The AtSUC5 sucrose transporter specifically expressed in the endosperm is involved in early seed development in Arabidopsis Plant J 43 2005 824 836 16146522 107 Chandran D. Reinders A. Ward J.M. Substrate specificity of the Arabidopsis thaliana sucrose transporter AtSUC2 J. Biol. Chem 278 2003 44320 44325 12954621 108 Chen L.-Q. Qu X.-Q. Hou B.-H. Sosso D. Osorio S. Fernie A.R. Frommer W.B. Sucrose efflux mediated by SWEET proteins as a key step for phloem transport Science 335 2012 207 211 22157085 109 Chen L.-Q. SWEET sugar transporters for phloem transport and pathogen nutrition New Phytol 201 2014 1150 1155 24649486 110 Smith A.M. Zeeman S.C. Smith S.M. Starch degradation Annu. Rev. Plant Biol 56 2005 73 98 15862090 111 Trethewey R.N. Aprees T. The role of the hexose transporter in the chloroplasts of Arabidopsis -thaliana L Planta 195 1994 168 174 112 Yu T.-S. Kofler H. Häusler R.E. Hille D. Flügge U.-I. Zeeman S.C. Smith A.M. Kossmann J. Lloyd J. Ritte G. Steup M. Lue W.-L. Chen J. Weber A. The Arabidopsis sex1 mutant is defective in the R1 protein, a general regulator of starch degradation in plants, and not in the chloroplast hexose transporter Plant Cell 13 2001 1907 1918 11487701 113 Scheidig A. Frohlich A. Schulze S. Lloyd J.R. Kossmann J. Downregulation of a chloroplast-targeted beta-amylase leads to a starch-excess phenotype in leaves Plant J 30 2002 581 591 12047632 114 Niittylä T. Messerli G. Trevisan M. Chen J. Smith A.M. Zeeman S.C. A previously unknown maltose transporter essential for starch degradation in leaves Science 303 2004 87 89 14704427 115 Reidel E.J. Turgeon R. Cheng L. A maltose transporter from apple is expressed in source and sink tissues and complements the Arabidopsis maltose export-defective mutant Plant Cell Physiol 49 2008 1607 1613 18776201 116 Ma Y. Slewinski T.L. Baker R.F. Braun D.M. Tie-dyed1 encodes a novel, phloem-expressed transmembrane protein that functions in carbohydrate partitioning Plant Physiol 149 2009 181 194 18923021 117 Ma Y. Baker R.F. Magallanes-Lundback M. DellaPenna D. Braun D.M. Tie-dyed1 and sucrose export defective1 act independently to promote carbohydrate export from maize leaves Planta 227 2008 527 538 17924136 118 Baker R.F. Slewinski T.L. Braun D.M. The tie-dyed pathway promotes symplastic trafficking in the phloem Plant Signal. Behav 8 2013 e24540 119 Bagnall N. Wang X.-D. Scofield G.N. Furbank R.T. Offler C.E. Patrick J.W. Sucrose transport-related genes are expressed in both maternal and filial tissues of developing wheat grains Funct. Plant Biol 27 2000 1009 1020 120 Scofield G.N. Hirose T. Aoki N. Furbank R.T. Involvement of the sucrose transporter, OsSUT1, in the long-distance pathway for assimilate transport in rice J. Exp. Bot 58 2007 3155 3169 17728297 121 Nunes C. O'Hara L.E. Primavesi L.F. Delatte T.L. Schluepmann H. Somsen G.W. Silva A.B. Fevereiro P.S. Wingler A. Paul M.J. The trehalose 6-phosphate/SnRK1 signaling pathway primes growth recovery following relief of sink limitation Plant Physiol 162 2013 1720 1732 23735508 122 Nuccio M.L. Wu J. Mowers R. Zhou H.P. Meghji M. Primavesi L.F. Paul M.J. Chen X. Gao Y. Haque E. Basu S.S. Lagrimini L.M. Expression of trehalose-6-phosphate phosphatase in maize ears improves yield in well-watered and drought conditions Nat. Biotechnol 33 2015 862 869 26473199 123 Henry C. Bledsoe S.W. Griffiths C.A. Kollman A. Paul M.J. Sakr S. Lagrimini L.M. Differential role for trehalose metabolism in salt-stressed maize Plant Physiol 169 2015 1072 1089 26269545 124 Martínez-Barajas E. Delatte T. Schluepmann H. de Jong G.J. Somsen G.W. Nunes C. Primavesi L.F. Coello P. Mitchell R.A.C. Paul M.J. Wheat grain development is characterized by remarkable trehalose 6-phosphate accumulation pregrain filling: tissue distribution and relationship to SNF1-related protein kinase1 activity Plant Physiol 156 2011 373 381 21402798 125 Lunn J.E. Delorge I. Figueroa C.M. Van Dijck P. Stitt M. Trehalose metabolism in plants Plant J 79 4 2014 544 567 24645920 126 Lunn J.E. Feil R. Hendriks J.H. Gibon Y. Morcuende R. Osuna D. Scheible W.R. Carillo P. Hajirezaei M.R. Stitt M. Sugar-induced increases in trehalose 6-phosphate are correlated with redox activation of ADPglucose pyrophosphorylase and higher rates of starch synthesis in Arabidopsis thaliana Biochem. J 397 2006 139 148 16551270 127 O'Hara L.E. Paul M.J. Wingler A. How do sugars regulate plant growth and development? New insight into the role of trehalose-6-phosphate Mol. Plant 6 2013 261 274 23100484 128 Nunes C. Primavesi L.F. Patel M.K. Martinez-Barajas E. Powers S.J. Sagar R. Fevereiro P.S. Davis B.G. Paul M.J. Inhibition of SnRK1 by metabolites: tissue-dependent effects and cooperative inhibition by glucose 1-phosphate in combination with trehalose 6-phosphate Plant Physiol. Biochem 63 2013 89 98 23257075 129 Eastmond P.J. van Dijken A.J. Spielman M. Kerr A. Tissier A.F. Dickinson H.G. Jones J.D. Smeekens S.C. Graham I.A. Trehalose-6-phosphate synthase 1, which catalyses the first step in trehalose synthesis, is essential for Arabidopsis embryo maturation Plant J 29 2002 225 235 11851922 130 Nagele T. Weckwerth W. Mathematical modeling reveals that metabolic feedback regulation of SnRK1 and hexokinase is sufficient to control sugar homeostasis from energy depletion to full recovery Front. Plant Sci 5 2014 365 25120550 131 Baena-Gonzalez E. Rolland F. Thevelein J.M. Sheen J. A central integrator of transcription networks in plant stress and energy signalling Nature 448 2007 938 942 17671505 132 Zhang Y. Primavesi L.F. Jhurreea D. Andralojc P.J. Mitchell R.A. Powers S.J. Schluepmann H. Delatte T. Wingler A. Paul M.J. Inhibition of SNF1-related protein kinase1 activity and regulation of metabolic pathways by trehalose-6-phosphate Plant Physiol. 149 2009 1860 1871 19193861 133 Paul M.J. Jhurreea D. Zhang Y. Primavesi L.F. Delatte T. Schluepmann H. Wingler A. Upregulation of biosynthetic processes associated with growth by trehalose 6-phosphate Plant Signal. Behav. 5 2010 386 392 20139731 134 Schluepmann H. Paul M. Trehalose metabolites in Arabidopsis -elusive, active and central Arabidopsis Book 7 2009 e0122 135 Schluepmann H. Berke L. Sanchez-Perez G.F. Metabolism control over growth: a case for trehalose-6-phosphate in plants J. Exp. Bot. 2011 136 Kim J.Y. Mahe A. Brangeon J. Prioul J.L. A maize vacuolar invertase, IVR2, is induced by water stress. Organ/tissue specificity and diurnal modulation of expression Plant Physiol. 124 2000 71 84 10982423 137 Zinselmeier S.A. Lauer M.J. Boyer J.S. Reversing drought-induced losses in grain yield: sucrose maintains embryo growth in maize Crop Sci. 35 1995 1390 1400 138 Campos H. Cooper M. Habben J.E. Edmeades G.O. Schussler J.R. Improving drought tolerance in maize: a view from industry Field Crop Res 90 2004 19 34 139 Kolbe A. Tiessen A. Schluepmann H. Paul M. Ulrich S. Geigenberger P. Trehalose 6-phosphate regulates starch synthesis via posttranslational redox activation of ADP-glucose pyrophosphorylase Proc. Natl. Acad. Sci. U. S. A. 102 2005 11118 11123 16046541 140 Martins M.C.M. Hejazi M. Fettke J. Steup M. Feil R. Krause U. Arrivault S. Vosloh D. Figueroa C.M. Ivakov A. Yadav U.P. Piques M. Metzner D. Stitt M. Lunn J.E. Feedback inhibition of starch degradation in Arabidopsis leaves mediated by trehalose 6-phosphate Plant Physiol 163 2013 1142 1163 24043444 141 Kretzschmar T. Pelayo M.A.F. Trijatmiko K.R. Gabunada L.F.M. Alam R. Jimenez R. Mendioro M.S. Slamet-Loedin I.H. Sreenivasulu N. Bailey-Serres J. Ismail A.M. Mackill D.J. Septiningsih E.M. A trehalose-6-phosphate phosphatase enhances anaerobic germination tolerance in rice Nat. Plants 15124 2015 142 Wingler A. Delatte T.L. O'Hara L.E. Primavesi L.F. Jhurreea D. Paul M.J. Schluepmann H. Trehalose 6-phosphate is required for the onset of leaf senescence associated with high carbon availability Plant Physiol 158 2012 1241 1251 22247267 143 Wahl V. Ponnu J. Schlereth A. Arrivault S. Langenecker T. Franke A. Feil R. Lunn J.E. Stitt M. Schmid M. Regulation of flowering by trehalose-6-phosphate signaling in Arabidopsis thaliana Science 339 2013 704 707 23393265 144 Tognetti J.A. Pontis H.G. Martinez-Noel G.M. Sucrose signaling in plants: a world yet to be explored Plant Signal. Behav 8 2013 e23316 145 Mason M.G. Ross J.J. Babst B.A. Wienclaw B.N. Beveridge C.A. Sugar demand, not auxin, is the initial regulator of apical dominance Proc. Natl. Acad. Sci 111 2014 6092 6097 24711430 146 Sweetlove L.J. Kossmann J. Riesmeier J.W. Trethewey R.N. Hill S.A. The control of source to sink carbon flux during tuber development in potato Plant J 15 1998 697 706 147 Slafer G.A. Elia M. Savin R. García G.A. Terrile I.I. Ferrante A. Miralles D.J. González F.G. Fruiting efficiency: an alternative trait to further rise wheat yield Food Energy Secur 4 2015 92 109 148 Long S.P. Marshall-Colon A. Zhu X.-G. Meeting the global food demand of the future by engineering crop photosynthesis and yield potential Cell 161 2015 56 66 25815985

PMC005xxxxxx/PMC5001787.txt

==== Front Biochim Biophys ActaBiochim. Biophys. ActaBiochimica et Biophysica Acta0006-3002Elsevier Pub. Co S0005-2728(16)30575-810.1016/j.bbabio.2016.07.003ArticleIdentifying involvement of Lys251/Asp252 pair in electron transfer and associated proton transfer at the quinone reduction site of Rhodobacter capsulatus cytochrome bc1 Kuleta Patryk aSarewicz Marcin aPostila Pekka bRóg Tomasz cdOsyczka Artur artur.osyczka@uj.edu.pla⁎a Department of Molecular Biophysics, Faculty of Biochemistry, Biophysics and Biotechnology, Jagiellonian University in Krakow, Gronostajowa 7, 30-387 Kraków, Polandb Structural Bioinformatics Laboratory, Biochemistry, Faculty of Science and Engineering, Åbo Akademi University, Turku, Finlandc Department of Physics, Tampere University of Technology, P.O. Box 692, FI-33101 Tampere, Finlandd Department of Physics, University of Helsinki, P.O. Box 64, FI-00014 Helsinki, Finland⁎ Corresponding author. artur.osyczka@uj.edu.pl1 10 2016 10 2016 1857 10 1661 1668 20 4 2016 22 6 2016 10 7 2016 © 2016 The Authors2016This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).Describing dynamics of proton transfers in proteins is challenging, but crucial for understanding processes which use them for biological functions. In cytochrome bc1, one of the key enzymes of respiration or photosynthesis, proton transfers engage in oxidation of quinol (QH2) and reduction of quinone (Q) taking place at two distinct catalytic sites. Here we evaluated by site-directed mutagenesis the contribution of Lys251/Asp252 pair (bacterial numbering) in electron transfers and associated with it proton uptake to the quinone reduction site (Qi site). We showed that the absence of protonable group at position 251 or 252 significantly changes the equilibrium levels of electronic reactions including the Qi-site mediated oxidation of heme bH, reverse reduction of heme bH by quinol and heme bH/Qi semiquinone equilibrium. This implicates the role of H-bonding network in binding of quinone/semiquinone and defining thermodynamic properties of Q/SQ/QH2 triad. The Lys251/Asp252 proton path is disabled only when both protonable groups are removed. With just one protonable residue from this pair, the entrance of protons to the catalytic site is sustained, albeit at lower rates, indicating that protons can travel through parallel routes, possibly involving water molecules. This shows that proton paths display engineering tolerance for change as long as all the elements available for functional cooperation secure efficient proton delivery to the catalytic site. Highlights • Contribution of D252 and K251 in Qi catalysis of cyt bc1 is explored by mutagenesis. • The proton path is disabled only when both protonable groups are eliminated. • H-bonding network influences thermodynamic properties of Q/SQ/QH2 triad in Qi site. • Parallel routes for proton transfer to the Qi site are implicated. Keywords Cytochrome bc1Mitochondrial complex IIIElectron transferProton transferQuinone ==== Body 1 Introduction Proton translocation across energy conserving membrane is crucial for generation of proton motive force. In Peter Mitchell's redox loop mechanism, proton translocation is achieved by a functional coupling of two reactions: an oxidation of quinol with release of two protons at one side of the membrane and a reduction of quinone with uptake of two protons at the opposite side of the membrane [1], [2], [3]. The quinol oxidation and quinone reduction sites can be located in two separate enzymes (bacterial examples [4]), or they can be assembled within one enzyme. The latter case concerns cytochrome bc1, a key component of many photosynthetic and respiratory systems including mitochondrial respiration [5], [6]. Cytochrome bc1 is a functional dimer [7]. The quinol oxidation and quinone reduction sites are located within cytochrome b subunit, which together with cytochrome c1 and iron-sulfur (ISP) subunit form the catalytic core of the monomer [8]. The quinol oxidation and quinone reduction sites are named the Qo and Qi sites, respectively. In the Qo site, the oxidation of quinol releases two protons to the intermembrane space. The electrons from this reaction are directed into two separate cofactor chains. The high potential c-chain transfers one electron to cytochrome c via iron-sulfur cluster [2Fe-2S], while the low potential b-chain delivers the second electron through hemes bL and bH to the Qi site. The sequential reduction of quinone to quinol through a semiquinone intermediate (SQi) is associated with an uptake of two protons from the mitochondrial matrix or cytoplasm [9], [10]. It follows that a complete reduction of one quinone molecule at the Qi site requires oxidation of two quinol molecules at the Qo site. In addition, the electron transfer between two hemes bL is possible [7], [11], [12], [13]. This secures functional connection of the two Qo and two Qi sites in the dimer. While the electron paths within cytochrome bc1 are well defined, the proton paths are much less known. This is in part due to the lack of methods that can directly monitor proton transfers. While uncertainties related with proton transfers concern both the Qo and Qi sites, here we focus just on the Qi site. Before X-ray structures of cytochrome bc1 were known, early site-directed mutagenesis successfully identified several key protonable residues associated with the operation of the Qi site [10], [14], [15]. However, the majority of models incorporating the protonation/deprotonation steps at this site were inferred from the inspection of X-ray structures [16], [17], [18]. Complementary studies based on electron paramagnetic resonance spectroscopy provided information on paramagnetic semiquinone bound to the Qi site [19], [20], [21]. In addition, Poisson-Boltzmann electrostatic calculations described redox-linked protonation state changes for this site [22]. All these studies point towards several important polar residues (His217, Asp252, Lys251, Asn221 in bacterial numbering) that can potentially be involved in the substrate binding (Q and SQi) and/or its protonation/deprotonation. Besides these amino acid side chains, cardiolipin (CL) was also postulated to facilitate proton transfers at the entry point from the protein exterior (dimer interface) to the Qi site. In this scenario, CL together with a neighboring lysine residue (Lys251) and water molecules can form the CL/K pathway delivering protons to the site [16], [23], [24]. Our recent MD simulation study [25] suggests that the role of Lys251 is more direct than the prior CL/K pathway hypothesis implied. After acquiring a proton from the dianionic CL head group the positively charged Lys251 could rotate into the Qi site to form a salt bridge with the deprotonated and negatively-charged Asp252 side chain. This fully bent Lys251 conformation, which is not seen in any substrate-bound X-ray crystal structures, results from semiquinone binding in the simulations, but pKa calculations indicate that the switch-like motion would be pH-dependent and possible even without a bound substrate at the Qi site. The rotation of the Lys251 side-chain implicates the possibility of functional connection between Lys251 and Asp252 for proton transfers to the Qi site. In view of this new finding, we examined the consequences of replacements of Lys251 and Asp252 with non-protonable residues for the functioning of cytochrome bc1 in vivo and for the kinetics of electron and proton transfers. Comparative analysis of separate replacements of either Lys251 or Asp252 side chains (single mutants) and simultaneous replacements of both side chains (double mutants) supports the idea that functional cooperation between Lys251 and Asp252 facilitates proton transfers to the Qi site. It also reveals a limited plasticity of this path to accommodate a lack of one, but not two of protonable groups from the Lys251/Asp252 pair. 2 Methods 2.1 Mutant preparation Rhodobacter (R.) capsulatus cells containing substitutions at 251 and 252 positions in cytochrome b subunit were obtained using a genetic system originally developed by Dr. F. Daldal [26]. Mutations K251M, D252A, D252N were introduced in the cytochrome b gene using QuikChange site-directed mutagenesis system (Stratagene) and the following PCR primers:D252A_F: 5′-TAT TTC GTG ATC AAG GCG CTG TTC GCG CTG GCC-3′; D252A_R: 5′-CAG CGC GAA CAG CGC CTT GAT CAC GAA ATA CGG-3′; D252N_F: 5′-TTC GTG ATC AAG AAC CTG TTC GCG CTG GC-3′; D252N_R: 5′-AG CGC GAA CAG GTT CTT GAT CAC GAA ATA CGG-3′; K251M_F: 5′-G TAT TTC GTG ATC ATG GAC CTG TTC GCG C-3′; K251M_R: 5′-C GAA CAG GTC CAT GAT CAC GAA ATA CGG C-3′; K251M/D252A_F: 5′-G TAT TTC GTG ATC ATG GCG CTG TTC GCG CTG GCC C-3′; K251M/D252A_R: 5′-GC GAA CAG CGC CAT GAT CAC GAA ATA CGG C-3′; K251M/D252N_F: 5′-CCG TAT TTC GTG ATC ATG AAC CTG TTC GCG CTG GCC C-3′; K251M/D252N_R: 5′-GGC CAG CGC GAA CAG GTT CAT GAT CAC GAA ATA CGG C-3′. As a template DNA pPET1 plasmid containing wild type (WT) petABC operon was used. The BstXI-XmaI fragment of the operon containing the desired mutations, and no other mutations, were inserted into pMTS1 vector and introduced into MT-RBC1 R. capsulatus strain using triparental crossing [26]. The presence of introduced mutations was confirmed by sequence analysis of petB gene on plasmid isolated from mutated R. capsulatus strains. R. capsulatus bacteria were grown under semiaerobic or photoheterotrophic conditions as described previously [27]. To test for the occurrence of reversion mutations, 100 μl of 1 l overnight liquid culture of the mutant strains were spread on mineral-peptone-yeast extract (MPYE) plates and kept in selective photosynthetic cultures for 12 days. Single colonies that acquired the Ps+ phenotype (photosynthetic competence) were isolated, and reversion mutations were identified by sequencing the entire petABC operon. 2.2 Isolation of chromatophores and protein purification Procedure described previously in ref. [28] was used to obtain the chromatophore membranes from R. capsulatus cells growing under semiaerobic conditions. After isolation, chromatophores were homogenized and suspended in MOPS pH 7.0 or Tris pH 9.0 buffer (for light-induced electron transfer measurements) or in 50 mM Tris pH 8.0 buffer containing 100 mM NaCl, 0.01% DDM and 20% glycerol (for protein purification). Cytochrome bc1 complexes were isolated from detergent-solubilized chromatophores using ion-exchange chromatography (DEAE-BioGel A) as described [28]. 2.3 Light-induced electron transfer measurements Double-wavelength time-resolved optical spectrophotometer [29] was used to measure the kinetics of electron transfer through hemes of cytochrome bc1 in chromatophores. Transient kinetics of hemes b were measured at 560–570 nm after activation by single saturating flash (~ 10 μs). Measurements were performed at pH 7.0 (50 mM MOPS, 100 mM KCl, 1 mM EDTA) or pH 9.0 (50 mM Tris, 100 mM KCl, 1 mM EDTA) under conditions of low (100 mV) or high (200 mV, 250 mV) ambient redox potential. Experiments were performed under anaerobic conditions in the presence of redox mediators and valinomycin as described in [29] except the carotenoid bandshift measurements for which the valinomycin was omitted. The rates of flash-induced electron transfer reactions were calculated from single exponential function fitted to: heme bH reduction in the presence of antimycin, bH re-oxidation without inhibitors and to heme bH reduction from reverse reaction in the presence of myxothiazol (Table 1). 2.4 EPR measurements of semiquinone CW EPR spectra of semiquinone were obtained for isolated cytochrome bc1 complexes. Samples of WT and mutants were measured at 200 K in 50 mM Tris buffer pH 8.0 containing 100 mM KCl, 0.01% DDM and 1 mM EDTA. All spectra were obtained using the following parameters: microwave frequency – 9.39 GHz, sweep width - 180 G, modulation amplitude – 10 G, microwave power – 1.9 mW. Semiquinone was generated in samples by incubation of 50 μM cytochrome bc1 with myxothiazol (Qo site inhibitor) and subsequent addition of 2,3-dimethoxy-5-methyl-6-decyl-1,4-benzohydroquinone (DBH2) as a substrate. The negative control was obtained by addition of antimycin (Qi site inhibitor) to samples treated previously with myxothiazol and DBH2. Both DBH2 and myxothiazol were used at final concentration of 200 μM while antimycin was used at 400 μM. Quantitative EPR analysis of the semiquinone was performed using 4-Hydroxy-TEMPO (TEMPOL) as a standard as described in [30]. To obtain the calibration curve, TEMPOL was measured under the same buffer, temperature and EPR parameters conditions as those used for SQi measurements. 3 Results 3.1 General biochemical and phenotypic properties of mutants of D252 and K251 Conclusions drawn from MD simulations described by Postila et al. [25] and other studies [10], [18], [19] point out four important side chains in SQ binding: Lys251, Asp252, Asn221 and His217 (Fig. 1B). From those we chose Lys251 and Asp252 for experimental testing through site-directed mutagenesis. For this purpose we constructed three single mutants K251M, D252A, D252N and two double mutants K251M/D252A, K251M/D252N. The rationale behind the substitutions of Lys to Met and Asp to Asn was to change the protonable side chains into the non-protonable ones with minimal structural distortions. The substitution of Asp to Ala also tested the removal of protonable group with, possibly, additional structural effects. The properties of those mutants and the most insightful kinetic data are summarized in Table 1 and Fig. 2, Fig. 3, Fig. 4, Fig. 5, Fig. 6, Fig. 7. The electrophoretic analysis of isolated complexes indicated that in all cases the mutant cells expressed cytochrome bc1 with all three catalytic subunits (SDS-page profiles showed the presence of three bands corresponding to cyt c1, cyt b and the FeS subunit). The difference optical spectra of all mutated complexes in the isolated form were similar to that of the native complex. The ability to grow under photosynthetic (Ps) conditions, which tests functionality of cytochrome bc1 in vivo [12], [13], [26], [31], [32] indicated that among the mutants only K251M showed a Ps + growth rate comparable to WT (Table 1). D252A showed a very weak Ps growth indicating severe functional impediment. The Ps growth in D252N was better than D252A, however still less robust than that of WT. Both double mutants did not grow under photosynthetic conditions indicating that cytochrome bc1 is not functional in vivo (Table 1). Incubation of D252A under photosynthetic conditions allowed us to isolate single colonies that exhibited faster Ps growth than original D252A. The DNA sequence analysis of these cells revealed that Ala at position 252 was replaced by Glu. In addition, the reversions were observed for the double mutants: K251M/D252A or K251M/D252N regained Ps + phenotype by placing Glu or Asp at position 252, respectively (Table 1). 3.2 Kinetics of light-induced electron transfer To assay the Qi site function in the mutants we analyzed the rates and amplitudes of light-induced electron transfer in chromatophore membranes under various redox conditions in the absence or presence of inhibitors specifically inactivating Qo or Qi sites [29], [33], [34]. Kinetic transients shown in Fig. 2 compare redox changes of heme bH (measured at 560–570 nm) under ambient redox potential setting hemes b oxidized and the quinone pool half-reduced prior to flash activation. Under these conditions, heme bH in the native enzyme undergoes light-induced reduction followed by re-oxidation (Fig. 2A, black trace). The reduction phase is associated with the oxidation of quinol at the Qo site. The re-oxidation phase occurs through the action of the Qi site (reduction of quinone to semiquinone and then semiquinone to quinol) and is blocked by antimycin, a potent inhibitor of this site (Fig. 2A, red trace) [35]. In the presence of both antimycin and myxothiazol (inhibitor of the Qo site [36]) the enzyme is fully blocked and changes in the redox state of heme bH do not occur (Fig. 2A, blue trace). The kinetic transients shown in Fig. 2 indicate that the mutants do not impede the reduction phase observed in the presence of antimycin (red traces in Fig. 2, and rates in Table 1). However, the re-oxidation phase observed in the absence of any inhibitor is clearly slowed down or blocked (Fig. 2, black traces, and rates in Table 1). In the group of single mutants D252A and D252N showed approximately six fold decrease in the rate of this phase, comparing to WT while in K251M, the slowing was less severe (did not exceed two times). In double mutants (K251M/D252A, K251M/D252N), re-oxidation of hemes b did not occur on a millisecond timescale (Table 1). Kinetic transients shown in Fig. 3 compare redox changes of heme bH under ambient redox potential setting hemes b and quinone pool oxidized prior to flash activation. Under these conditions the amount of quinol molecules after flash activation is limited and approximately only one quinol is oxidized in every Qo site. This leads to reduction of heme bH which equilibrates with the occupant of the Qi site. This equilibration is reflected in a difference in amplitudes of heme bH reduction in the absence and presence of antimycin (black and red, respectively). While the reduction rates in the presence of antimycin in all mutants are similar and comparable to WT (Table 1) the level of heme bH reduction in the absence of any inhibitors is elevated in the mutants. In single mutants (K251M, D252A, D252N) this level approaches approximately 70% of the maximum reduction level (seen in the presence of antimycin), in the double mutants, it reaches the maximum reduction level (the amplitude of black and red trace are comparable). Kinetic transients shown in Fig. 4 (blue traces) monitor the electron transfer from QH2 to heme bH at the Qi site (reverse reaction) under conditions where the Qo site is blocked by myxothiazol and the reduction power of Q pool is increased (by increasing pH). Reduction of heme bH under these conditions is not observed on a millisecond time scale in D252A and in both double mutants. In D252N this reaction is 70 times slower than in WT (see the rates in Table 1). In K251M, the slowing of the rate is not as severe as in D252N (5 times). At the same time, the amplitude of reverse heme bH reduction in K252M is much higher and, unlike in WT, exceeds the amplitude of heme b reduction in the absence of inhibitors (compare blue vs black in WT and K251M). 3.3 Monitoring electrogenic reactions associated with cytochrome bc1 To get information on proton uptake from bulk solution to the Qi site, we conducted a series of measurements of electrogenic reactions associated with the operation of cytochrome bc1 by following the antimycin-sensitive phase of carotenoid bandshift (Fig. 5 and Table 1) [37], [38]. In K251M this phase is comparable to WT. D252A and D252N show decrease in the amplitude of this phase which in D252A additionally has a clearly slower rate. In contrast to single mutants, both double mutants (K251M/D252A, K251M/D252N) do not reveal antimycin-sensitive phase of carotenoid bandshift. 3.4 Testing the SQi levels by EPR Semiquinone in the Qi is observed by EPR as antimycin-sensitive radical signal with gx transition – 2.004 (Fig. 6). Typically, the signal is generated in the samples of isolated cytochrome bc1 exposed to excess of quinol in the presence of myxothiazol. These conditions favor reverse reaction in the Qi site in which reduction of heme bH by QH2 leads to formation of stable SQi [9], [19], [39], [40], [41]. Fig. 6 shows that under these conditions (and with comparable concentrations of cytochrome bc1) clear SQi signal can be observed only in WT and D252N (Fig. 6A, C). Quantitative estimation of SQi concentration indicated the ratio [SQi]/[cytochrome bc1] of 0.34 and 0.18 for WT and D252N, respectively. Traces of SQi signals were observed in D252A while no SQi signal was detected in K251M and double mutants (K251M/D252A, K251M/D252N). 4 Discussion 4.1 Experimental evidence for involvement of Lys251 and Asp252 in electron/proton reactions in the Qi site The roles of Lys251 and Asp252 in proton management of the Qi site, suggested by MD simulations [25] are supported by the effects of mutations observed here and in previous studies [16], [18], [19]. The results consistently indicate that mutating Lys251 and/or Asp252 alters the operation of the Qi site without much influence on the Qo site. The unaffected Qo site was inferred from little influence of the mutations on the rates of Qo site-mediated heme bH reduction (Fig. 2, Fig. 3, red traces). The influence of mutations on the Qi site was revealed by various changes in both the electron transfer reactions associated with redox reactions of the Qi site and cytochrome bc1-related proton translocation. The observation that the rate of the re-oxidation of heme bH (Fig. 2, black traces) was slowed down (single mutants) or blocked (double mutants) indicates impediments in electron and proton reactions that involve first electron transfer from heme bH to Q and subsequent electron transfer from heme bH to SQ to complete Q reduction. Similar slowing of the re-oxidation of heme bH was observed in K251M mutant of R. sphaeroides, but not in the other mutant at this position (K251I) for which the kinetics comparable to WT were reported [14]. The two mutants of Asp252 (D252A and D252N) in this species exhibited lack of heme bH re-oxidation in the light-induced kinetics in the absence of inhibitors [14]. This was clearly a more severe impediment comparing to the respective mutants shown here. The redox equilibrium level between heme bH and Q or SQ was shifted in the mutants towards reduction of heme bH in comparison to WT (Fig. 3, black vs red traces), implicating that heme bH in mutants faces difficulty in delivering electron to quinone occupying the Qi site. This effect is apparently not a result of a changing in the redox midpoint potential (Em) of heme bH given the values of Em determined by redox potentiometry (Table 1). These changes of equilibrium are also evident from the measurements of reverse reactions at the Qi site, associated with electron transfer from quinol to oxidized heme bH (Fig. 4). For all these mutants the process of proton uptake from bulk solutions to the Qi site in the mutants, was inferred from the measurements of blue-shift of absorption spectra of carotenoids (carotenoid bandshift) upon generation of transmembrane electric field. The antimycin-sensitive phase of carotenoid bandshift is associated with the action of cytochrome bc1 complex. Concerning the previous studies [38], [42], [43], [44] and our results we assume that this phase reflects the reactions associated with two protons uptake from aqueous phase into the Qi site after the full quinone reduction is completed. This concerns protonation of oxygen atoms at both the C-1 (through the K251/D252 path) and C-4 groups (through the H217 path) of reduced quinone. In light of this assumption, the diminished amplitude of the carotenoid bandshift phase in D252A and D252N, and additional slowing in D252A, reflect overall difficulty in uptake of protons to the Qi site, while the elimination of this phase in double mutants indicate much more severe blocking of this process. Single K251M does not influence much the proton uptake, as indicated by similar rate and amplitude of the carotenoid bandshift phase in this mutant (comparing to WT). The mutants of Asp252 in R. sphaeroides also affected this phase: D252N showed a slowing, with diminished amplitude while in D252A this phase was abolished. K251M showed a slower phase without amplitude change. In all three cases, changes in the carotenoid bandshift appear to be more severe in R. sphaeroides than the effects of respective mutants shown here [14]. They, however, seem to reflect the same phenomenon: perturbed proton transfers to the Qi site. This, in view of electron transfer measurements, MD simulations and crystal structure data, is most likely associated with the hampered K251/D252 path affecting protonation of quinone C-1 carbonyl. The role of His217 in C-4 carbonyl protonation is inferred from previous studies which showed that replacing His217 to Asp or Arg yielded enzymatically active complexes functional in vivo but replacement to Leu deactivated the enzyme leading to loss of its functional competence in vivo [10]. Interestingly, H217L fully abolished the antimycin-sensitive phase of carotenoid bandshift, similarly to the effects of double mutants reported here. Thus, the lack of this phase in H217L or double mutants suggests that blocking of just one proton path (either K251/D252 path or H217 path) eliminates the proton uptake in both paths, implicating functional coupling (connection) between them. We note that, if this and other mutational works including [9], [14], are considered, there is a correlation between the occurrence of antimycin-sensitive carotenoid bandshift phase and the functionality of cytochrome bc1 in vivo: only mutants that show this phase at measurable rates and amplitudes are able to grow photosynthetically. This is understandable, if one considers that the efficiency of proton transfers ultimately defines proton motive generating capacity of the enzyme in vivo. This further substantiates the notion that this phase reflects the protons uptake from aqueous phase into the Qi site. Additional indication for involvement of D252 in proton transfer came from the observation that barely functional D252A and non-functional K251M/D252A or K251M/D252N mutants regained functionality by restoring protonable group (either E or D) at position 252 (Table 1). 4.2 The role of H-bonding network in binding of quinone/semiquinone and defining thermodynamic properties of Q/SQ/QH2 triad Considering all kinetic traces shown in (Fig. 2, Fig. 3, Fig. 4), the data from measurements of carotenoid bandshift (Fig. 5) and the EPR data on SQi (Fig. 6) we may draw the general conclusions on the influence of the mutations on changing the equilibrium of electron transfer and associated with it protonation/deprotonation within the Qi site. The most obvious results are found for the double mutants for which the mechanistic picture is rather simple. Removing of two important protonable side chains within the Qi site exerts a synergistic effect on both electron transfer (there is neither Q/SQ reduction in forward mode (Fig. 2E, F) nor QH2 oxidation via reverse reaction (Fig. 4E, F) nor detectable SQi (Fig. 6E, F)) and proton transfer (no observable cytochrome bc1-mediated proton transfers from outside of the protein to the Qi site (Fig. 5E, F)). All these effects could result from a lack or improper binding of substrate at the site. The more complex effects are associated with single replacements of either K251 or D252 with non-protonable amino acids. Although the reactions associated with electron transfer between Q or QH2 and heme bH are generally similar for K251M, D252A and D252N we notice some differences that result from different effect of Lys and Asp on Q/SQ/QH2 binding and proton transfer between protein interior and exterior. The sharpest differences between Lys and Asp mutants become visible when analyzing traces in which only theoretically one-electron reactions are involved. It is clear that when Q is awaiting electron from heme bH in all three mutants K251M, D252A or D252N the electron is mostly retained at the level of heme bH as if the potential of Q/SQ couple was lowered. For K251M, it may reflect a higher degree of deprotonation of Asp carboxyl group that cannot be stabilized by interaction with amine group of Lys which leads to destabilization (weaker binding) of Q or SQ within the Qi site. This destabilization seems to be even more severe for mutant having Asp replaced with non-protonable residues (D252A and D252N) for which there is no direct partner for quinone or semiquinone that may deliver proton and stabilize the binding. Interestingly, when considering reverse reaction (QH2 oxidation by heme bH in the Qi site) the differences between the mutants shed light on the proton reactions associated with the SQ/QH2 couple. A lack of QH2 oxidation in D252A mutant indicates that deprotonation of QH2 is blocked when direct proton exchanger (Asp) is replaced by hydrophobic residue. As a result, the semiquinone at the Qi site cannot be effectively formed (Fig. 6B) nor detectable heme bH reduction is observed (Fig. 4B). This is even though the proton path from the site to the bulk still exists (with the help of Lys251). D252N mutant encounters similar difficulty, yet the reverse reaction follows but at a very slow rate when compared to WT. In contrast to Ala in D252A, the polar Asn does not repel water molecules from the vicinity of quinone. They, in turn, may alleviate the lack of COO- group of Asp, however they are not as efficient in proton exchange as the K251/D252 pair. Thus, the reverse reaction leads to the reduction of heme bH. This reaction is two orders of magnitude slower than WT but proceeds to higher level (Fig. 4C, Table 1). Correspondingly, clear EPR signal of SQi can be detected in this mutant, although its amplitude is lower, when compared to WT (Fig. 6C). In K251M, unlike in D252A or D252N, the efficiency of reverse reaction is unexpectedly high, exceeding the level of WT, as if the interior of the protein was much more alkaline. To explain this, we assume that amine group of Lys251 in WT stabilizes “proper” protonation of Asp carboxyl group and the removal of the amine group in the mutants promotes fast deprotonation of SQ/QH2 within the site. Consequently, protons from QH2 are sequentially removed with a help of Asp and then full deprotonation promotes transfer of two electrons to the b-chain yielding high level of reduced hemes b. This apparent lowering of the redox potential of QH2/SQ/Q triad, induced by a very efficient deprotonation, leads to disappearance of the semiquinone EPR signal (Fig. 6D) due to the fact, that upon reverse reaction, the Qi site is overwhelmingly occupied by Q instead of being occupied by QH2 or SQ. In summary, the changes in electron transfer drawn from the reverse reactions associated with different deprotonation reactions allow us to make a general picture of possible equilibration states of Qi-site occupant and heme bH (Fig. 7). Single mutant D252A and double mutants K251M/D252A and K251M/D252N show neither semiquinone signal nor reduced heme bH as the impaired deprotonation of QH2 prevents any efficient reactions in the site. In WT, Asp252 side chain interacting with K251 allows the deprotonation of QH2 promoting a generation of relatively high level of SQ and moderate level of heme bH reduction. It can be envisaged that in this case amount of QH2 oxidized to SQ equals the amount of reduced heme bH. In D252N the deprotonation is even more efficient than in WT, however this is not associated with an elevated level of SQ. This is simply because the electronic equilibrium is shifted from SQ to heme bH yielding lower amplitude of SQ and higher level of bH heme reduced. In this case more than one electron from QH2 is transferred to the b-chain. In K251M, two protons are removed from the vicinity of the bound QH2 of SQ which leads to the most efficient reverse reaction - two electrons from QH2 eventually go to the b-chain. Thus in equilibrium the Qi site is occupied by Q instead of SQ while the level of reduced heme bH is the highest among the tested cytochrome bc1 forms. 4.3 Parallel routes for proton transfer to the Qi site In several studies, Lys251 and Asp252 have been considered as good candidates for residues securing proton delivery from the peripheral CL to the C-1 carbonyl of quinone [16], [18], [19], [20], [22], [24]. The possible cooperation of these two residues in proton transfer became most evident in recent MD simulations which demonstrated that the side chain of Lys251 can rotate from the periphery of the complex towards the Qi site where formation of a salt bridge with the side chain of Asp252 is possible. In view of this observation, the most obvious scenario leading to protonation of the C-1 carbonyl of quinone involves a sequential protonation of Lys251 and Asp252, as described in detail by Postila et al. [25]. We emphasize, however, that in light of experimental results, any scenario assuming a sequential mechanism of transfer of protons involving Lys251 and Asp252 should be considered as a possible, but certainly not the unique path available for protons to enter the Qi site. Alternative pathway/pathways omitting either Lys251 or Asp252 must exist in single mutants having non-protonable side chains at either of these positions (K251M or D252N), as these mutants still retain much of the electron and proton transfer capabilities and remain functional in vivo. This could be result of another protonable group/groups, possibly water molecules, taking over the function of the original side chains that are missing in the mutants, or a reminiscence of natural existence of parallel (multiple) paths for protons in native protein [45]. The latter explanation is quite reasonable in light of the multiplicity for proton paths considered in the case of other quinone binding sites, such as the QB site of photosynthetic reaction center [46], [47], [48]. However, the double mutants show that the simultaneous presence of non-protonable side chains at both positions (K251M/D252A, K251M/D252N) effectively deactivates proton entry to the Qi site which yields mutants non-functional in vivo with fully inactive Qi site. This indicates that at least one of the protonable side chains at either position 251 or 252 must by present. In addition, in R. sphaeroides it was observed that the inversion of charges at positions 251 and 252 (double mutant K251D/D252K) had little effect on enzymatic activity and did not affect the function of enzyme in vivo [49]. This all indicates that proton paths in this system display engineering tolerance for change as long as all the elements available for functional cooperation secure efficient proton delivery to the catalytic site. Transparency document Transparency document. Image 2 Acknowledgements This work was supported by: The Wellcome Trust (09078/Z/10/Z) International Senior Research Fellowship (to A.O.), Academy of Finland (project no. 179571) (Center of Excellence in Biomembrane Research, postdoctoral research fellowship) (to T.R. and P.P.), the Paulo Foundation, and the European Research Council Advanced Grant (project CROWDED-PRO-LIPIDS). We thank Dr. Ewelina Cieluch and Dr. Robert Ekiert for help with preparation of site-directed mutants used in this work. The Transparency document associated with this article can be found, in online version. Fig. 1 A - crystal structure of dimeric cytochrome bc1 from R. capsulatus (PDB: 1ZRT) [8]. Subunits in both monomers are: cytochrome b - green, cytochrome c1 - light gray, ISP subunit - dark gray. Hemes are red sticks. B – close-up view of the Qi site with antimycin (yellow sticks) bound at the site. The protonable residues in the vicinity of the quinone binding site are indicated. C - view of the Qi site as in B with molecular surface added to visualize the entrance to the cavity. B and C show the structure of the Qi site with antimycin from R.sphaeroides (PDB: 2QJP) [50]. Fig. 1Fig. 2 Light-induced heme b reduction and re-oxidation under low ambient redox potential. Transient kinetics at 560–570 nm were followed for WT (A), single mutants D252A (B), D252N (C), K251M (D) and double mutants K251 M/D252A (E), K251M/D252N (F). Traces were recorded without inhibitors (black), after inhibition with antimycin (red), and subsequent inhibition with myxothiazol (blue) at pH 7 and ambient potential of 100 mV. Fig. 2Fig. 3 Light-induced heme b reduction under high ambient redox potential. Transient kinetics at 560–570 nm were followed for WT (A), single mutants D252A (B), D252N (C), K251M (D) and double mutants K251M/D252A (E), K251M/D252N (F). Traces were recorded without inhibitors (black), after inhibition with antimycin (red), and subsequent inhibition with myxothiazol (blue) at pH 7 and ambient potential of 200 mV. Fig. 3Fig. 4 Reduction of heme bH through reverse electron transfer at the Qi site. Transient kinetics at 560–570 nm were followed for WT (A), single mutants D252A (B), D252N (C), K251M (D) and double mutants K251M/D252A (E), K251M/D252N (F). Traces were recorded without inhibitors (black), with myxothiazol (blue), and with both myxothiazol and antimycin present (red) at pH 9 and ambient potential of 250 mV. Fig. 4Fig. 5 Antimycin-sensitive component of carotenoid bandshift measured for WT (A), D252A (B), D252N (C), K251M (D), K251M/D252A (E), K251M/D252N (F) at pH 7 and ambient potential of 100 mV. Traces were obtained by subtracting transients at 475–490 nm measured in the presence of antimycin from the transients measured without inhibitors. Fig. 5Fig. 6 CW EPR spectra of semiquinone radical in isolated cytochrome bc1. Spectra measured for WT (A), D252A (B), D252N (C), K251M (D), K251M/D252A (E), K251M/D252N (F) at pH 8 with excess of DBH2 in the presence of myxothiazol (black). Spectra of the same samples subsequently inhibited with antimycin are shown in red. Conditions of measurements are in the Methods section. Fig. 6Fig. 7 Electron distribution between QH2/SQ/Q and heme bH in WT and mutants. The length of the red bars (X-axis) indicate level of heme bH reduction. Blue areas represent the oxidation level of QH2 (more blue as more quinones occupy the Qi site). The relative level of SQi is showed as amplitude between small arrows (Y-axis). Fig. 7Table 1 Selected properties of cytochrome bc1 mutants. Table 1 Phenotypea Reversions Light-induced heme b reduction Light-induced heme b re-oxidation Heme b reduction from Qi reverse reaction Carotenoid bandshift phase Em of hemes bH bL s− 1 s− 1 s− 1 [mV] WT +++ − 1040 185 400 + 35 − 120 D252A − (+) D252E 1120 32 0 + 43 − 118 D252N ++ − 830 25 6 + 28 − 107 K251M +++ − 1080 100 80 + 36 − 109 K251M/D252A − K251M/D252E 1250 0 0 − ndb nd K251M/D252N − K251M 1200 0 0 − nd nd a +++, indicates Ps growth comparable to WT; ++, indicates Ps growth slower than WT (colonies appear on Ps plates with approximately one day delay comparing to WT); − (+), indicates very weak Ps growth (small colonies appear with approximately five days of delay comparing to WT). b nd, not determined. ==== Refs References 1 Mitchell P. The protonmotive Q cycle: a general formulation FEBS Lett. 59 1975 137 139 1227927 2 Crofts A.R. Meinhardt S.W. Jones K.R. Snozzi M. The role of the quinone pool in the cyclic electron-transfer chain of Rhodopseudomonas sphaeroides : a modified Q-cycle mechanism Biochim. Biophys. Acta 723 1983 202 218 21494412 3 Osyczka A. Moser C.C. Dutton P.L. Fixing the Q cycle Trends Biochem. Sci. 30 2005 176 182 15817393 4 Simon J. van Spanning R.J.M. Richardson D.J. The organisation of proton motive and non-proton motive redox loops in prokaryotic respiratory systems Biochim. Biophys. Acta 1777 2008 1480 1490 18930017 5 Nicholls D.G. Ferguson S.J. Bioenergetics 4 fourth ed. 2013 Academic Press Amsterdam 6 Dikalov S. Khramtsov V. Zimmer G. Determination of rate constants of the reactions of thiols with superoxide radical by electron paramagnetic resonance: critical remarks on spectrophotometric approaches Arch. Biochem. Biophys. 326 1996 207 218 8611025 7 Świerczek M. Cieluch E. Sarewicz M. Borek A. Moser C.C. Dutton P.L. Osyczka A. An electronic bus bar lies in the core of cytochrome bc 1 Science 329 2010 451 454 20651150 8 Berry E.A. Huang L.-S. Saechao L.K. Pon N.G. Valkova-Valchanova M. Daldal F. X-ray structure of Rhodobacter capsulatus cytochrome bc 1 : comparison with its mitochondrial and chloroplast counterparts Photosynth. Res. 81 2004 251 275 16034531 9 Robertson D.E. Prince R.C. Bowyer J.R. Matsuura K. Dutton P.L. Ohnishi T. Thermodynamic properties of the semiquinone and its binding site in the ubiquinol-cytochrome c (c 2 ) oxidoreductase of respiratory and photosynthetic systems J. Biol. Chem. 259 1984 1758 1763 6319410 10 Gray K.A. Dutton P.L. Daldal F. Requirement of histidine 217 for ubiquinone reductase activity (Qi site) in the cytochrome bc 1 complex Biochemistry 33 1994 723 733 8292600 11 Lanciano P. Lee D.-W.D.-W. Yang H. Darrouzet E. Daldal F. Intermonomer electron transfer between the low-potential b hemes of cytochrome bc 1 Biochemistry 50 2011 1651 1663 21261281 12 Czapla M. Cieluch E. Borek A. Sarewicz M. Osyczka A. Catalytically-relevant electron transfer between two hemes b L in the hybrid cytochrome bc 1 -like complex containing a fusion of Rhodobacter sphaeroides and capsulatus cytochromes b Biochim. Biophys. Acta 1827 2013 751 760 23428397 13 Ekiert R. Czapla M. Sarewicz M. Osyczka A. Hybrid fusions show that inter-monomer electron transfer robustly supports cytochrome bc 1 function in vivo Biochem. Biophys. Res. Commun. 451 2014 270 275 25089001 14 Crofts A. Hacker B. Barquera B. Yun C.-H. Gennis R. Structure and function of the bc -complex of Rhodobacter sphaeroides Biochim. Biophys. Acta 1101 1992 162 165 1321665 15 Hacker B. Barquera B. Crofts A.R. Gennis R.B. Characterization of mutations in the cytochrome b subunit of the bc 1 complex of Rhodobacter sphaeroides that affect the quinone reductase site (Qc ) Biochemistry 32 1993 4403 4410 8386545 16 Lange C. Nett J.H. Trumpower B.L. Hunte C. Specific roles of protein-phospholipid interactions in the yeast cytochrome bc 1 complex structure EMBO J. 20 2001 6591 6600 11726495 17 Hunte C. Palsdottir H. Trumpower B.L. Protonmotive pathways and mechanisms in the cytochrome bc 1 complex FEBS Lett. 545 2003 39 46 12788490 18 Gao X. Wen X. Esser L. Quinn B. Yu L. Yu C.-A. Xia D. Structural basis for the quinone reduction in the bc 1 complex: a comparative analysis of crystal structures of mitochondrial cytochrome bc 1 with bound substrate and inhibitors at the Qi site Biochemistry 42 2003 9067 9080 12885240 19 Kolling D.R.J. Samoilova R.I. Holland J.T. Berry E.A. Dikanov S.A. Crofts A.R. Exploration of ligands to the Qi site semiquinone in the bc 1 complex using high-resolution EPR J. Biol. Chem. 278 2003 39747 39754 12874282 20 Dikanov S.A. Samoilova R.I. Kolling D.R.J. Holland J.T. Crofts A.R. Hydrogen bonds involved in binding the Qi-site semiquinone in the bc1 complex, identified through deuterium exchange using pulsed EPR J. Biol. Chem. 279 2004 15814 15823 14736869 21 Dikanov S.A. Holland J.T. Endeward B. Kolling D.R.J. Samoilova R.I. Prisner T.F. Crofts A.R. Hydrogen bonds between nitrogen donors and the semiquinone in the Qi -site of the bc 1 complex J. Biol. Chem. 282 2007 25831 25841 17616531 22 Klingen A.R. Palsdottir H. Hunte C. Ullmann G.M. Redox-linked protonation state changes in cytochrome bc 1 identified by Poisson-Boltzmann electrostatics calculations Biochim. Biophys. Acta 1767 2007 204 221 17349966 23 Wenz T. Hielscher R. Hellwig P. Schägger H. Richers S. Hunte C. Role of phospholipids in respiratory cytochrome bc 1 complex catalysis and supercomplex formation Biochim. Biophys. Acta 1787 2009 609 616 19254687 24 Pöyry S. Cramariuc O. Postila P.A. Kaszuba K. Sarewicz M. Osyczka A. Vattulainen I. Róg T. Atomistic simulations indicate cardiolipin to have an integral role in the structure of the cytochrome bc 1 complex Biochim. Biophys. Acta 1827 2013 769 778 23529178 25 Postila P. Kaszuba K. Kuleta P. Vattulainen I. Marcin S. Osyczka A. Róg T. Switch-like pH-dependent Negative Feedback Mechanism Supplies Protons to the Qi Site of the Cytochrome bc1 Complex Sci. Rep. 2016 (under review) 26 Atta-Asafo-Adjei E. Daldal F. Size of the amino acid side chain at position 158 of cytochrome b is critical for an active cytochrome bc 1 complex and for photosynthetic growth of Rhodobacter capsulatus Proc. Natl. Acad. Sci. U. S. A. 88 1991 492 496 1846443 27 Czapla M. Borek A. Sarewicz M. Osyczka A. Fusing two cytochromes b of Rhodobacter capsulatus cytochrome bc 1 using various linkers defines a set of protein templates for asymmetric mutagenesis Protein Eng. Des. Sel. 25 2012 15 25 22119789 28 Valkova-Valchanova M.B. Saribas A.S. Gibney B.R. Dutton P.L. Daldal F. Isolation and characterization of a two-subunit cytochrome b-c 1 subcomplex from Rhodobacter capsulatus and reconstitution of its ubihydroquinone oxidation (Qo ) site with purified Fe-S protein subunit Biochemistry 37 1998 16242 16251 9819216 29 Cieluch E. Pietryga K. Sarewicz M. Osyczka A. Visualizing changes in electron distribution in coupled chains of cytochrome bc 1 by modifying barrier for electron transfer between the FeS cluster and heme c 1 Biochim. Biophys. Acta 1797 2010 296 303 19917265 30 Sarewicz M. Borek A. Daldal F. Froncisz W. Osyczka A. Demonstration of short-lived complexes of cytochrome c with cytochrome bc 1 by EPR spectroscopy: implications for the mechanism of interprotein electron transfer J. Biol. Chem. 283 2008 24826 24836 18617515 31 Davidson E. Ohnishi T. Atta-Asafo-Adjei E. Daldal F. Potential ligands to the [2Fe2S] Rieske cluster of the cytochrome bc 1 of Rhodobacter capsulatus probed by site-directed mutagenesis Biochemistry 31 1992 3342 3351 1313292 32 Osyczka A. Dutton P.L. Moser C.C. Darrouzet E. Daldal F. Controlling the functionality of cytochrome c 1 redox potentials in the Rhodobacter capsulatus bc1 complex through disulfide anchoring of a loop and a B-branched amino acid near the heme-ligating methionine Biochemistry 40 2001 14547 14556 11724568 33 Osyczka A. Moser C.C. Daldal F. Dutton P.L. Reversible redox energy coupling in electron transfer chains Nature 427 2004 607 612 14961113 34 Shinkarev V.P. Crofts A.R. Wraight C.A. The electric field generated by photosynthetic reaction center induces rapid reversed electron transfer in the bc 1 complex Biochemistry 40 2001 12584 12590 11601982 35 Wikström M.K.F. Berden J.A. Oxidoreduction of cytochrome b in the presence of antimycin Biochim. Biophys. Acta 283 1972 403 420 4346389 36 von Jagow G. Ljungdahl P.O. Graf P. Ohnishi T. Trumpower B.L. An inhibitor of mitochondrial respiration which binds to cytochrome b and displaces quinone from the iron-sulfur protein of the cytochrome bc1 complex J. Biol. Chem. 259 1984 6318 6326 6327677 37 Mulkidjanian A.Y. Mamedov M.D. Semenov A.Y. Shinkarev V.P. Verkhovsky M.I. Drachev L.A. Partial reversion of the electrogenic reaction in the ubiquinol Cytochrome c 2 -oxidoreductase of Rhodobacter sphaeroides chromatophores under neutral and alkaline conditions FEBS Lett. 277 1990 127 130 2176609 38 Robertson D.E. Dutton P.L. The nature and magnitude of the charge-separation reactions of ubiquinol cytochrome c 2 oxidoreductase Biochim. Biophys. Acta 935 1988 273 291 2844257 39 Sarewicz M. Dutka M. Pintscher S. Osyczka A. Triplet state of the semiquinone-Rieske cluster as an intermediate of electronic bifurcation catalyzed by cytochrome bc 1 Biochemistry 52 2013 6388 6395 23941428 40 Kauten R. Tsai A.-L. Palmer G. The kinetics of reduction of yeast complex III by a substrate analog J. Biol. Chem. 262 1987 8658 8667 3036818 41 De Vries S. Albracht S.P.J. Berden J.A. Marres C.A.M. Slater E.C. The effect of pH, ubiquinone depletion and myxothiazol on the reduction kinetics of the prosthetic groups of ubiquinol: cytochrome c oxidoreductase Biochim. Biophys. Acta 723 1983 91 103 6299337 42 Jackson J.B. Crofts A.R. The kinetics of light induced carotenoid changes in Rhodopseudomonas sphaeroides and their relation to electrical field generation across the chromatophore membrane Eur. J. Biochem. 18 1971 120 130 5540508 43 Mulkidjanian A. Mamedow M.D. Drachev L.A. Slow electrogenis events in the cytochrome bc 1 -complex of Rhodobacter sphaeroides FEBS 284 1991 227 231 44 Mulkidianian A. Junge W. Calibration and time resolution of lumenal pH-transients in chromatophores of Rhodobacter capsulatus following a single turnover flash of light: proton release by the cytochrome bc 1 -complex is strongly electrogenic FEBS Lett. 353 1994 189 193 7926049 45 Wraight C.A. Chance and design-proton transfer in water, channels and bioenergetic proteins Biochim. Biophys. Acta 1757 2006 886 912 16934216 46 Nabedryk E. Breton J. Coupling of electron transfer to proton uptake at the QB site of the bacterial reaction center: a perspective from FTIR difference spectroscopy Biochim. Biophys. Acta 1777 2008 1229 1248 18671937 47 Stowell M.H.B. McPhillips T.M. Rees D.C. Soltis S.M. Abresch E. Feher G. Light-induced structural changes in photosynthetic reaction center: implications for mechanism of electron-proton transfers Science 276 1997 812 816 9115209 48 Maróti Á. Wraight C.A. Maróti P. The rate of second electron transfer to QB - in bacterial reaction center of impaired proton delivery shows hydrogen-isotope effect Biochim. Biophys. Acta 1847 2014 223 230 25445317 49 Crofts A.R. Barquera B. Bechmann G. Guergova M. Salecedo-Hernandez R. Hacker B. Hong S. Gennis R.B. Structure and function in the bc 1 -complex of Rhodobacter sphaeroides Mathis P. Photosynth. From Light to Biosph. 1995 Kluwer Academic Publishers Dordrecht 493 500 50 Esser L. Elberry M. Zhou F. Yu C.-A. Yu L. Xia D. Inhibitor-complexed structures of the cytochrome bc 1 from the photosynthetic bacterium Rhodobacter sphaeroides J. Biol. Chem. 283 2008 2846 2857 18039651

PMC005xxxxxx/PMC5001788.txt

==== Front Nat CommunNat CommunNature Communications2041-1723Nature Publishing Group ncomms1138310.1038/ncomms1138327136393ArticleThe topography of mutational processes in breast cancer genomes Morganella Sandro 1Alexandrov Ludmil B. 234Glodzik Dominik 2Zou Xueqing 2Davies Helen 2Staaf Johan 5Sieuwerts Anieta M. 6Brinkman Arie B. 7Martin Sancha 2Ramakrishna Manasa 2Butler Adam 2Kim Hyung-Yong 8Borg Åke 5Sotiriou Christos 9Futreal P. Andrew 110Campbell Peter J. 2Span Paul N. 11http://orcid.org/0000-0002-1930-6638Van Laere Steven 12Lakhani Sunil R. 1314http://orcid.org/0000-0003-1879-2555Eyfjord Jorunn E. 15Thompson Alastair M. 1617Stunnenberg Hendrik G. 7van de Vijver Marc J. 18Martens John W. M. 6Børresen-Dale Anne-Lise 1920Richardson Andrea L. 2122Kong Gu 8Thomas Gilles 23Sale Julian 24Rada Cristina 24http://orcid.org/0000-0003-4898-5550Stratton Michael R. 2Birney Ewan 1Nik-Zainal Serena a2251 European Molecular Biology Laboratory, European Bioinformatics Institute, Wellcome Trust Genome Campus, Cambridgeshire CB10 1SD, UK2 Wellcome Trust Sanger Institute, Cambridge CB10 1SA, UK3 Theoretical Biology and Biophysics (T-6), Los Alamos National Laboratory, Los Alamos NM 87545, New Mexico, USA4 Center for Nonlinear Studies, Los Alamos National Laboratory, Los Alamos NM 87545, New Mexico, USA5 Division of Oncology and Pathology, Department of Clinical Sciences Lund, Lund University, Lund SE-223 81, Sweden6 Department of Medical Oncology, Erasmus MC Cancer Institute and Cancer Genomics Netherlands, Erasmus University Medical Center, Rotterdam 3015CN, The Netherlands7 Radboud University, Faculty of Science, Department of Molecular Biology, 6525GA Nijmegen, The Netherlands8 Department of Pathology, College of Medicine, Hanyang University, Seoul 133-791, South Korea9 Breast Cancer Translational Research Laboratory, Université Libre de Bruxelles, Institut Jules Bordet, Bd de Waterloo 121, B-1000 Brussels, Belgium10 Department of Genomic Medicine, UT MD Anderson Cancer Center, Houston, Texas 77230, USA11 Department of Radiation Oncology, and department of Laboratory Medicine, Radboud university medical center, Nijmegen 6525GA, The Netherlands12 Translational Cancer Research Unit, GZA Hospitals Sint-Augustinus, Wilrijk, Belgium and Center for Oncological Research, University of Antwerp, Antwerp B-2610, Belgium13 Centre for Clinical Research and School of Medicine, University of Queensland, Brisbane, Queensland 4059, Australia14 Pathology Queensland, The Royal Brisbane and Women's Hospital, Brisbane, Queensland 4029, Australia15 Cancer Research Laboratory, Faculty of Medicine, University of Iceland, 101 Reykjavik, Iceland16 Department of Breast Surgical Oncology, University of Texas MD Anderson Cancer Center, 1400 Pressler Street,Houston, Texas 77030, USA17 Department of Surgical Oncology, University of Dundee, Dundee DD1 9SY, UK18 Department of Pathology, Academic Medical Center, Meibergdreef 9, 1105 AZ Amsterdam, The Netherlands19 Department of Cancer Genetics, Institute for Cancer Research, Oslo University Hospital, The Norwegian Radium Hospital, Oslo 0310, Norway20 K.G. Jebsen Centre for Breast Cancer Research, Institute for Clinical Medicine, University of Oslo, Oslo 0310, Norway21 Department of Pathology, Brigham and Women's Hospital, Boston, Massachusetts 02115, USA22 Department of Cancer Biology, Dana-Farber Cancer Institute, Boston, Massachusetts 02215, USA23 Synergie Lyon Cancer, Centre Léon Bérard, 28 rue Laënnec, Lyon Cedex 08, France24 MRC Laboratory of Molecular Biology, Francis Crick Avenue, Cambridge CB2 0QH, UK25 East Anglian Medical Genetics Service, Cambridge University Hospitals NHS Foundation Trust, Cambridge CB2 9NB, UKa snz@sanger.ac.uk02 05 2016 2016 7 1138301 12 2015 18 03 2016 Copyright © 2016, Nature Publishing Group, a division of Macmillan Publishers Limited. All Rights Reserved.2016Nature Publishing Group, a division of Macmillan Publishers Limited. All Rights Reserved.This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/Somatic mutations in human cancers show unevenness in genomic distribution that correlate with aspects of genome structure and function. These mutations are, however, generated by multiple mutational processes operating through the cellular lineage between the fertilized egg and the cancer cell, each composed of specific DNA damage and repair components and leaving its own characteristic mutational signature on the genome. Using somatic mutation catalogues from 560 breast cancer whole-genome sequences, here we show that each of 12 base substitution, 2 insertion/deletion (indel) and 6 rearrangement mutational signatures present in breast tissue, exhibit distinct relationships with genomic features relating to transcription, DNA replication and chromatin organization. This signature-based approach permits visualization of the genomic distribution of mutational processes associated with APOBEC enzymes, mismatch repair deficiency and homologous recombinational repair deficiency, as well as mutational processes of unknown aetiology. Furthermore, it highlights mechanistic insights including a putative replication-dependent mechanism of APOBEC-related mutagenesis. Mutational signatures provide evidence of the mechanism of action of a given mutagen and are found in cancer cells. Here, using 560 breast cancer genomes, the authors demonstrate that mutational signatures are frequently associated with genomic architecture including nucleosome positioning and replication timing. ==== Body Correlations between the density of somatic mutations and various features of genomic structure and function have customarily been performed on aggregated cancer mutations across many cancer types123456789. These reports show similar general conclusions, for example, that substitution mutations are enriched in genomic regions that undergo replication late while rearrangements are enriched in early replicating regions123456789 or that specific genomic landmarks like chromatin organization are variably associated with mutation distribution910. The interpretation of these historic analyses is, however, complicated, because somatic mutations do not arise from a single, universal mutagenic process. They occur due to numerous mutational processes that have occurred throughout the lifetime of the cancer patient11121314 and may be distinct in different tissues. Consider analyses based on simple substitution classes across multiple cancers. C>T transitions, for example, could arise from disparate mutational processes including deamination of methylated cytosines, deamination by APOBEC cytidine deaminases, exposure to ultraviolet irradiation or mismatch repair (MMR) deficiency. The interpretation of how C>T mutations are distributed relative to any genomic landmark would thus be limited by the complexity of mutational processes that contribute to C>T mutations. In addition, previous analyses commonly combined data across several cancer types with diverse tissues of origin. However, exposures to DNA-damaging agents are likely to be different between tissues (for example, ultraviolet damage occurs in skin but not colorectal tissue) and DNA repair pathways may behave differently in cells of different organs. Moreover, replicative, transcriptional and chromatin dynamics may be distinct from one tissue to another, further hampering interpretation of such aggregated somatic mutation data10. Each mutational process will leave its own specific pattern on the genome or mutational signature11 regardless of whether it arose as a pre-neoplastic process or post-malignant transformation. Recent advances in the mathematical extraction of mutational signatures14 from cancer sequences have led to the discovery of 21 such signatures in 30 different cancer types14. In a recent article of 560 highly curated whole-genome sequenced (WGS) breast cancers15, we extracted 12 base substitution mutational signatures from 3,479,652 base substitutions (signatures 1, 2, 3, 5, 6, 8, 13, 17, 18, 20, 26 and 30). These signatures were based on a 96-mutation classification that incorporates the base substitution type (expressed as the pyrimidine of a mutated Watson–Crick base pair, C>A, C>G, C>T, T>A, T>C, T>G) and the immediate flanking sequence context of the mutated base (four possible 5′ and four possible 3′ bases)1114. We also analysed 77,695 rearrangements that were classified according to rearrangement type (deletions, tandem duplications, inversions and translocations), size (range 1 kilobase to >1 Mb) and whether they were focal or genomically dispersed, to extract six novel rearrangement signatures (RS1–RS6)15. These had different predominating features including being mainly characterized by tandem duplications (RS1 and RS3), deletions (RS5), clustered rearrangements (RS2, RS4) or translocations (RS2). In addition, 371,993 indels were categorized into two distinct signatures. ‘Repeat-mediated' deletions share the identical motif as a flanking polynucleotide repeat tract, are small (<3 bp) and arise from erroneous repair of insertion–deletion loops at polynucleotide tracts, the onus of post-replicative MMR16. In contrast, microhomology-mediated deletions show homology of several nucleotides between the start of the deletion and the flanking sequence of the deletion junction. They are usually larger (≥3 bp) than repeat-mediated deletions and are associated with repair by microhomology-mediated end joining mechanisms. The significance of these signatures is clear. They are a proxy for the biological processes that have gone awry in breast tissue (see Table 1 for summary of signatures, their characteristics and putative aetiologies). Some associations include homologous recombination (HR) repair deficiency with signatures 3 and 8, microhomology-mediated indels, RS1, RS3 and RS5, putative activity of the APOBEC family of cytidine deaminases with signatures 2 and 13, MMR deficiency with signatures 6, 20 and 26 and an excess of repeat-mediated indels and deamination of methylated cytosines with signature 1. Aetiologies of the remaining signatures (signatures 5, 17, 18, 30; RS2, RS4 and RS6) are currently unknown (Table 1). This set of 560 breast cancers is the largest cohort of WGS cancers of a single tissue-type to date providing an exceptional opportunity to gain insights into mutagenic processes of a specific tissue. We thus set out to comprehensively explore how mutation signatures in human breast cancers are influenced by genomic architecture. Critically, we were able to assign probabilistic estimates for individual somatic mutations to each mutational signature for every sample. Thus, by studying the genomic distribution of mutations as mutational signatures, we are able to interpret how mutagenic processes in breast tissues are influenced by cellular activities such as replication, transcription or by physical features like nucleosome occupancy. Results Diverse temporal relationships with replication DNA replication begins at origins (or near clusters of origins) of replication and propagates bidirectionally from each starting point1718 with some regions of the genome copied sooner (early replicating) than others (late replicating)19. Using replication-sequencing (Repli-Seq)20 data generated from the breast cancer cell line, MCF-7 (ref. 20), early- and late-replicating regions were determined (Supplementary Fig. 1a–b) and relationships between mutations attributed to each signature and replication time were explored (Fig.1, Supplementary Fig. 1c–e). Base-substitution signatures 1, 2, 3, 5, 8, 17, 18, 20 and 26 showed increases in mutation density from early to late replication, in keeping with previously described observations on aggregated substitutions. However, each had a distinctive gradient (Fig. 1b, Supplementary Fig. 1c–e) underscoring the individuality of each signature. In contrast, signature 6, 13 and 30 showed the unexpected tendency of relatively constant mutation densities through all replication time domains. All six rearrangement signatures (RS1–RS6) from the 560 breast cancers15 were enriched in early replication. However, the gradient of change from early to late replication was variable between them (Fig. 1c). There was an approximately twofold reduction in rearrangement frequencies between the earliest and latest replication domains for RS1, RS3, RS4 and RS6. In contrast, RS2 and RS5 had flatter gradients with a greater proportion of rearrangements found in late replication domains than the other rearrangement signatures. Somatic deletions were generally enriched late in replication. Repeat-mediated deletions showed a steep gradient with more mutations in late-replication time domains. Ten cancers with overwhelming indel mutagenesis (range 2,535–66,764) associated with MMR deficiency demonstrated a particularly steep gradient. In contrast, microhomology-mediated deletions demonstrated a gradual slope of increasing frequency towards late replication domains (Fig. 1d). Thus, the signature-based approach permits higher resolution observations of distinctive variation between different mutational processes including behaviours different from those found when all mutations are considered together. Direction of replication influences mutational distribution Replication fork migration from early to late replicative regions generates replicative strands that act as templates for DNA synthesis in a continuous and discontinuous manner, respectively18 (Supplementary Fig. 2a). Through knowing the direction of replication fork migration relative to the p-to-q orientation of the genome, transition zones could be assigned to p-to-q leading or p-to-q lagging replicative strands, respectively (Methods section, Supplementary Fig. 2b–d). We were conservative in our assignments excluding the first and last 25 kb of the latest replication domains and discarding regions of 10 kb or less indicative of where potential replicative strand switches could have occurred2122 (Methods section). We thus explored whether the direction of replication influenced mutational processes through differences in mutation distribution between replicative strands (Fig. 2a,b, Supplementary Fig. 2e–i). The level of asymmetry between strands is referred to as strand imbalance. A strand imbalance of 20% implies that one strand has 20% more mutations than the other (for example, for every 100 mutations on one strand, there are 120 on the other). We found that the level of asymmetry was different between the various base substitution signatures (Fig. 2b, Supplementary Table 1). Signatures 2, 13 and 26 exhibited strong replicative strand asymmetries with imbalances >30% for each of the signatures (P value <2.2e−16); signatures 1, 3, 5, 6, 17, 18 and 20 had weaker asymmetries (P value <2e−4 and strand imbalance <13%) and signatures 8, and 30 did not exhibit asymmetries of distribution between replication strands (P value >0.01 and strand imbalance <3%). Asymmetric mutation distribution between replicative strands was thus observed more markedly for a variety of different biological processes including putative APOBEC-related mutagenesis (signatures 2 and 13), and MMR deficiency (signatures 6, 20 and 26—all with different biases), implying that the direction of travel of a replication fork can influence somatic mutation accumulation in diverse mutational mechanisms. APOBEC deamination of cytosine to uracil (C>U) is thought to initiate mutations of signatures 2 and 13 (refs 11, 23, 24, 25). Knowledge of which of the Watson–Crick base pair is targeted by the APOBEC enzyme enables insight into the preferred replicative strand and indicates that APOBEC-related mutagenesis occurs at a higher rate on the p-to-q lagging replicative strand. APOBECs require single-stranded DNA as a substrate for cytosine deamination2627 and thus the replication process itself could provide physiological opportunities for mutagenesis. Why APOBEC mutagenesis favours the lagging strand is unclear but could be due to differential availability of single-stranded DNA between leading and lagging strands for APOBEC deamination. Subtype heterogeneity in breast cancer and mutation distribution A variety of breast cancer subtypes exist and these have been historically classified according to transcriptomic profiles. We sought to understand whether typical classifications such as oestrogen receptor subtype (positive or negative) or putative cell-of-origin (basal of luminal A/B) showed differences in the behaviour of mutational signatures, as this could confound the interpretation of our analyses. We found that mutational signature relationships to replication time (Supplementary Fig. 1g–k) and to replication strand (Supplementary Fig. 2e–i) were highly similar regardless of whether breast cancers were oestrogen receptor positive or negative or whether they were basal or luminal. Therefore, heterogeneity of the main breast cancer subtypes does not appear to influence the distribution of mutation signatures, suggesting that mutational processes behave relatively similarly in cells from the same tissue. It remains to be seen, however, whether mutational signatures will differ in their distributions when cancers of different tissue types are contrasted to one another. Processive mutagenesis Previous work28 showed that APOBEC-related signatures 2 and 13 demonstrated strand-coordinated mutagenesis where pairs of adjacent mutations of the same reference allele were observed on the same strand more frequently than expected28. The underlying reason for this observation is unknown. Here we extend the strand-coordination analysis to identify long stretches of 10 or more successive mutations occurring on the same DNA strand (that is, successive mutations may be C>T…C>T…C>T or G>A…G>A…G>A but not C>T…G>A…C>T)28 for all mutation signatures. We found that long processive groups were features of signatures 2, 13, 6, 26 and 17 and were particularly over-represented in signature 13. 157 such processive groups were identified in 27 breast cancers, 76% of these from signature 13, which also had the longest processive group containing 19 point mutations (Fig. 2c, Supplementary Table 2). The longest processive stretch of mutations covered ∼1.1  Mb although most processive stretches were tens to hundreds of kilobases in length (Supplementary Table 3). This suggests that mutational processes generating processive mutations can do so for remarkably long stretches, perhaps suggesting that long stretches of single-stranded DNA exist in cells and/or that individual proteins track one of the two DNA strands over long distances. Transcriptional strand biases Base substitutions falling within the footprint of a gene, corresponding to ∼40% of the human genome, were classified according to whether they were on the ‘transcribed' strand (the non-coding strand), which forms the template for the primary mRNA transcript, or the ‘non-transcribed' strand112930 (Fig. 2a,b, Supplementary Fig. 3). Base-substitution signatures of breast cancer showed variation in transcriptional strand asymmetries. Signatures 2, 3, 5, 8, 13, 18 and 26 showed some transcriptional strand bias (P value <0.01 and strand imbalance up to 10.7%), while signatures 1, 6, 17, 20 and 30 showed no asymmetrybetween transcriptional strands (P-value >0.01). While transcriptional strand bias can result from the involvement of transcription-coupled nucleotide excision repair that often acts on large, DNA distorting adducts29, the biases detected here in breast tissue are novel, and not easily ascribed to known double-helix distorting agents. These observations imply that other currently undefined transcription-coupled DNA damage and/or repair processes may be at play313233 in breast tissue. The results also suggest that DNA replication has overall a stronger effect on the mutational landscape than transcription (Fig. 2a,b, Supplementary Table 1). Mutation signatures and nucleosome occupancy Finally, we examined how mutations due to different mutational signatures were distributed relative to nucleosome positions (Fig. 3, Supplementary Fig. 4). Nucleosomes consist of an octamer of histone proteins wrapped by ∼147 bp of ‘core DNA' and are separated from the next nucleosome by ∼60–80 bp of ‘linker DNA' sequence18. Reference regions indicating stable nucleosome occupancy were defined based on MNase experiments performed on an ENCODE line, K562 (refs 4, 34). Variant-to-nucleosome distances were calculated for each variant in each signature (Fig. 3a,b). A randomization was performed correcting for systematic variation in AT/CG content of mutational signatures and of core and linker DNA regions, and the distribution of observed mutations (Fig. 3b, coloured lines) was compared with that of the randomization (Fig. 3b, grey lines) for each signature. Signature 17 and 18 mutations demonstrated enrichment at nucleosome core DNA sequences and showed a marked periodicity at ∼200–220 bp intervals, the approximate internucleosome distance1835, contrasting with their predicted distributions through simulations and distinctly different when compared with all other signatures. By contrast, signature 26 mutations, associated with MMR deficiency, were more frequent at linker DNA sequences distant from nucleosomes636 (Fig. 3b). We also observed that repeat-mediated deletions, particularly those from MMR-deficient cancers were enriched at linker sequences (Fig. 3a) in keeping with previous reports37. Thus MMR deficiency increases the likelihood of indels as well as base substitutions occurring between nucleosomes. Intriguingly, it appears to do so in only one of the three distinct MMR-associated substitution signatures. Base-substitution signatures 1, 2, 3, 5, 6, 8, 13, 20 and 30 did not exhibit a different distribution to that expected from the randomization experiment suggesting that nucleosome positioning does not influence their underlying mutational processes. Discussion This signature-based analysis is a novel way of visualizing in vivo mutagenesis providing a powerful means of revealing the natural experiments that occur in human cells. We find that different somatic mutational signatures demonstrate distinct partialities in replicative and transcriptional strands have variable profiles across replication time and are differentially influenced by physical genomic attributes such as nucleosome positioning (Table 1). The observed profiles are out-of-keeping with the profiles expected through randomization experiments that correct for per-sample mutation burden, AT/GC content of each signature and AT/GC content of the genome. Reference coordinates for replication-related analyses were generated from Repli-Seq data for the cell line, MCF-7. Of publicly available Repli-Seq data sets, this ductal breast carcinoma cell line was selected because it was most closely related to the breast cancers in terms of tissue-of-origin. A large proportion of the earliest (average 59.8%, range 51.7–69%) and latest (average 77.9%, range 52.6–85%) replication domains are shared between MCF-7 and all other ENCODE cell lines (Supplementary Fig. 5), and analytical comparisons suggest that strong biological signals such as the replicative strand bias for signatures 2 and 13, remains convincingly consistent with only minor variation between cell lines (Supplementary Fig. 5, Supplementary Table 4). Thus, although a cell line will never be perfectly representative of what occurs in vivo in any cancer, MCF-7 appears to be a reasonable proxy for generating reference coordinates for assessing mutational distribution of breast cancers. The signature-based analysis enables us to distinguish between similar mutational processes and provides mechanistic insights. Here we show that replication may be a source of single-stranded DNA over relatively long genomic distances, providing the potential for APOBEC-related deamination damage that initiates signatures 2 and 13. However, mathematical extractions differentiate these signatures and suggest that signature 2 is composed predominantly of C>T transitions, while signature 13 comprises mainly C>G transversions1114. These differences have been hypothesized to be due to the subsequent reparative step1113: it was postulated that C>G signature 13 transversions were fixed via uracil-N-glycosylase (UNG)- and REV1-polymerase (REV1)-dependent mechanism in base excision repair, while C>T signature 2 transitions were formed by replicative polymerases exerting the ‘A' rule (the preferential insertion of adenine opposite non-informative templates such as abasic sites or uracils38). Here analyses relating to replication dynamics show that first, signatures 13 and 2 have distinct replication strand biases, second, signature 13 has consistently longer processive groups than signature 2 and third, signature 13 mutations are more frequent early in replication than signature 2. We postulate that UNG/REV1-dependent uracil processing generating signature 13 mutations occurs earlier in replication, thus has more time and is more likely to process successive uracils. By contrast, the reparative process that generates signature 2 mutations acts on remaining incompletely processed uracils and/or abasic sites including those that are left at the end of the replication cycle, leading to the observed distribution of higher frequencies late in replication. Thus, we suggest a replication-related model of APOBEC mutagenesis: although the replication process itself springs forth opportunities for APOBEC-related DNA damage, it is possible that the variation in repair across replication time results in these disparate signatures (Fig. 4). Signatures 6, 20, 26 and an excess of repeat-mediated indels are all associated with defective MMR23. They are rare signatures (only 1.7%) in breast cancers, often found together in the same breast tumours and are overwhelming when they occur. However, they exhibit differing relationships with respect to replication time and direction, transcription and nucleosome occupancy (Table 1). Intriguingly, only one of the three MMR-related substitution signatures exhibits a distinctive distribution relative to nucleosome occupancy, an observation that would not have been appreciable without this signature-based approach applied to WGS data. Three of six rearrangement signatures (RS1, RS3 and RS5) are associated with defects in HR and are enriched early in replication. Microhomology-mediated indels and substitution signatures 3 and 8 are also associated with HR defects but are enriched late in replication. These signatures are often found together (albeit in differing quantities) in individual patients15 and likely represent compensatory methods of double-strand break (DSB) repair in the face of defective HR. Perhaps, back-up recombination-based repair pathways39404142 are more likely to be invoked early in replication because DSBs encountered early in S-phase are poorly tolerated42434445. In contrast, microhomology-mediated deletions represent the outcome of DSB resolution through microhomology-mediated end joining mechanisms, reflecting a contingency route for DSBs that have not been repaired successfully by recombination strategies4445 earlier in replication. Likewise, base substitution signatures 3 and 8 could similarly represent the outcome of back-up error-prone translesion synthesis activity in HR-deficient cancers46. Regardless, these different compensatory signatures exhibit distinctive behaviours across replication time, painting the physiology of HR deficiency as complex mutagenesis that has an instantly recognizable whole-genome profile with potential clinical relevance15. To the best of our knowledge, we provide the first systematic signature-based characterization of the genomic distribution of all classes of somatic mutations in human cancer. The power provided by large numbers of WGSs of a single cancer type affords a higher resolution perspective on the topography of biological processes underlying mutagenesis in breast tissue. We emphasized how detailed analyses help showcase the mechanistic contribution of replication dynamics to specific mutational processes (for example, APOBEC-related signatures 2 and 13). We also highlighted how multiple forms of DNA repair have an impact on mutation distribution leaving complex but distinctive global genomic profiles. Finally, the signature-based genomic variation seen here drives home a fundamental point regarding genomic analyses forthwith: statistical models involving mutability cannot assume uniform genomic mutation rates and must consider signature-dependent variation as a factor in all future analyses. Methods Data set All short-read sequencing data were aligned on the GRCh37/hg19 assembly15 and somatic substitutions, indels and rearrangements called and curated as previously described15. High-quality mutation calls were parsed through non-negative matrix factorization11121447 to extract mutation signatures. In all, twelve base substitution signatures and six rearrangement signatures were identified15. Indels were classified according to the properties of the breakpoint junctions into two signatures. Insights into mechanisms generating deletions are more certain than that of insertions, thus our analyses were restricted to deletions. The non-negative matrix factorization-based mutational signatures analysis revealed the substitution signatures of 12 mutational processes operative in breast cancer15. Furthermore, the analysis provided the number of somatic mutations assigned to each of these 12 signatures in each of the examined breast cancers (exposures). Using the patterns of the extracted mutational signatures and their contributions in each sample, we were able to assign an a posteriori probability for any individual substitution to be generated by any of the 12 mutational signatures in a given sample. The posterior probability for a given substitution with a trinucleotide context, k, to be generated by the mutational signature n in the sample g, (kng), was computed as the exposure of this sample to the signature n (eng), multiplied by the probability of the signature n to generate this particular mutation with trinucleotide context k (pkn). The a posteriori probabilities were then normalized to sum to 1 by using the number of mutations observed in the sample . Different methodologies were considered to associate the substitutions with the mutational processes that generated them: (i) maximum likelihood (the signature associated with each mutation was the one having the highest probability), (ii) maximum likelihood with probability threshold (same of maximum likelihood but here signatures with an a posteriori probability lower or equal to 0.5 were filtered out), (iii) belief propagation (the a posteriori probabilities ϑkng were propagated in the downstream analyses). We used the maximum likelihood approach to perform the analyses described in the main manuscript. This choice was motivated by the fact that this approach could be consistently applied to all downstream analyses and could be used to perform statistical tests. For example, belief propagation could not be used for the analysis of processive groups and it was not suitable for statistical tests requiring integer values. In addition, the thresholding method tended to result in reduced power. Regardless, the strong biological signals from analyses of particular signatures such as signatures 2 and 13 were robust and reproducible across all three approaches (to compare the different methodologies, please see results from the thresholding and belief propagation approach in Supplementary Fig. 6). Replication analyses Reference coordinates for replication landmarks were inferred from Repli-seq data obtained from the ENCODE project48 (https://www.encodeproject.org/). Cell lines were first isolated into six cell cycle fractions of newly replicated DNA (G1/G1b, S1, S2, S3, S4 and G2) and each fraction was sequenced. To visualize genome-wide replication patterns as a continuous function, percentage-normalization of sequencing tags was followed by a wavelet-smoothed transformation. The majority of origins do not fire as a part of a clear, deterministic programme, instead origin firing occurs both individually and as clusters495051. Replication domains were defined using Repli-seq signal: peaks (local maxima) in the smoothened profile correspond to replication initiation zones, while valleys (local minima) correspond to replication termination zones48. Replication time domains were modelled on conservatively defined transition zones in DNA replication time data. Repli-seq data were split into deciles with each segment containing exactly 10% of the observed signal. AT/CG content of the deciles were variable (Supplementary Fig. 1a), and the genome-wide distribution of the deciles was heterogeneous (Supplementary Fig. 1b). All analyses related to replication time domains were corrected for genomic size. In particular, in each decile a mutation density was computed as the total mutation count in each decile divided by the number of attributable bases (excluding ‘N's) contained in the relevant decile. In order for gradients to be comparable between signatures (given the variation in mutation rate between signatures), counts were then normalized to between 0 and 1. Results of analyses with absolute counts can also be found in Supplementary Fig. 1c–e. Finite difference approximations of second and first derivatives were used to identify Repli-seq signal local maxima (f ′′(x)<0) and local minima (f ′′(x)>0) corresponding to potential origin firing sites, and then to distinguish between leading (f ′(x)<0) and lagging (f ′(x)>0) strand, respectively (Supplementary Fig. 2a–b). Derivative functions were defined in agreement with p and q arm chromosome orientation. We named the replication strand as p2q leading and p2q lagging. To remain conservative in downstream assignments5253, we removed the last 25 kb of the latest zones of the replicating domains. We focused on long transitions between early and late replicative domains, discarding ambiguous mini-peaks or valleys that were <10 kb in length. It was possible to assign replication domains in 2,414,428,423 bp of the genome. The median length of assigned replication strand was 136,001 bp and the mean length was 196,907 bp, safely and conservatively within the limits described by recent alternative methods, including DNA combing54. Derived p2q leading and p2q lagging strands were comparable in genomic footprint, AT/CG content and in amount of transcribed/non-transcribed regions (Supplementary Fig. 2c–d). To investigate asymmetry relative to replication strands, all base substitutions were first described in the pyrimidine context and then orientated with respect to the relevant strand. Choice of reference cell line for mutational distribution Replication time-related coordinates for the main analyses reported in the manuscript were generated from MCF-7 cell line. This choice was motivated by the fact that this is a ductal breast carcinoma cell line and most closely represented our data set of breast cancers. Note that across the 14 cell lines available from ENCODE, on average 59.8% of the earliest replication time domain is shared between MCF-7 and each of the other cell lines (range 51.7–69%), and average 77.9% of the latest replication time domain (range 52.6–85%; Supplementary Fig. 5a). In other words, large swathes of the earliest and latest replication time domains are identical between MCF-7 and other cell lines. To further contrast the cell lines to identify the most appropriate source of reference coordinates for our analyses, we analysed the mutation density trend across the cell cycle for all cell lines where data were available from ENCODE. Cell lines showing a consistent increase of aggregated mutation density going from early to late replicative regions, should be preferred over the cell lines that exhibited a random trend. For each cell line, we extracted replication time deciles and counted the number of mutations falling in each of these domains. These counts were then corrected for the genomic size of each domain. In this way, we obtained the mutation density mij for each decile (i represents the i-th cell line, and j=1,2,..,10 the decile with 1 and 10 being the earliest and the latest decile, respectively). The mutation densities were ordered across replication time () to capture the overall trend of mutation accumulating across the cell cycle. Pearson's correlations were separately applied to assess the relationship between the distribution of mutations across replication time domains (expecting an increasing trend progressing through to late replication). The Pearson's test showed low P values for strong correlations across replication time. On the contrary, less-significant P values were observed for distributions that were poorly correlated and showed more randomly distributed mutations across replication time domains. Results of this comparison are showed in Supplementary Fig. 1f. Transcriptional strand characterization The nucleotide sequence of the primary mRNA transcript is identical to the sense/non-template/non-transcribed strand except that U replaces T, and is complementary to that of the anti-sense/template/transcribed strand (Supplementary Fig. 3a). All mutations were called on the + strand of the reference genome, were placed into the ‘pyrimidine' context and noted if so. Transcriptional strand was assigned for each pyrimidine-based mutation (Supplementary Fig. 3b for explanation of orientation). Regions of the genome with protein coding genes were used to assign transcriptional strands. On the total of 20,305 protein coding genes, 10,301 (677,912,252 bp) were on the + strand and 10,004 (646,112,188 bp) were on the − strand, respectively. Computing replication and transcription strand ratios All base substitution mutations were described in the pyrimidine context and orientated with respect to the replication and transcription strand (for example, an A>C observed on the p2q leading strand was counted as a T>G on the lagging). Given the broadly random orientation of both transcriptional direction of genes in the genome and replication strand (Supplementary Fig. 2c–d), our null hypothesis is that all the signatures would have a 50:50 distribution with respect to transcriptional or replicative strands. To ensure this hypothesis is robust to other features in the genome effecting mutation rates, such as local base composition, we randomized the position of the observed mutations keeping the local triplet context (see statistical analysis section below for more details about the approach used to generate the simulated data). The random simulations showed no bias towards either replication strand or transcription strand (all P values >0.05, binomial test). In contrast, many signatures showed striking bias either in replication strand or transcription strand with the deviating signatures showing strong statistical support (all P values <2e−16 binomial test; Fig. 1b and Supplementary Table 1). Supplementary Figure 2e and 2f show the overall distribution of the ratios for the six mutation classes and for the 12 signatures, respectively, across all the 560 samples (each dot represents a sample), with summary plots in Supplementary Fig. 2g–h. Note that the interpretation of transcriptional or replicative strand bias for six classes of base substitutions is restricted by the complexity of mutational mechanisms that contribute to each base substitution class (Fig. 2a,b). For example, C>T transitions exhibit lagging replicative strand bias (Supplementary Fig. 2g) but are components of signatures 1 (due to deamination of methylated cytosines), 2 (APOBEC related), 6, 20, 26 (MMR deficiency) and 30 (unknown aetiology). Hence, our analyses concentrate on exploring extracted base substitution signatures (Supplementary Fig. 2h) instead of the six classes of base substitutions (Fig. 2a, Supplementary Fig. 2g), but are provided for information. Processivity Processive groups were defined separately for each sample. Kataegis mutations were excluded from these analyses first as they have been previously highlighted to demonstrate strand-coordinated mutagenesis thus inclusion of kataegis would produce a biased signal. Adjacent somatic mutations were considered to be part of the same processive group if (i) they were associated with the same signature and (ii) they had the identical reference allele (complementary mutations were not considered processive e.g., A>G, T>C, A>G=non-processive while A>G, A>G, A>G or T>C, T>C, T>C=processive). The average mutation density for processive groups for each signature is provided in Supplementary Table 2. A total of 426,066 processive groups were identified. We characterized each group by using the number of substitutions involved in each of them (processive group length). Group length ranged from 2 to 19 (Supplementary Table 2). Results shown in Fig. 1c were generated by counting the number of groups having the specified length for each mutation signature. For visualization purposes, the absolute counts were log10 transformed. We used a set of simulated mutations to understand whether the observed processive behaviour was the consequence of the idiosyncrasies of individual samples (that is, the possibility of observing long processive groups in samples containing many mutations belonging to one signature may be higher than samples having fewer mutations or having mutations belonging to several different signatures). We generated a null distribution by using 100 random data sets that took the number and type of mutations relevant to each signature into consideration (more details on the approach used to generate the random simulations can be found in Relationships with nucleosome occupancy Section. To assess the probability of observing groups of a particular length, we compared the observed data to the null distribution. Let pnij be the number of processive groups of length i observed in the jth random dataset and associated with the signature n. We can compute the number of processive groups of length i observed across all the 100 simulated datasets for the signature n as , (J=1, …, 100), and we can assess the probability to observe a processive group of length L for the signature n as: . Bonferroni's correction was used to adjust for multiple testing. For each signature the replication strand ratio was computed by summing the mutations over all the groups having the specified length. Two-tailed binomial test was used separately on each group to assess the significance of the imbalance between leading and lagging strand, and 0.05 was used as the level of significance in this analysis. Note that groups with less than six mutations did not contain enough observations to obtain a significant P value from the two-tailed binomial test. Intriguingly, we also observed that all or nearly all mutations were on the same replicative strand within individual processive groups for signature 13. Indeed, lagging strand bias was ∼fourfold stronger for processive mutations than those that were not processive. The data implies that for signature 13, replication is not simply a source of single-stranded DNA, it permits processive deamination for exceptionally long genomic distances. Relationships with nucleosome occupancy Micrococcal nuclease sequencing (MNase-seq) data for the K562 cell line was obtained from the ENCODE project48. Please see http://genome.ucsc.edu/cgi-bin/hgTrackUi?db=hg19&g=wgEncodeSydhNsome for details of experiments and generation of nucleosome density signal maps. Although K562 is not a breast cancer cell line, our choice to use it for our analysis was motivated by several factors: it has been used in other research laboratories for similar analyses55, it is one of the two main reference lines archived in ENCODE with clear cell culture protocols for these experiments and it is the only cancer cell line available. To assess the relationships between signatures and nucleosome occupancy, we created a window of 2 kb centred around each mutation (within a signature), and obtained the nucleosome density signal observed within the 2 Kb window. We calculated the SUM of the signal observed across the window for all the mutations within a signature, and the number of mutations (COUNT) contributing to the signature. The average signal (y axis) is the SUM/COUNT for every position within the 2 kb window (Supplementary Fig. 4a–b). The nucleosome density signal distribution for K562 MNase data encompasses 575,649,742 loci, where the signal is a smoothed version of the total number of reads. The MNase signal has a skewed distribution with mode that lies in the interval 0.85–0.9, hence the averaged signal (accounting for all the mutations) for each signature lies in the region of 0.85–0.9. Note that every mutation that contributes towards a given signature is at a different genomic location and there are many thousands, or even tens or hundreds of thousands of mutations per signature. If mutations in a given signature bore no relationship to the position of nucleosomes, then when aggregated across thousands of mutations per signature, a flat line would be seen. However, if mutations in a particular signature were more frequent at core sequences, there would be a peak of nucleosome signal where the mutation is centred. If mutations were more frequent at linker sequences, there would be a trough. Other computational and statistical analyses All statistical analyses were performed in R (version 3.0.2): Pearson's correlation was performed with cor.test, binomial test was performed with binomial test, Bonferroni correction was performed by using p.adjust. Multivariate normal mixtures were computed by using normalmixEM function available as part of the R/CRAN mixtool package, an initial mixing proportion of 0.5 was used to compute three components (parameter lambda and k of the function, respectively). corrplot R package was used to generate Fig. 1c and Supplementary Fig. 1b and 5a. Data in bigWig format were preprocessed by using bigWigToBedGraph script. Perl EnsEMBL API (version 73) was used to extract the genome features of interest. bedtools (version 2.16.2) was used to identify intersection and union among genomic features, and to manipulate BED files. samtools (version 0.1.18) was used to extract subsequences from fasta files. Random mutations were generated in agreement with their flanking sequence context defined by the neighbouring bases immediately 5′ and 3′ to the mutated base and by the mutated base itself. We generated random simulations of the dataset obtained from the 560 breast cancers. We imposed the following constraints: (i) mutation class (ii) flanking sequence context (iii) overall mutation burden (iv) contribution of each mutation signature to each sample. To perform simulations of processive data, mutations for each signature were shuffled separately for each sample. Shuffled samples were then used to compute the processive groups. For each signature 100 simulations were used to compute the null distribution associated with the expected processive length associated. For each observed rearrangement we simulated both breakpoint junctions, and we kept the signature, and type (that is, translocation, inversion, deletion and tandem duplication) observed in the real dataset. Given a rearrangement we randomly picked the first breakpoint from one of the replication time deciles, then (i) if the rearrangement was not a translocation then we randomly pick the second breakpoint on the same chromosome of the first one (ii) if the rearrangement was a translocation we randomly picked also the second breakpoint without any constraint. Functional genomics experiments based on next-generation sequencing, such as Repli-seq and MNase-seq, often produce artefactual signals in certain regions of the genome. To filter out artefact-ridden regions that tend to show artificially high/low signals, we excluded from our analyses ENCODE blacklisted genomic regions (human, hg19/GRCh37): http://hgdownload.cse.ucsc.edu/goldenpath/hg19/encodeDCC/wgEncodeMapability/wgEncodeDacMapabilityConsensusExcludable.bed.gz. Additional information How to cite this article: Morganella, S. et al. The topography of mutational processes in breast cancer genomes. Nat. Commun. 7:11383 doi: 10.1038/ncomms11383 (2016). Supplementary Material Supplementary Information Supplementary Figures 1-6 and Supplementary Tables 1-4. This analysis has been performed on data from a project funded through the ICGC Breast Cancer Working group by the Breast Cancer Somatic Genetics Study (a European research project funded by the European Community's Seventh Framework Programme (FP7/2010-2014) under the grant agreement number 242006); the Triple Negative project funded by the Wellcome Trust (grant reference 077012/Z/05/Z) and the HER2+ project funded by Institut National du Cancer (INCa) in France (Grants No. 226-2009, 02-2011, 41-2012, 144-2008, 06-2012). The ICGC Asian Breast Cancer Project was funded through a grant of the Korean Health Technology R&D Project, Ministry of Health & Welfare, Republic of Korea (A111218-SC01). We would like to acknowledge the Wellcome Trust Sanger Institute Sequencing Core Facility, Core IT Facility and Cancer Genome Project Core IT team and Cancer Genome Project Core Laboratory team for general support. This research also used resources provided by the Los Alamos National Laboratory Institutional Computing Program, which is supported by the US Department of Energy National Nuclear Security Administration under Contract No. DE-AC52-06NA25396. Research performed at Los Alamos National Laboratory was carried out under the auspices of the National Nuclear Security Administration of the United States Department of Energy. S.M. is funded through the Breast Cancer Somatic Genetic Study (BASIS). L.B.A. is supported through a J. Robert Oppenheimer Fellowship at Los Alamos National Laboratory. D.G. is supported by the EU-FP7-SUPPRESSTEM project. A.L.R. is partially supported by the Dana-Farber/Harvard Cancer Center SPORE in Breast Cancer (NIH/NCI 5 P50 CA168504-02). M.S. was supported by the EU-FP7-DDR response project. C.S. is funded by FNRS (Fonds National de la Recherche Scientifique). G.K. is supported by National Research Foundation of Korea (NRF) grants NRF 2015R1A2A1A10052578. E.B. is funded by EMBL. SN-Z is a Wellcome Beit Fellow and personally funded by a Wellcome Trust Intermediate Fellowship (WT100183MA). S.N.-Z. and X.Z. also work under the auspices of the COMSIG Consortium, supported by a Wellcome Trust Strategic Award (101126/B/13/Z). We would like to acknowledge all members of the ICGC Breast Cancer Working Group, ICGC Asian Breast Cancer Project, and Oslo Breast Cancer Consortium (OSBREAC). Author contributions S.M., M.R.S., E.B., and S.N.-Z. designed the study, analysed the data and wrote the manuscript. D.G., X.Z., H.D. and J.S., performed curation and contributed towards genomic and copy number analyses. L.B.A., A.M.S., A.B.B., M.R., H.-Y.K. A.B., C.S., P.A.F., P.J.C., P.N.S., S.V.L., S.R.L., J.E.E., A.M.T., H.G.S., M.J. van de Vijver, J.W.M.M., A.-L.B.-D., A.L.R., G.K., G.T., J.S., C.R contributed towards analyses. S.Martin was project coordinator. A.B. contributed IT expertise. All authors discussed the results and commented on the manuscript. Figure 1 Distribution of all mutations across the cell cycle. Replication domains were identified by using conservatively defined transition zones in DNA replication time data. Data were separated into deciles, with each segment containing exactly 10% of the observed replication time signal. Normalized mutation density per decile is presented for early (left) to late (right) replication domains. (a) Aggregated distribution of mutations (green), rearrangements (purple) and indels (orange) across the cell cycle. (b) Distribution of the 12 base substitution signatures across the cell cycle. Dashed grey lines represent the predicted distribution of mutations for each signature based on simulations that take into account mutation burden and sequence characteristics of individual mutations and of the signatures that were estimated to be present in each patient (Methods section). (c) Distribution of the six rearrangement signatures across the cell cycle. Dashed grey lines represent the predicted distribution of mutations for each signature based on simulations. (d) Distribution of the indel signatures across the cell cycle. Figure 2 Replication and transcriptional strand bias and strand-coordinated mutagenesis of mutational signatures. Forest plots showing replication (blue) and transcription (orange) strand bias for the 6 base substitution classes (a) and for the 12 base substitution signatures (b). Mutations were oriented in the pyrimidine context (the current convention for characterizing mutational signatures). Observed distribution between strands is shown as a diamond for replication and circle for transcriptional strands with 95% confidence intervals, against an expected probability of 0.5 (Supplementary Table 1 for values). (c) Relationship between processive group lengths (columns) and mutational signatures (rows). Processive groups were defined as sets of adjacent substitutions of the same mutational signature sharing the same reference allele, and the group length indicates the number of adjacent substitutions within each group. The size of each circle represents the number of groups (log10) observed for the specified group length (column) for each signature (row). The intensity of colour of each circle indicates significance of the likelihood of detection of a processive group of a defined length (−log10 of the P value obtained by comparing observed data to simulations, further details in Methods section). Figure 3 Relationship between mutational signatures and nucleosome occupancy. The distribution of the signal of nucleosome density (y axis) is shown in a 2 kb window centred on each mutation (position 0 on the x axis), for each signature. The averaged signal was calculated as the total amount of signal observed at each point divided by total number of mutations contributing to that signal. (a) Nucleosome density for aggregated substitutions (green), and for deletions observed in MMR-proficient (blue) and MMR-deficient (orange) samples. (b) Nucleosome density for the twelve base substitution signatures (note the degree of variation between substitution signatures relative to aggregated substitutions in a). The grey line shows the distribution predicted by simulations if mutations from each signature were randomly distributed. The analysis reveals that most of the observed distributions showed similar trends to those expected from simulations, apart from signatures 17, 18 and 26 and to a lesser extent signatures 5 and 8. Figure 4 A replication-related model of mutagenesis for putative APOBEC-related signatures 2 and 13. 1. During replication, transient moments of increased availability of single-stranded DNA (ssDNA) (for example, uncoupling between leading and lagging replicative strands or delays in elongation of the nascent lagging strand by Okazaki fragments) could occur, exposing ssDNA for APOBEC deamination, potentially for long genomic tracts. 2. Uracil-N-glycosylase (UNG) acts to remove undesirable uracils leaving a trail of abasic sites in its wake. Divergence of mutational processes occurs from this point. 3A Earlier in replication, error-prone translesion polymerases such as REV1 have been postulated to insert cytosines opposite abasic sites to avoid detrimental replication fork stalling or collapse. 4A The final outcome is stretches of successive C>G transversions at a TpC sequence context characteristic of signature 13. 3B Alternatively, uracils and abasic sites that are not fixed via REV1, undergo contingency processing, for example, the ‘A' rule. 4B The final outcome is of C>T mutations at a TpC sequence context. Table 1 Summary of relationships between each mutational signature and various genomic features. Mutational signature Mutation type Predominant features of signature Associated mutational process Transcriptional strand Replicative strand Replication time Chromatin organization 1 Sub C>T at CpG Deamination of methyl-cytosine (age associated)   Some bias Enriched late   5 Sub T>C Uncertain (age associated) Some bias Some bias Enriched late Slight enrichment at linker 2 Sub C>T at TpCpN APOBEC related Some bias Strong lagging strand bias Enriched late   13 Sub C>G at TpCpN APOBEC related Some bias Strong lagging strand bias Flat   6 Sub C>T (and C>A and T>C) MMR deficient   Some bias Flat   20 Sub C>A (and C>T and T>C) MMR deficient   Some bias Enriched late   26 Sub T>C MMR deficient Some bias Strong bias Enriched late Enriched at linker 3 Sub   HR deficient Some bias Some bias Enriched late   8 Sub C>A amplified by HR deficiency? Some bias   Enriched late   18 Sub C>A Uncertain Some bias Some bias Enriched late Enriched at nucleosomes and periodic 17 Sub T>G Uncertain   Some bias Enriched late Enriched at nucleosomes and periodic 30 Sub C>T Uncertain     Flat   RS1 Rearr Large tandem duplications (>100 kb) Uncertain type of HR deficiency? NA NA Enriched early   RS2 Rearr Dispersed translocations   NA NA Enriched early   RS3 Rearr Small tandem duplications (<10 kb) HR deficiency (BRCA1) NA NA Enriched early   RS4 Rearr Clustered translocations   NA NA Enriched early   RS5 Rearr Deletions HR deficient NA NA Enriched early   RS6 Rearr Other clustered rearrangements   NA NA Enriched early   Repeat-med Indel <3 bp indel at polynucleotide repeat tract amplified when MMR deficient NA NA Enriched late Enriched at linker and periodic Microhom Indel ≥3 bp indel with microhomology at breakpoint junction HR deficient NA NA Enriched late   HR, homologous recombination; indel, insertions/deletions; rearr, rearrangement; RS, rearrangement signature; sub, substitution. The 20 mutational signatures are noted in the left most column. This is followed by information on mutation classes, features that predominantly characterize each signature and associated aetiologies, if known. Relationships relating to transcriptional strands, replication time and strands and chromatin organization are also noted. ==== Refs De S. & Michor F. DNA replication timing and long-range DNA interactions predict mutational landscapes of cancer genomes . Nat. Biotechnol. 29 , 1103 –1108 (2011 ).22101487 Drier Y. . Somatic rearrangements across cancer reveal classes of samples with distinct patterns of DNA breakage and rearrangement-induced hypermutability . Genome Res. 23 , 228 –235 (2013 ).23124520 Koren A. . Differential relationship of DNA replication timing to different forms of human mutation and variation . Am. J. Hum. Genet. 91 , 1033 –1040 (2012 ).23176822 Kundaje A. . Ubiquitous heterogeneity and asymmetry of the chromatin environment at regulatory elements . Genome Res. 22 , 1735 –1747 (2012 ).22955985 Liu L. , De S. & Michor F. DNA replication timing and higher-order nuclear organization determine single-nucleotide substitution patterns in cancer genomes . Nat. Commun. 4 , 1502 (2013 ).23422670 Supek F. & Lehner B. Differential DNA mismatch repair underlies mutation rate variation across the human genome . Nature 521 , 81 –84 (2015 ).25707793 Woo Y. H. & Li W. H. DNA replication timing and selection shape the landscape of nucleotide variation in cancer genomes . Nat. Commun. 3 , 1004 (2012 ).22893128 Stamatoyannopoulos J. A. . Human mutation rate associated with DNA replication timing . Nat. Genet. 41 , 393 –395 (2009 ).19287383 Schuster-Bockler B. & Lehner B. Chromatin organization is a major influence on regional mutation rates in human cancer cells . Nature 488 , 504 –507 (2012 ).22820252 Polak P. . Cell-of-origin chromatin organization shapes the mutational landscape of cancer . Nature 518 , 360 –364 (2015 ).25693567 Nik-Zainal S. . Mutational processes molding the genomes of 21 breast cancers . Cell 149 , 979 –993 (2012 ).22608084 Nik-Zainal S. . The life history of 21 breast cancers . Cell 149 , 994 –1007 (2012 ).22608083 Helleday T. , Eshtad S. & Nik-Zainal S. Mechanisms underlying mutational signatures in human cancers . Nat. Rev. Genet. 15 , 585 –598 (2014 ).24981601 Alexandrov L. B. . Signatures of mutational processes in human cancer . Nature 500 , 415 –421 (2013 ).23945592 Nik-Zainal S. . Landscape of somatic mutations in 560 breast cancer whole-genome sequences . Nature http://dx.doi.org/10.1038/nature17676 (2016 ). Pena-Diaz J. & Jiricny J. Mammalian mismatch repair: error-free or error-prone? Trends Biochem. Sci. 37 , 206 –214 (2012 ).22475811 O'Donnell M. , Langston L. & Stillman B. Principles and concepts of DNA replication in bacteria, archaea, and eukarya . Cold Spring Harb. Perspect. Biol. 5 , a010108 (2013 ).23818497 Alberts B. Molecular Biology of the Cell 5th edn Garland Science (2008 ). Fragkos M. , Ganier O. , Coulombe P. & Mechali M. DNA replication origin activation in space and time . Nat. Rev. Mol. Cell Biol. 16 , 360 –374 (2015 ).25999062 ENCODE https://http://www.encodeproject.org (2012 ). Guan Z. . Decreased replication origin activity in temporal transition regions . J. Cell Biol. 187 , 623 –635 (2009 ).19951913 Guilbaud G. . Evidence for sequential and increasing activation of replication origins along replication timing gradients in the human genome . PLoS Comput. Biol. 7 , e1002322 (2011 ).22219720 Signatures C. M http://cancer.sanger.ac.uk/cosmic/signatures (2015 ). Conticello S. G. The AID/APOBEC family of nucleic acid mutators . Genome Biol. 9 , 229 (2008 ).18598372 Harris R. S. , Petersen-Mahrt S. K. & Neuberger M. S. RNA editing enzyme APOBEC1 and some of its homologs can act as DNA mutators . Mol. Cell 10 , 1247 –1253 (2002 ).12453430 Byeon I. J. . NMR structure of human restriction factor APOBEC3A reveals substrate binding and enzyme specificity . Nat. Commun. 4 , 1890 (2013 ).23695684 Holtz C. M. , Sadler H. A. & Mansky L. M. APOBEC3G cytosine deamination hotspots are defined by both sequence context and single-stranded DNA secondary structure . Nucleic Acids Res. 41 , 6139 –6148 (2013 ).23620282 Nik-Zainal S. . Association of a germline copy number polymorphism of APOBEC3A and APOBEC3B with burden of putative APOBEC-dependent mutations in breast cancer . Nat. Genet. 46 , 487 –491 (2014 ).24728294 Nouspikel T. DNA repair in mammalian cells : Nucleotide excision repair: variations on versatility . Cell. Mol. Life Sci. 66 , 994 –1009 (2009 ).19153657 Pleasance E. D. . A comprehensive catalogue of somatic mutations from a human cancer genome . Nature 463 , 191 –196 (2010 ).20016485 Hanawalt P. C. & Spivak G. Transcription-coupled DNA repair: two decades of progress and surprises . Nature Rev. Mol. Cell Biol. 9 , 958 –970 (2008 ).19023283 Wang Y. , Sheppard T. L. , Tornaletti S. , Maeda L. S. & Hanawalt P. C. Transcriptional inhibition by an oxidized abasic site in DNA . Chem. Res. Toxicol. 19 , 234 –241 (2006 ).16485899 Tornaletti S. , Maeda L. S. & Hanawalt P. C. Transcription arrest at an abasic site in the transcribed strand of template DNA . Chem. Res. Toxicol. 19 , 1215 –1220 (2006 ).16978026 Tolstorukov M. Y. , Volfovsky N. , Stephens R. M. & Park P. J. Impact of chromatin structure on sequence variability in the human genome . Nat. Struct. Mol. Biol. 18 , 510 –515 (2011 ).21399641 Sasaki S. . Chromatin-associated periodicity in genetic variation downstream of transcriptional start sites . Science 323 , 401 –404 (2009 ).19074313 Schopf B. . Interplay between mismatch repair and chromatin assembly . Proc. Natl Acad. Sci. USA 109 , 1895 –1900 (2012 ).22232658 Lujan S. A. . Heterogeneous polymerase fidelity and mismatch repair bias genome variation and composition . Genome Res. 24 , 1751 –1764 (2014 ).25217194 Strauss B. S. The "A" rule revisited: polymerases as determinants of mutational specificity . DNA Repair 1 , 125 –135 (2002 ).12509259 Kass E. M. & Jasin M. Collaboration and competition between DNA double-strand break repair pathways . FEBS Lett. 584 , 3703 –3708 (2010 ).20691183 Mateos-Gomez P. A. . Mammalian polymerase theta promotes alternative NHEJ and suppresses recombination . Nature 518 , 254 –257 (2015 ).25642960 Ceccaldi R. . Homologous-recombination-deficient tumours are dependent on Poltheta-mediated repair . Nature 518 , 258 –262 (2015 ).25642963 Costantino L. . Break-induced replication repair of damaged forks induces genomic duplications in human cells . Science 343 , 88 –91 (2014 ).24310611 Durant S. T. & Nickoloff J. A. Good timing in the cell cycle for precise DNA repair by BRCA1 . Cell Cycle 4 , 1216 –1222 (2005 ).16103751 Aguilera A. & Gomez-Gonzalez B. Genome instability: a mechanistic view of its causes and consequences . Nat. Rev. Genet. 9 , 204 –217 (2008 ).18227811 Carr A. M. & Lambert S. Replication stress-induced genome instability: the dark side of replication maintenance by homologous recombination . J. Mol. Biol. 425 , 4733 –4744 (2013 ).23643490 Simpson L. J. & Sale J. E. Rev1 is essential for DNA damage tolerance and non-templated immunoglobulin gene mutation in a vertebrate cell line . EMBO J. 22 , 1654 –1664 (2003 ).12660171 Alexandrov L. B. , Nik-Zainal S. , Wedge D. C. , Campbell P. J. & Stratton M. R. Deciphering signatures of mutational processes operative in human cancer . Cell Rep. 3 , 246 –259 (2013 ).23318258 Consortium E. P. An integrated encyclopedia of DNA elements in the human genome . Nature 489 , 57 –74 (2012 ).22955616 Kaykov A. & Nurse P. The spatial and temporal organization of origin firing during the S-phase of fission yeast . Genome Res. 25 , 391 –401 (2015 ).25650245 Gilbert D. M. Making sense of eukaryotic DNA replication origins . Science 294 , 96 –100 (2001 ).11588251 Rhind N. DNA replication timing: random thoughts about origin firing . Nat. Cell Biol. 8 , 1313 –1316 (2006 ).17139278 Vengrova S. & Dalgaard J. Z. RNase-sensitive DNA modification(s) initiates S. pombe mating-type switching . Genes Dev. 18 , 794 –804 (2004 ).15059961 Arcangioli B. & de Lahondes R. Fission yeast switches mating type by a replication-recombination coupled process . EMBO J. 19 , 1389 –1396 (2000 ).10716938 Kaykov A. , Taillefumier T. , Bensimon A. & Nurse P. Molecular combing of single DNA molecules on the 10 megabase scale . Sci. Rep. 6 , 19636 (2016 ).26781994 Gaffney D. J. . Controls of nucleosome positioning in the human genome . PLoS Genet. 8 , e1003036 (2012 ).23166509

PMC005xxxxxx/PMC5001791.txt

==== Front Rambam Maimonides Med JRambam Maimonides Med JRMMJRambam Maimonides Medical Journal2076-9172Rambam Health Care Campus 10.5041/RMMJ.10246rmmj-7-3-e0019Original ResearchEffectiveness of Inferior Vena Cava Filters without Anticoagulation Therapy for Prophylaxis of Recurrent Pulmonary Embolism Zektser Miri M.D.1Bartal Carmi M.D.2Zeller Lior M.D.1Nevzorov Roman M.D.1Jotkowitz Alan M.D.1Stavi Vered M.D.1Romanyuk Vitaly M.D.3Chudakov Gregory M.D.3Barski Leonid M.D.1* 1 Department of Internal Medicine F, Soroka University Medical Center, Beer-Sheva, Israel 2 Department of Internal Medicine E, Soroka University Medical Center, Beer-Sheva, Israel 3 Department of Radiology, Soroka University Medical Center, Beer-Sheva, Israel* To whom correspondence should be addressed. E-mail:lbarski@bgu.ac.il7 2016 28 7 2016 7 3 e0019Copyright: © 2016 Zekster et al.2016This is an open-access article. All its content, except where otherwise noted, is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.Objective The optimal treatment of deep vein thrombosis (DVT) is anticoagulation therapy. Inferior vena cava filter (IVC) placement is another option for the prevention of pulmonary embolism (PE) in patients with deep vein thrombosis. This is used mostly in patients with a contraindication to anticoagulant therapy. The purpose of the present study was to compare the two options. Methods A retrospective cohort study of two groups of patients with DVT: patients who received an IVC filter and did not receive anticoagulation due to contraindications; and patients with DVT and similar burden of comorbidity treated with anticoagulation without IVC insertion. To adjust for a potential misbalance in baseline characteristics between the two groups, we performed matching for age, gender, and Charlson’s index, which is used to compute the burden of comorbid conditions. The primary outcome was an occurrence of a PE. Results We studied 1,742 patients hospitalized with the diagnosis of DVT in our hospital;93 patients from this population received IVC filters. Charlson’s score index was significantly higher in the IVC filter group compared with the anticoagulation group. After matching of the groups of patients according to Charlson’s score index there were no significant differences in primary outcomes. Conclusion Inferior vena cava filter without anticoagulation may be an alternative option for prevention of PE in patients with contraindications to anticoagulant therapy. Deep vein thrombosisinferior vena cava filterpulmonary embolism ==== Body INTRODUCTION The most effective treatment for patients with deep vein thrombosis (DVT) is anticoagulation therapy. Traditionally, anticoagulant therapy involves parenteral anticoagulants (heparin, low-molecular heaprins), overlapping with and followed by oral vitamin K antagonists. Recently, new or direct oral anticoagulants (NOACs or DOACs), including the factor Xa inhibitors rivaroxaban, apixaban, and edoxaban, and the direct thrombin inhibitor dabigatran etexilate have been developed. These drugs have been approved for the treatment of acute deep vein thrombosis (DVT) and pulmonary embolism (PE).1–5 Another therapeutic option for DVT treatment is inferior vena cava (IVC) filter placement. The only widely accepted and validated indications for IVC filter placement are: (1) absolute contraindication to therapeutic anticoagulation; (2) failure of anticoagulation when there is acute proximal venous thrombosis; and (3) life-threatening hemorrhage on anticoagulation. There are other situations in which the indication for IVC filter placement is controversial.6 In the last decade there has been increased use of filters for treatment of DVT, including their use, in addition to anticoagulant therapy, in patients with PE and/or a large clot burden, poor cardiopulmonary reserve, or a suspected increased risk for recurrence.7–9 Their use has also been advocated by several guidelines.10–12 The results of the most recent investigations do not support such a strategy, demonstrating that, compared with anticoagulation alone, placement of a retrievable IVC filter for 3 months in addition to anticoagulation provided no benefit in terms of PE recurrence or mortality in patients presenting with an acute symptomatic PE.13 A number of percutaneous IVC filters have been developed since the introduction of the Greenfield filter in 1973. The newer devices are designed to optimize flow dynamics, maximize clot-trapping capacity, and increase ease of insertion.6,14–16 The newest devices are potentially retrievable. Despite the obvious theoretical benefit, carefully controlled trials that demonstrate the ability of IVC filters to decrease recurrence rates or mortality from PE have not been performed.6,14,17 While the data suggest that recurrent embolism is unusual following filter insertion (2%–4% in most series), strong scientific evidence that IVC filters prevent death from PE is not currently available.18 A systematic review of retrievable IVC filters estimated an incidence of PE following filter placement of 1.3%.19 An impact on case fatality rate with vena cava filters was shown in unstable patients, whether or not they received thrombolytic therapy, and in stable patients who received thrombolytic therapy. Vena cava filters were associated with a lower all-cause in-hospital case fatality rate compared with anticoagulation therapy only without IVC filters among unstable patients who received thrombolytic therapy (7.6% versus 18%) and lower all-cause case fatality rate in unstable patients who did not receive thrombolytic therapy (33% versus 51%).7 Previous studies suggest that the addition of IVC filter placement to standard anticoagulation results in a reduction in the risk of subsequent PE; however, filter placement increases the incidence of DVT. Filter placement does not appear to increase the overall risk of venous thromboembolism (VTE), post-thrombotic syndrome, or mortality.20,21 Of note, neither these studies nor any others have compared the efficacy of anticoagulation with IVC filter placement in the absence of anticoagulation. Whether concomitant anticoagulant therapy should be utilized following filter placement is unknown. Although filter placement protects the pulmonary vascular bed, it does nothing to lessen the thrombotic predisposition or to decrease the incidence of lower-extremity venous thrombosis. Small thrombi are capable of passing through patent filters or through collaterals around obstructed filters; furthermore, direct thrombus extension can occur through the filter itself. Because patients with IVC filters are at risk for IVC thrombosis, insertion site thrombosis, and recurrence of the initial thromboembolic event, continued use of anticoagulants when there are no contraindications is advisable.14,22 The purpose of our study was to investigate the natural history of patients with DVT and IVC filter insertion without anticoagulation therapy and compare them with patients on anticoagulation only without IVC filter. METHODS We performed a retrospective cohort study of patients with DVT admitted to the Soroka University Medical Center, a 1,000-bed tertiary care teaching hospital that serves as the only tertiary referral hospital for southern Israel (estimated population 1,000,000) between January 1, 2006 and January 1, 2010. Two groups of patients with DVT were compared: patients who received an IVC filter and no anticoagulation; and patients with DVT and similar burden of comorbidity treated with anticoagulation without IVC insertion. All patients in the IVC filter group were treated with IVC filter insertion and did not receive any anticoagulant drugs. In patients in the IVC filter group the contraindications for anticoagulant therapy were active bleeding, recent surgery, pre-existing life-threatening bleeding, various coagulopathies, and the inability to receive anticoagulant therapy per the decision of the senior attending physician. To adjust for a potential misbalance in baseline characteristics between the group of patients who received anticoagulation and the patients that were implanted with an IVC filter, we performed matching for age, gender, and Charlson’s index, which is used to compute the burden of comorbid conditions. Charlson’s index is a list of 19 conditions each of which has a weight assigned from 1 to 6, derived from relative risk estimates of a proportional hazard regression model. The higher the score, the more severe the burden of comorbidity is, and the likelyhood of mortality increases.23 Charlson’s index is a well-validated indicator of overall disease burden and can be accurately calculated based on ICD-9 diagnoses. The primary outcome was an occurrence of a PE. The secondary outcomes were 1-year all-cause mortality, 2-year all-cause mortality, a recurrent hospitalization for a thrombotic event in the first year after DVT diagnoses, and length of hospital stay. The discharge diagnoses (ICD-9) were used to identify the subjects with DVT. Deep vein thrombosis was defined as a positive compression ultrasonography study of the lower extremities. Pulmonary embolism was defined as the obstruction of the pulmonary artery or one of its branches due to thrombus revealed either on angiographic computed tomography study or ventilation-perfusion scan. Information on the patients’ demographic characteristics, ICD-9 diagnoses, medications, and clinical and laboratory data was obtained during a comprehensive medical chart review and from the computerized hospital database. The study was approved by the Institutional Review Board prior to its initiation. The following types of IVC filters were used in the present study: Simon Nitinol Filter (SNF) (Bard Peripheral Vascular, Inc., Temple, AZ, USA), a non-retrievable filter, was inserted to 41 patients of this study; OptEase Filter (Cordis Endovascular; a Johnson & Johnson Company, Warren, NJ, USA), a retrievable filter, was inserted to 37 patients; Cook Celect Filter (William Cook Europe, Bjaeverskov, Denmark), a retrievable filter, was inserted to 11 patients; and ALN Filter (ALN Implants Chirurgicaux, Ghisonaccia, France), a retrievable filter, was inserted to 5 patients. All filters were placed and positioned by ultrasound guide with femoral or jugular insertion sites. No complications were demonstrated in filter placement. Statistical Analysis The results are presented as the mean ± standard deviation for continuous variables and as number and percentage of total patients for categorical data. Student’s t test was used for comparison of the continuous variables and chi-square test for categorical data with the use of Fisher’s exact test if needed. We used the Mann–Whitney test for the comparison of continuous variables not distributed normally and presented as median and interquartile range (IQR). Survival curves were calculated by the Kaplan–Meier method, and comparison between groups of patients was performed by log-rank test. For multi-variable analysis, the binary logistic regression model was applied. The initial selection of the variables entered into the model was based on univariate analysis significance with an inclusion criterion of P<0.10. The results of multivariate analysis were presented as the hazard ratio (HR) with 95% confidence interval (CI). A two-sided P value <0.05 was considered statistically significant. The statistical analysis was performed with SPSS software (version 17.0). RESULTS From January 2006 to December 2010 at Soroka University Medical Center there were 1,742 patients with DVT, and 93 patients from this population received IVC filters. General characteristics of the entire cohort are shown in Table 1. No age differences were found between the two groups of patients. The majority of the patients in the anticoagulation group were women (58.8% compared to 48.4% in IVC filter group, P=0.048). Table 1 General Patient Characteristics. Characteristics IVC Filter Group n=93 Anticoagulation Group n=1649 P Value Age (y), mean±SD 65.2±17.2 62.7±22.6 0.2 Female gender, n (%) 45 (48.4) 969 (58.8) 0.048 History of myocardial infarction, n (%) 14 (15.1) 328 (19.9) 0.3 Chronic heart failure, n (%) 6 (6.5) 211 (12.8) 0.07 Peripheral vascular disease, n (%) 9 (9.7) 301 (18.3) 0.035 Dementia, n (%) 11 (11.8) 214 (13.0) 0.7 Chronic obstructive lung disease, n (%) 4 (4.3) 170 (10.3) 0.06 Connective tissue disease, n (%) 5 (5.4) 52 (3.2) 0.2 History of stroke, n (%) 19 (20.4) 218 (13.2) 0.048 Liver cirrhosis, n (%) 2 (2.2) 48 (2.9) 0.7 Chronic renal failure, n (%) 34 (36.6) 559 (33.9) 0.6 Diabetes mellitus, n (%) 23 (24.7) 453 (27.5) 0.6 Solid tumor, n (%) 41 (44.1) 403 (24.4) <0.001 Leukemia, n (%) 1 (1.1) 13 (0.8) 0.5 Lymphoma, n (%) 4 (4.3) 44 (2.7) 0.3 Metastatic tumor, n (%) 26 (28.0) 189 (11.5) <0.001 Charlson score, median (IQR) 4 (2; 8) 3 (0; 6) <0.001 IQR, interquartile range (25th; 75th percentiles). Patients in the IVC filter group compared with the anticoagulation group had more chronic medical conditions prior to hospitalization, including peripheral vascular disease, cerebral vascular disease, hemiplegia, and solid and metastatic tumors. Therefore, Charlson’s score index was significantly higher in patients in the IVC filter group compared with the anticoagulation group [4 (IQR 2; 8) versus 3 (IQR 0; 6), P<0.001]. Clinical outcomes before matching are shown in Table 2. The rate of 1-year and 2-year mortality before matching according to severe comorbid diseases was significantly higher in patients in the IVC filter group compared to the anticoagulation group (49.5% versus 24.3%, P<0.001; 52.7% versus 30.5%, P<0.001, respectively). No significant difference was found between the two groups of patients for 1-year readmission. Table 2 Outcomes. Outcomes IVC Filter Group n=93 Anticoagulation Group n=1649 P Value 1-Year readmission, n (%) 83 (89.5) 1402 (85.0) 0.3 1-Year mortality, n (%) 46 (49.5) 400 (24.3) <0.001 2-Year mortality, n (%) 49 (52.7) 503 (30.5) <0.001 Multivariate analysis of factors associated with 1-year mortality (Table 3) revealed that advanced age and metastatic tumor were independent predictors (HR 1.01, 95% CI 1.006–1.01; HR 6.1, 95% CI 4.99–7.42, respectively). Table 3 Multivariate Analysis for 1-Year Mortality. Variables Hazard Ratio 95% CI P Value Age (the increment for each year) 1.01 1.006–1.01 <0.001 Metastatic tumor 6.1 4.99–7.42 <0.001 Figure 1 shows the Kaplan–Meier survival plots for 1-year survival stratified by IVC filter and anticoagulation and demonstrates a decreased survival of patients in the IVC filter group compared to the anticoagulation group. Figure 1 Kaplan–Meier Survival Plots for 1-year Survival. Stratified by IVC filter and anticoagulation. Log-rank test P<0.001. After matching both groups of patients according to Charlson’s score index there were no significant differences in primary (occurrence of PE) or secondary outcomes (1-year readmission, 1-year mortality, 2-year mortality) (Table 4). Table 4 Outcomes after Matching. Outcomes IVC Filter Group n=92 Anticoagulation Group n=92 P Value Pulmonary embolism, n (%) 6 (6.5) 3 (3.2) 0.5 1-Year readmission, n (%) 82 (89.1) 79 (85.9) 0.6 1-Year mortality, n (%) 45 (48.9) 32 (34.8) 0.1 2-Year mortality, n (%) 48 (52.2) 39 (42.4) 0.2 DISSCUSSION In the present study IVC filters were used for the prophylaxis of recurrent PE in patients with DVT. The insertion of an IVC filter was a treatment option only for those patients who were diagnosed recently with a DVT, with contraindications to anticoagulant therapy. This is the only strict indication for IVC insertion published in guidelines of the American College of Chest Physicians,24 American Heart Association,12 and the British Committee for Standards in Hematology.22 There is no consensus on the role of IVC filters in reducing mortality or recurrent PE in patients with other indications, such as patients with VTE despite anticoagulation, patients with recent VTE requiring anticoagulation while awaiting surgery, or primary prevention in high-risk patients. This could be a possible explanation for the observed outcomes in our patients: after performing matching of the two groups of patients we found no significant differences in primary and secondary outcomes between the two groups. This confirms the utility of using IVC filters in patients with absolute contraindications to anticoagulation. It is possible that if, in our institution, the IVC was inserted for other indications the results would be less favorable in the filter group. The non-selective use of IVC filters is associated with unacceptable morbidity and mortality, and only a few patients among those surveyed would have benefited from an IVC filter.25 The mortality rate in our study was high, even before matching the group of patients with DVT without IVC filter insertion. In fact, the majority of patients in both groups in this study hospitalized with a diagnosis of DVT had several severe comorbidities, including malignant metastatic tumors, cerebral vascular disease with hemiplegia, myocardial infarction, congestive heart failure, chronic lung diseases, chronic liver diseases, and chronic kidney diseases; they also had a high Charlson’s score index. It is possible that other less complicated patients with diagnoses of DVT may be treated as outpatients with good results.7,19,26 In the IVC filter group about 50% of the patients had cancer, and 28% had a metastatic tumor; the percentage was significantly higher than in the no-IVC filter group. Cancer itself, or its treatment, may result in certain clinical complications that make systemic anticoagulation very risky. Many recent studies questioned the need to insert IVC filters in cancer patients, particularly in those with advanced-stage disease whose survival is short and in whom prevention of PE may be of little clinical benefit and could be a poor utilization of resources.27,28 Although in our study we did not perform formal subgroup analysis, the IVC filter group, which had a higher proportion of cancer patients, did not have a higher 1- or 2-year mortality compared to the other group. Although more than 50% of our patients were inserted with retrievable filters, none of the filters was removed during our 2 years of observation. In this group of patients almost 50% died in the first 2 years, and as was mentioned previously this group of patients had significant comorbidities. No randomized clinical trials have been performed to compare retrievable and non-retrievable filters. In one large retrospective study of retrievable filters placed for various indications, only 18.7% of the filters were successfully retrieved.29 Significant complications after IVC filter insertion were not demonstrated in this study. Our study did not reveal an increased incidence of hospitalization due to recurrent thrombotic events, and this might allay the fears of those clinicians who might be reluctant to place IVC filters due to the concern of the filter being a nexus for thrombosis. Only 10 patients in the group of filter insertion received anticoagulation during 2 years of follow-up. This demonstrates that the contraindications for anticoagulation in the majority of patients were not temporary, but rather persistent. In addition, significant complications after IVC filter insertion were not demonstrated in this study. Not surprisingly, multivariate analysis for 1-year mortality confirms that advanced age and metastatic tumor were predictors of 1-year mortality in the present study. The present study adds to the growing body of literature that an effective treatment for patients with DVT and an absolute contraindication to anticoagulation is an IVC filter. LIMITATIONS OF THIS STUDY This was a retrospective study performed in only one medical center. The patients on anticoagulation in our study did not receive a NOAC. CONCLUSION Inferior vena cava filter without anticoagulation may be an alternative therapeutic option for prevention of PE in patients with contraindications to anticoagulant therapy. Conflict of interest: No potential conflict of interest relevant to this article was reported. Abbreviations DOACdirect oral anticoagulant DVTdeep vein thrombosis IVCinferior vena cava NOACnew oral anticoagulant PEpulmonary embolism VTEvenous thromboembolism ==== Refs REFERENCES 1 Franco P Losub DI New anticoagulants in the management of venous thromboembolism in women Thromb Res 2015 135 Suppl 1 S5 7 10.1016/S0049-3848(15)50431-3 25903537 2 Arcelus JI Domènech P Fernández-Capitan Mdel C Rivaroxaban in the treatment of venous thromboembolism and the prevention of recurrences: a practical approach Clin Appl Thromb Hemost 2015 21 297 308 10.1177/1076029614561321 25504999 3 Yeh CH Hogg K Weitz JI Overview of the new oral anticoagulants: opportunities and challenges Arterioscler Thromb Vasc Biol 2015 35 1056 65 10.1161/ATVBAHA.115.303397 25792448 4 Sharifi M Freeman W Bay C Sharifi M Schwartz F Low incidence of post-thrombotic syndrome in patients treated with new oral anticoagulants and percutaneous endovenous intervention for lower extremity deep venous thrombosis Vasc Med 2015 20 112 16 10.1177/1358863X14553882 25832599 5 Bacchus F Schulman S Clinical experience with the new oral anticoagulants for treatment of venous thromboembolism Arterioscler Thromb Vasc Biol 2015 35 513 19 10.1161/ATVBAHA.114.303396 25717178 6 Streiff MB Vena caval filters: a comprehensive review Blood 2000 95 3669 77 10845895 7 Stein PD Matta F Keyes DC Willyerd GL Impact of vena cava filters on in hospital case fatality rate from pulmonary embolism Am J Med 2012 125 478 84 10.1016/j.amjmed.2011.05.025 22310013 8 Stein PD Kayali F Olson RE Twenty-one–year trends in the use of inferior vena cava filters Arch Intern Med 2004 164 1541 5 10.1001/archinte.164.14.1541 15277286 9 Spencer FA Bates SM Goldberg RJ A population-based study of inferior vena cava filters in patients with acute venous thromboembolism Arch Intern Med 2010 170 1456 62 10.1001/archinternmed.2010.272 20837832 10 Kaufman JA Kinney TB Streiff MB Guidelines for the use of retrievable and convertible vena cava filters: report from the Society of Interventional Radiology multidisciplinary consensus conference J Vasc Interv Radiol 2006 17 449 59 10.1097/01.RVI.0000203418-39769.0D 16567669 11 Caplin DM Nikolic B Kalva SP Quality improvement guidelines for the performance of inferior vena cava filter placement for the prevention of pulmonary embolism J Vasc Interv Radiol 2011 22 1499 506 10.1016/j.jvir.2011.07.012 21890380 12 Jaff MR McMurtry MS Archer SL Management of massive and submassive pulmonary embolism, iliofemoral deep vein thrombosis, and chronic thromboembolic pulmonary hypertension: a scientific statement from the American Heart Association Circulation 2011 123 1788 830 10.1161/CIR.0b013e318214914f 21422387 13 Mismetti P Laporte S Pellerin O PREPIC2 Study Group Effect of a retrievable inferior vena cava filter plus anticoagulation vs anticoagulation alone on risk of recurrent pulmonary embolism: a randomized clinical trial JAMA 2015 313 1627 35 10.1001/jama.2015.3780 25919526 14 Becker DM Philbrick JT Selby JB Inferior vena cava filters. Indications, safety, effectiveness Arch Intern Med 1992 152 1985 94 10.1001/archinte.1992.00400220023005 1417371 15 Kinney TB Update on inferior vena cava filters J Vasc Interv Radiol 2003 14 425 40 10.1097/01.RVI.0000064860.87207.77 12682199 16 Kercher K Sing RF Overview of current inferior vena cava filters Am Surg 2003 69 643 8 12953819 17 Girard P Stern JB Parent F Medical literature and vena cava filters: so far so weak Chest 2002 122 963 7 10.1378/chest.122.3.963 12226040 18 Ku GH Billett HH Long lives, short indications. The case for removable inferior cava filters Thromb Haemost 2005 93 17 22 15630485 19 Angel LF Tapson V Galgon RE Restrepo MI Kaufman J Systematic review of the use of retrievable inferior vena cava filters J Vasc Interv Radiol 2011 22 1522 1530.e3 10.1016/j.jvir.2011.08.024 22024114 20 Decousus H Leizorovicz A Parent F A clinical trial of vena caval filters in the prevention of pulmonary embolism in patients with proximal deep-vein thrombosis. Prévention du Risque d’Embolie Pulmonaire par Interruption Cave Study Group N Engl J Med 1998 338 409 15 10.1056/NEJM199802123380701 9459643 21 PREPIC Study Group Eight-year follow-up of patients with permanent vena cava filters in the prevention of pulmonary embolism: the PREPIC (Prevention du Risque d’Embolie Pulmonaire par Interruption Cave) randomized study Circulation 2005 112 416 22 10.1161/CIRCULATIONAHA.104.512834 16009794 22 British Committee for Standards in Haematology Writing Group Baglin TP Brush J Streiff M Guidelines on use of vena cava filters Br J Haematol 2006 134 590 5 10.1111/j.1365-2141.2006.06226.x 16869824 23 O’Connel RL Lim LL Utility of the Charlson comorbidity index computed from routinely collected hospital discharge diagnosis codes Methods Inf Med 2000 39 7 11 10786063 24 Kearon C Akl EA Comerota AJ Antithrombotic therapy and prevention of thrombosis, 9th ed: American College of Chest Physicians evidence-based clinical practice guidelines Chest 2012 141 e419S e94S 22315268 25 Thomsen MB Lindblad B Bergqvist D Fatal pulmonary embolism in an unselected series: the possible role of caval filters in prevention Eur J Surg 1994 160 553 9 7849157 26 Carson JL Kelley MA Duff A The clinical course of pulmonary embolism N Engl J Med 1992 326 1240 5 10.1056/NEJM199205073261902 1560799 27 Karabinis VD Mehta V Dupont EL Matsumoto T Kerstein MD Potential of overuse of the inferior vena cava filter Surg Gynecol Obstet 1993 177 463 7 8211597 28 Jarrett BP Dougherty MJ Calligaro KD Inferior vena cava filters in malignant disease J Vasc Surg 2002 36 704 7 10.1016/S0741-5214(02)00121-0 12368729 29 Barginear MF Lesser M Akerman ML Need for inferior vena cava filters in cancer patients: a surrogate marker for poor outcome Clin Appl Thromb Hemost 2009 15 263 9 10.1177/1076029608315165 18385149

PMC005xxxxxx/PMC5001792.txt

==== Front Rambam Maimonides Med JRambam Maimonides Med JRMMJRambam Maimonides Medical Journal2076-9172Rambam Health Care Campus 10.5041/RMMJ.10247rmmj-7-3-e0020Original ResearchFor Better Orchiopexy, Processus Vaginalis Should Be Dissected and a High Ligation Should Be Performed Sonmez Kaan M.D.ProfessorKarabulut Ramazan M.D.Professor*Turkyilmaz Zafer M.D.ProfessorKaya Cem M.D.Pehlivan Yildiz M.D.Basaklar A. Can M.D.ProfessorDepartment of Pediatric Surgery, Medical Faculty of Gazi University, Ankara, Turkey* To whom correspondence should be addressed. E-mail:karabulutr@yahoo.com7 2016 28 7 2016 7 3 e0020Copyright: © 2016 Sonmez et al.2016This is an open-access article. All its content, except where otherwise noted, is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.Objective Data on the prevalence of patent processus vaginalis (PPV) and hernia in patients with cryptorchidism are controversial. While some pediatric surgeons do not dissect the processus vaginalis (PV), most prefer to do so to prevent hernia formation and to achieve an effective orchiopexy outcome. This study was performed to evaluate the importance of dissection and high ligation of the PV during treatment of undescended testis (UT). Methods The clinical findings and surgical procedures of 55 patients with UT were retrospectively investigated. Results The mean patient age was 2.5 (range 1.0–12.0) years. Non-palpable testis (NPT) was located on the right and left side in 39 and 16 patients, respectively. Ultrasonography revealed no testis in 10 patients and an atrophic testis in 7 patients. Seven patients had a parent with an inguinal hernia, and the silk sign or a PPV was detected during inguinoscrotal examination in 22 patients. Undescended testis repair was performed by an inguinal approach in all patients. The inguinal canal was opened in all patients; 42 patients had a wider-than-normal internal ring (>2.5 cm), and the posterior wall of the inguinal canal was consequently weakened. Two-stage orchiopexy was performed in 2 patients, and 15 underwent the Prentiss maneuver. In the remaining patients, the dissection was easily done, and the orchiopexy was performed without any difficulty. Scrotal edema and wound infection occurred in five and two patients, respectively. One patient presented with an atrophic testis, and three had recurrent UT. Inguinal hernia was not observed in any of the patients during the study period, and all procedures were performed on an outpatient basis. Conclusion High ligation of the PV is an effective method for successful orchiopexy and prevention of inguinal hernia in patients with NPT and UT. Herniahigh ligationprocessus vaginalisundescended testis ==== Body INTRODUCTION The incidence of undescended testis (UT) in pediatric patients >1 year of age is 2.2%. Approximately 20% of these patients have a non-palpable testis (NPT). In about half of patients with NPT, one testis is located in the abdominal cavity; the remaining testis is often atrophic, either secondary to antenatal torsion in utero or agenesis.1,2 Data on the prevalence of patent processus vaginalis (PPV) and hernia in patients with cryptorchidism are controversial.3 While some pediatric surgeons do not dissect the processus vaginalis (PV), most prefer to do so to prevent hernia formation and to achieve an effective orchiopexy outcome.4,5 The aim of the present study was to evaluate the importance of dissection of the PV from the spermatic cord and vessels during correction procedures for UT at the internal ring level or in an intra-abdominal location. METHODS In total, 1,300 patients with UT and 55 with intra-abdominal NPT underwent operations in our clinic from 1998 to 2013. This retrospective study reviewed the records of the 55 children who underwent inguinal orchiopexy for treatment of intra-abdominal NPT. The age, period from diagnosis to operation date, side of UT, clinical findings, and surgical procedures were retrospectively investigated. RESULTS The mean age of the patients was 2.5 (range 1.0–12.0) years. The NPT was located on the right side in 39 patients and on the left side in 16 patients. Preoperative physical examination revealed a palpable mobile mass at the internal ring level that was thought to be the UT in 41 (75.5%) patients; the testes were not palpable in the remaining patients. Scrotal and abdominal ultrasonography was performed in all patients. Ultrasonography revealed no testis in 10 patients and an atrophic testis in 7 patients. Seven patients had a parent with an inguinal hernia. The silk sign or a PPV was detected during inguinoscrotal examination in 22 patients. No patients displayed inguinal pain or signs of incarceration such as nausea, vomiting, or erythema of the scrotum or inguinal area. Undescended testis repair was performed by an inguinal approach in all patients. The inguinal canal was opened in all patients; in 42 patients, the internal ring was wider than normal (>2.5 cm), and the posterior wall of the inguinal canal was consequently weakened. Furthermore, the PV was very delicate and fragile and the testes seemed to be in the PV during testicular retraction. The PPV was oriented as in inguinal hernia repair, allowing for free movement of intra-abdominal fluid. The PV was dissected from the pampiniform plexus and vas deferens during the surgical descent of the testis. Specifically, the PV was dissected to the level of the preperitoneal fat tissue, and high ligation was performed. During this procedure, the PV was very wide and thin in 15 patients and was opened. On the other hand, a purse string could be performed in 22 patients, and high ligation was performed. In patients in whom the testis could be moved into the widened PV sac and in cases involving extreme tension, PV dissection began from the region corresponding to the mid-distance between the internal ring and the testis. The aim of this maneuver was circumferentially to dissect the PV from a narrower diameter and allow for easier dissection of the PV, even in cases involving perforation of the PV. Two patients required two-stage orchiopexy, and the Prentiss maneuver was performed in 15 patients, effectively reducing the testis to its anatomical position. In the remaining patients, the dissection was easily performed, and the testes were put into the scrotum without any difficulty. A subdartos pouch was created in all patients, and the orchiopexy was completed. Scrotal edema and wound infection occurred in five and two patients, respectively, in the first postoperative week. The mean follow-up period was 2.5 years (range 8 months to 9 years). Testicular atrophy occurred in one patient, and recurrent UT occurred in three. One of these three patients underwent correction by the Prentiss maneuver, and the remaining two underwent classical orchiopexy. No patients developed an inguinal hernia. All patients were treated by pediatric surgeons on an outpatient basis. DISCUSSION At about the time of birth, the portion of the PV between the testicle and the abdominal cavity is obliterated, leaving the peritoneal cavity separate from the PV, which surrounds the testicle. If obliteration of the PV is incomplete, a variety of anomalies can occur. Most true cases of UT (90%) are associated with a PPV, with the exception of retractile testes. In one series, laparoscopy was performed in patients with UT in whom the testes were non-palpable; the testes were visualized in 47 patients, and 43 (91%) had a PPV. Of 77 patients in whom the PV was closed, 75 (97%) had vanished or absent testes.6 In another study, epididymal anomalies in patients with PPV included inguinal hernia, hydrocele, and UT. The incidence of PPV associated with UT was higher than that of other inguinal anomalies (69%), and epididymal anomalies were more frequently associated with undescended (72%) than descended (34%) testes. The study concluded that androgenic stimulation is mandatory for PV closure and that PPV occurs with a high incidence in patients with UT.7 In another study by Radmayr et al.,8 inguinal hernias were detected in 26% of patients with UT. Traditionally, the recommended treatment for UT is high ligation of the hernial sac after proper dissection up to the deep ring. Jain et al.4 evaluated 450 cases of orchiopexy in male patients from 6 months to 10 years of age. None of them had demonstrable hernias. All of the children underwent the orchiopexy san ligation technique of orchiopexy. The sac was opened directly while keeping the testis down and separated from the cord structures up to the deep ring. In their technique, the sac remained unligated and was pushed below the deep ring. No patients developed inguinal hernias in the follow-up period. However, they selected patients in whom the testes were palpable.4 Mohta et al. suggested not ligating the sac in pediatric inguinal hernia repair.5 A prospective study of laparoscopic inguinal hernia repair in children by Schier9 showed no difference between simple suturing and resection of the hernia sac. In addition, it was emphasized that an open internal inguinal ring is not an inguinal hernia. During laparoscopic orchiopexy, Handa et al.10 showed that closure of the internal ring is not necessary. Blackburn et al.11 incised the indirect hernia sac but did not perform high ligation in their case series of Fowler–Stephens orchiopexy. On the other hand, Varela-Cives et al.12 investigated PPV in patients with UT who underwent herniography. Ninety-five of 376 patients with UT had PPV (25.3%), and 31 of 244 patients without UT had PPV (12.7%); the difference between these two groups of patients reached statistical significance (P=0.0001). The incidence of PPV was higher at younger ages and reached 41.3% among all patients with NPT.12 In two other prospective studies of pediatric patients with UT (excluding those with clinically overt hernia and hydrocele), surgical exploration revealed a PPV prevalence of 71% and 77%, respectively.13,14 In patients with UT, the presence of PPV is related to both patient age and testis position. In patients aged <2 years, the prevalence of PPV was 68.7%. This percentage was dramatically lower at 27.2% in patients aged 2–6 years and lower still at 11.2% in patients aged 6–12 years.12 Undescended testis located cranial to the external inguinal ring was significantly more frequently associated with PPV than UT located more caudally. In concordance with the literature, we perform PV dissection in all of our patients with UT to aid the performance of the orchiopexy. Patients with NPT were specifically chosen for the present study because the incidence of PPV is higher in such cases. During the orchiopexy procedure in these patients, the PV should be dissected and high ligation should be performed to provide a tension-free repair. If an overt hernia is present, expeditious hernia repair with orchiopexy at the age of presentation is undertaken. Otherwise, the hernia should be repaired at the time of orchiopexy. Riquelme et al.15 found inguinal hernias in 23 cases (69.9%): in 16 (76.%) palpable testes and 7 (58.3%) NPT during laparoscopic orchiopexy of 64 patients. They found no inguinal hernias on the side opposite the UT.15 In our study, 7 patients were suspected to have a hernia by the family, and 22 patients with the silk sign were diagnosed by physical examination during a visit to a physician. A wide PPV was observed intraoperatively in all patients with NPT. We observed no hernias in the follow-up period after PV dissection with high ligation, suggesting the suitability of this procedure. We chose to perform open orchiopexy because we believe that thorough dissection and very high ligation are key factors in the success of the procedure. Furthermore, this approach allows for controlled incision of the PV and access to the abdomen to search for the UT. Patent PV dissection is initiated between the internal ring and testis. This enables dissection of a smaller circumference of the PV and enables better repair if the PV is inadvertently ruptured than does repair from the inner ring level. Very high ligation of the PV enables a tension-free orchiopexy outcome. Furthermore, it supports performance of the Prentiss maneuver and prevents hernia formation in patients with a wide PPV. CONCLUSION During open or laparoscopic orchiopexy, especially in patients with UT involving NPT, the PV should be dissected and high ligation should be performed to prevent hernia formation and provide a good outcome. Conflict of interest: No potential conflict of interest relevant to this article was reported. Abbreviations NPTnon-palpable testis PPVpatent processus vaginalis PVprocessus vaginalis UTundescended testis ==== Refs REFERENCES 1 Berkowitz GS Lapinski RH Dolgin SE Gazella CA Holzman IR Prevalence and natural history of cryptorchidism Pediatrics 1993 92 44 9 8100060 2 Levitt SB Kogan SJ Engel RM The impalpable testis: a rational approach to management J Urol 1978 120 515 20 30847 3 Leung AK Robson WL Current status of cryptorchidism Adv Pediatr 2004 51 351 77 15366780 4 Jain VK Singh S Garge S Joshi M Sanghvi J Orchidopexy san ligation technique of orchidopexy Afr J Paediatr Surg 2011 8 112 14 10.4103/0189-6725.79073 21478603 5 Mohta A Jain N Irniraya KP Saluja SS Sharma S Gupta A Nonligation of hernial sac during herniotomy: a prospective study Pediatr Surg Int 2003 19 451 2 10.1007/s00383-002-0940-y 12774253 6 Elder JS Laparoscopy for impalpable testes: significance of the patent processus vaginalis J Urol 1994 152 776 8 7912744 7 Barthold JS Redman JF Association of epididymal anomalies with patent processus vaginalis in hernia, hydrocele and cryptorchidism J Urol 1996 156 2054 6 10.1016/S0022-5347(01)65434-X 8911389 8 Radmayr C Corvin S Studen M Bartsch G Janetschek G Cryptorchidism, open processus vaginalis, and associated hernia: laparoscopic approach to the internal inguinal ring Eur Urol 1999 36 631 4 10.1159/000020058 10559618 9 Schier F Laparoscopic inguinal hernia repair – a prospective personal series of 542 children J Pediatr Surg 2006 41 1081 4 10.1016/j.jpedsurg.2006.02.028 16769338 10 Handa R Kale R Harjai MM Laparoscopic orchiopexy: is closure of the internal ring necessary? J Postgrad Med 2005 51 266 8 16388167 11 Blackburn SC Adams SD Mahomed AA Risk of hernia occurrence where division of an indirect inguinal sac without ligation is undertaken J Laparoendosc Adv Surg Tech A 2012 22 713 14 10.1089/lap.2012.0011 22827694 12 Varela-Cives R Bautista-Casasnovas A Taboada-Santomil P Relevance of herniography for accurate diagnosis of patent processus vaginalis in cryptorchidism Int Braz J Urol 2008 34 57 62 10.1590/S1677-55382008000100009 18341722 13 Adamsen S Aronson S Borjesson B Prospective evaluation of human chorionic gonadotropin in the treatment of cryptorchidism Acta Chir Scand 1989 155 509 14 2574929 14 Hazebroek FW de Muinck Keizer-Schrama SM van Maarschalkerweerd M Visser HK Molenaar JC Why luteinizing-hormone-releasing-hormone nasal spray will not replace orchiopexy in the treatment of boys with undescended testes J Pediatr Surg 1987 22 1177 82 10.1016/S0022-3468(87)80732-7 3326926 15 Riquelme M Aranda A Rodriguez C Cortinas J Carmona G Riquelme-Q M Incidence and management of the inguinal hernia during laparoscopic orchiopexy in palpable cryptoorchidism: preliminary report Pediatr Surg Int 2007 23 301 4 10.1007/s00383-007-1876-z 17287944

PMC005xxxxxx/PMC5001793.txt

==== Front Rambam Maimonides Med JRambam Maimonides Med JRMMJRambam Maimonides Medical Journal2076-9172Rambam Health Care Campus 10.5041/RMMJ.10248rmmj-7-3-e0021Original ResearchTransversus Abdominis Plane Versus Ilioinguinal and Iliohypogastric Nerve Blocks for Analgesia Following Open Inguinal Herniorrhaphy* Stav Anatoli M.D.12**Reytman Leonid M.D.23Stav Michael-Yohay M.S.2Troitsa Anton M.D.4Kirshon Mark M.D.4Alfici Ricardo M.D.25Dudkiewicz Mickey M.D.26Sternberg Ahud M.D.24 1 Postanesthesia Care Unit, Hillel Yaffe Medical Center, Hadera, Israel 2 The Ruth and Bruce Rappaport Faculty of Medicine, Technion–Israel Institute of Technology, Haifa, Israel 3 Department of Anesthesiology, Hillel Yaffe Medical Center, Hadera, Israel 4 Department of Surgery A, Hillel Yaffe Medical Center, Hadera, Israel 5 Department of Surgery B, Hillel Yaffe Medical Center, Hadera, Israel 6 Director-General, Hillel Yaffe Medical Center, Hadera, Israel** To whom correspondence should be addressed. E-mail:anatoli99@bezeqint.net7 2016 28 7 2016 7 3 e0021Copyright: © 2016 Stav et al.2016This is an open-access article. All its content, except where otherwise noted, is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.Objectives We hypothesized that preoperative (pre-op) ultrasound (US)-guided posterior transversus abdominis plane block (TAP) and US-guided ilioinguinal and iliohypogastric nerve block (ILI+IHG) will produce a comparable analgesia after Lichtenstein patch tension-free method of open inguinal hernia repair in adult men. The genital branch of the genitofemoral nerve will be blocked separately. Methods This is a prospective, randomized, controlled, and observer-blinded clinical study. A total of 166 adult men were randomly assigned to one of three groups: a pre-op TAP group, a pre-op ILI+IHG group, and a control group. An intraoperative block of the genital branch of the genitofemoral nerve was performed in all patients in all three groups, followed by postoperative patient-controlled intravenous analgesia with morphine. The pain intensity and morphine consumption immediately after surgery and during the 24 hours after surgery were compared between the groups. Results A total of 149 patients completed the study protocol. The intensity of pain immediately after surgery and morphine consumption were similar in the two “block” groups; however, they were significantly decreased compared with the control group. During the 24 hours after surgery, morphine consumption in the ILI+IHG group decreased compared with the TAP group, as well as in each “block” group versus the control group. Twenty-four hours after surgery, all evaluated parameters were similar. Conclusion Ultrasound-guided ILI+IHG provided better pain control than US-guided posterior TAP following the Lichtenstein patch tension-free method of open inguinal hernia repair in men during 24 hours after surgery. (ClinicalTrials.gov number: NCT01429480.) Analgesianerve blockopen herniorrhaphypainpost-opultrasound ==== Body INTRODUCTION The Lichtenstein patch tension-free method of open inguinal hernia repair1 is commonly used under general anesthesia with the facilitation of ilioinguinal and iliohypogastric nerve blocks (ILI+IHG) to improve postoperative (post-op) analgesia.2 The PROSPECT Working Group (http://www.postoppain.org/working-group/) published a procedure-specific review and guideline for analgesia following inguinal hernia repair in 2012, in which ILI+IHG were strongly recommended preoperatively (pre-op) and during surgery (grade A recommendation based on randomized clinical trials).3 According to the new guideline published in February 2016, the use of peripheral regional anesthetic techniques is strongly recommended as a component of multimodal analgesia for pain management following open inguinal hernia repair.4 Ultrasound (US)-guided posterior transversus abdominis plane block (TAP) has been described as an appropriate method of post-op analgesia for abdominal wall incisions below the umbilicus. Inguinal hernia repair was also included.5–7 Aveline et al. compared the efficacy of pre-op US-guided TAP to conventional (landmark-guided technique) ILI+IHG and concluded that better post-op pain relief was achieved following US-guided TAP.8 Fredrickson et al. concluded that, in children, US-guided ILI+IHG provides more effective analgesia compared with US-guided TAP following inguinal herniotomy, hydrocelectomy, and orchiopexy.9 Wang et al. conducted a meta-analysis to evaluate the clinical efficacy of US-guided ILI+IHG and US-guided TAP for perioperative analgesia in patients undergoing open inguinal surgery; however, only four articles were included in the meta-analysis:10 two randomized controlled trials were performed in pediatric patients,11,12 and two trials were conducted in adults.8,13 The authors concluded that US-guided ILI+IHG or TAP is associated with improved perioperative analgesia compared with the landmark-based technique. Anatomically, sensory innervation of the male inguinal region is from the Th12–L2 nerves, and “the spermatic sympathetic plexus contains the sensory fibers for the testis.”9 The ilioinguinal and iliohypogastric nerves (branches of Th12 and L1) pass between the internal oblique and transversus abdominis muscles at the level immediately superior to the anterior superior iliac spine. More medial ilioinguinal and iliohypogastric nerves pierce the internal oblique abdominal muscle and lie near the internal inguinal ring between it and the external oblique muscles.9 The ilioinguinal and iliohypogastric nerves innervate part of the structures in the inguinal canal and lie on the anterior surface of the spermatic cord.9 The genital branch of the genitofemoral nerve (L1, L2) lies on the posterior aspect of the spermatic cord after entering the inguinal canal through the internal inguinal ring;9 it provides motor and sensory innervation to the cremaster muscle and anterolateral aspect of the scrotum in men.14 Although ILI+IHG has been shown to provide insufficient analgesia, the addition of a genitofemoral nerve block may improve the analgesic effect.15 According to the described technique for both evaluated blocks, local anesthetic should be injected into the fascial plane between the internal oblique and transversus abdominis muscles.5–7,16 There is a difference in the volume and anatomical point of injection. The recommended volume of injected local anesthetic for TAP is 20–30 mL, as compared with 10 mL for ILI+IHG.16 The point of needle insertion for an US-guided TAP is more proximal compared with US-guided ILI+IHG.5–7,16 We suggest that the ilioinguinal and iliohypogastric nerves may be appropriately blocked using both US-guided techniques, TAP and ILI+IHG, since both techniques produce the same block of the ilioinguinal and iliohypogastric nerves; the only difference is that TAP is a compartment block, while ILI+IHG blocks the truncal. In order to compare the effect of TAP and ILI+IHG on post-op pain intensity, pain must be eliminated from the structures innervated by the genital branch of the genitofemoral nerve; an intraoperative block of the genital branch of the genitofemoral nerve should be used in all patients; and the same pharmacological treatment should be offered for post-op pain. Our model includes post-op, patient-controlled analgesia with intravenous morphine (PCA). No previous trials have compared US-guided TAP versus US-guided ILI+IHG in adult men undergoing inguinal hernia repair using the open Lichtenstein patch tension-free method. We tested the hypothesis that both US-guided TAP and US-guided ILI+IHG provide comparable post-op analgesia, together with morphine consumption, following Lichtenstein patch hernioplasty during the first 24 hours post-op. METHODS This is a prospective, randomized, controlled, and observer-blinded clinical study; enrollment began after receiving Hillel Yaffe Review Board approval and written informed patient consent. Study eligibility was as follows: male; over the age of 18; physical status of I–III based on the American Society of Anesthesiologists criteria; and scheduled to undergo elective inguinal hernioplasty using the unilateral open Lichtenstein patch tension-free method. The rationale only to include males in the trial was based on the aim to compare data among maximally identical groups. In total, 186 patients were eligible for the study. Exclusion criteria were recurrent hernia, previous episode of incarceration, inguino-scrotal hernia, or sliding inguinal hernia.17 It should be noted that manipulating sliding hernias during an open herniorrhaphy can produce visceral pain, whereas pre- or intraoperative blocks of the somatic peripheral nerves (i.e. TAP, ILI+IHG, and genital branch of the genitofemoral nerve) do not cause visceral pain. Additional exclusion criteria included being under the age of 18, presence of a local skin infection near the block injection site, allergy to local anesthetics, demonstrated opioid dependency, international normalized ratio (INR) greater than 1.4, platelet count less than 100,000, chronic pain, dementia, or an inability to comprehend the pain scale or use the PCA device. Patients were instructed regarding the use of a 100 mm visual analogue scale (VAS) graded from 0 (without pain) to 100 (intolerable pain). Based on the exclusion criteria, 20 patients were excluded from the study. The inclusion group (166 adult male patients) was randomized into three groups: TAP, ILI+IHG, and control (Figure 1). Randomization was done using a computer-generated table of random numbers, placing them in a sealed envelope, and then opening the envelope on the morning of surgery. Patients in the control group did not undergo a pre-op peripheral nerve block; however, the surgeon intraoperatively blocked the genital branch of the genitofemoral nerve in all patients in all three groups. Figure 1 CONSORT Flow Chart. *TAP group, transversus abdominis plane block group; **ILI+IHG, ilioinguinal and iliohypogastric nerve blocks group; +GFNB, genitofemoral nerve block; ++PCA, patient-controlled analgesia with morphine. The patients and the investigator who collected the data during the post-op period were blinded to the group assignments throughout the trial. Premedication was limited to intravenous (IV) fentanyl (0.5 μg/kg), midazolam (0.03 mg/kg), and local anesthesia via injection of lidocaine (3–5 mL; 10 mg/mL). Standard American Society of Anesthesiologists monitoring and supplemental oxygen (40% via mask; 5 L/min) were applied throughout the procedure. The same anesthesiologists (A. Stav or L.R.) performed all pre-op US-guided peripheral nerve blocks using an ultrasound system (SonoSite S-Nerve, SonoSite, Bothell, WA, USA) with a 6–13 MHz linear array transducer (L25x). A 22-gauge 80 mm Pajunk Sono Tap cannula (Pajunk® Medical Produkte GmbH, Geisingen, Germany) was used to perform the blocks. All pre-op blocks were performed using aseptic techniques in the “block room.” The local anesthetic injected for the nerve blocks comprised bupivacaine 5 mg/mL with adrenaline 5 μg/mL. The same volume of 20 mL of local anesthetic was injected for TAP and ILI+IHG to ensure similar local anesthetic doses and volumes, and to facilitate appropriate comparison of perioperative pain and morphine consumption between the groups. The TAP and ILI+IHG procedures were conducted using techniques described by Hadzic.16 The genital branch of the genitofemoral nerve was blocked intraoperatively in all patients in all three groups by the surgeon during exploration of the spermatic cord under direct vision at the level of the internal inguinal ring. “All the genital branches passed through the ventral aspect of the internal ring, entered the ring and continued within the spermatic cord.”17 The block was induced by injecting 10 mL of bupivacaine 5 mg/mL without adrenaline. The success of the sensory block was tested by pin-prick in the inguinal crease (from the projection of the internal inguinal ring and medial up to the projection of the external ring) 30 min after completion of the pre-op procedure (both groups) by an independent investigator who was blinded to the previously performed block. Patients were additionally excluded from the study if a failed block was diagnosed (normal sensation). A standardized general anesthesia protocol was used for all patients. Lornoxicam 8 mg IV was injected 30 min before the end of surgery. All operations were performed by or under the supervision of the same experienced surgeons (A.T. or M.K.). Morphine was initiated via PCA IV in the postanesthesia care unit (PACU) for all patients. The PCA protocol was as follows: a loading bolus of 2 mg IV morphine; 1 mg for the subsequent bolus, followed by a lockout period of 5 min, with a maximal dosage of 50 mg within a 4-hour period. Patient-controlled analgesia continued post-op for 24 hours. Variables assessed before and during surgery included patient characteristics (age, height, and body weight (BW)). Body mass index (BMI) was calculated using the following formula: BMI = BW (kg)/height2 (m2). Preoperative pain at rest and during motion was measured using a visual analogue scale (VAS).18 If the oxygen saturation decreased during surgery, or bradycardia, low blood pressure, or other events persisted for more than 2 min, the patient was excluded from the study. The aim of this exclusion was to remove potential temporary intraoperative brain tissue oxygen desaturation, which may influence mental state19 and pain intensity estimation during the post-op period when assessed by VAS. Pain intensity in the PACU was measured using a modified nurses’ assessment of postoperative pain scale (NP):20 0=no pain, or the patient is asleep; 1=mild pain; 2=moderate pain; 3=severe pain; 4=intolerable pain. If moderate, severe, or intolerable pain was experienced by the patient in the PACU immediately post-op, then morphine was initially injected using a titration method until pain relief was achieved,21,22 followed by PCA. Morphine consumption in the PACU, post-op nausea and vomiting (PONV) (yes/no), and time of stay were assessed and statistically compared between the groups. In addition, data were collected after discharge from the PACU, i.e. during post-op day 0 regarding pain intensity at rest and during motion (measured by VAS), morphine consumption after PACU discharge and up to 24 hours post-op, and other analgesic consumption (oral dipyrone (metamizole), IV or oral paracetamol (acetaminophen), or non-steroidal anti-inflammatory drugs). The investigator collecting the data 24 hours after PACU discharge was blinded regarding the type of previously performed block. The primary end-point of the study included three parameters: pain intensity immediately post-op, morphine and other analgesic consumption 24 hours post-op, and pain intensity following termination of the effect of the peripheral nerve blocks in the “block” groups, i.e. 24 hours post-op. The parameters for the secondary end-point were: PONV (yes/no), PACU time of stay, and the correlation between block performance time and the patients’ body mass index (BMI). Statistical Analysis A priori power analysis was performed using “G* Power 3.0.10”© with a fixed effects, omnibus, one-way ANOVA for all groups. A total sample size of 66 was considered adequate to achieve an effect size of 0.5 with an α error probability of 0.05 and a power (1-β error probability) of 0.95. After all exclusions, our study included 149 patients. Statistical analysis was performed using “IBM SPSS Statistics 20”©. Continuous numerical parameters were analyzed according to the Shapiro–Wilk test for distribution normality, followed by the Levene test for homogeneity of the variances (if a normal distribution was determined). Parameters with a normal distribution and homogeneous variances were compared by one-way ANOVA followed by Tukey’s post-hoc test, if necessary. The Kruskal–Wallis test was used when an abnormal distribution of the continuous variables was detected. The Mann–Whitney post-hoc test was used following the Kruskal–Wallis test, if necessary. The intensity of VAS-measured pain was also analyzed using the Kruskal–Wallis test followed by the Mann–Whitney test, if necessary. Frequency tables and Pearson chi-square tests were used to compare the proportions between the categorical variables among the groups. Pearson correlation was used to assess the correlation between BMI and block performance time in each group. A value of P<0.05 was considered statistically significant. RESULTS Study data were collected between October, 2011 and September, 2015. The results of our trial are presented according to the standards of reporting Clinical Trials (CONSORT) statement using a CONSORT flow chart (Figure 1). The data obtained from 149 patients were statistically analyzed until the end of the study. The pre-op variables were comparable among the groups (Table 1). There were no differences in surgery duration, and there was no correlation between the block performance time and BMI. There were no complications during or immediately after the blocks in the TAP and ILI+IHG groups. Table 1 Preoperative Data. TAP (n=50) ILI+IHG Group (n=49) Control Group (n=50) P Age (years) 50±17 46±19 49±16 NS Height (m) 1.75±0.07 1.73±0.07 1.73±0.08 NS BW (kg) 77.83±12.15 77.24±10.79 74.38±11.92 NS BMI 25.47±3.55 25.84±3.67 24.66±3.44 NS Pain* (rest) in mm 6.54±12.84 9.12±19.68 5.34±13.56 NS Pain* (motion) in mm 39.66±27.51 28.61±29.17 29.60±32.10 NS Block performance time (min) 11.48±4.82 11.80±4.07 NS All values are presented as mean±SD. * Pain intensity at rest and during motion were measured by visual analogue scale (VAS) in mm. BMI, body mass index; ILI+IHG, ilioinguinal and iliohypogastric nerve blocks group; NS, no statistically significant difference between groups; TAP, transversus abdominis plane block group. Pain intensity in the PACU and morphine consumption was not significantly different between the TAP and ILI+IHG groups; however, they were significantly decreased when compared with the control group (Table 2). This was one parameter of the primary end-point of our study. Table 2 Pain Intensity and Drug Consumption. TAP (n=50) ILI+IHG (n=49) CG (n=50) P value TAP versus ILI+IHG TAP versus CG ILI+IHG versus CG Pain intensity (by NP scale) in PACU 0.72 0.86 1.6 0.373 0.001 0.008 Morphine consumption (mg) in PACU 4.24 3.94 7.08 0.573 0.049 0.017 Pain at rest, 24 h post operation by visual analogue scale (mm, 1–100) 37.46 30.84 35.28 NS NS NS Pain at motion, 24 h post operation by visual analogue scale (mm, 1–100) 46.37 39.87 51.21 NS NS NS Dipyrone consumption during 24 h after operation 327 41 380 <0.0001 0.37 <0.0001 Morphine consumption, first day after PACU discharge 18.76 10.69 18.12 0.004 0.901 0.001 Control group: Patients in this group received intraoperative genitofemoral block only. CG, control group; ILI+IHG, Ilioinguinal and iliohypogastric nerve blocks group; NP, modified nurses’ assessment of postoperative pain scale; PACU, post anesthesia care unit; TAP, transversus abdominis plane block group. See text for details. Prophylaxis of PONV with 8 mg ondansetron IV prior to induction of anesthesia was effective. No PONV occurred in the 147 evaluated patients. One patient in the TAP group and one patient in the control group suffered from nausea, but no vomiting occurred during the immediate post-op period or during the 24 hours post-op. This was the secondary end-point of the trial. The duration of PACU stay was similar in all groups. Following PACU discharge, the patients in all groups were treated with PCA. Pain intensity (VAS) at rest and during motion 24 hours post-op was similar in all groups (Table 2). Morphine and dipyrone consumption following PACU discharge and up to 24 hours post-op were significantly decreased in the ILI+IHG group compared with the other two groups. There were no differences between the TAP and the control group (Table 2). Analgesic consumption during the 24 hours post-op was another parameter of the primary end-point of our trial. DISCUSSION Acute pain following open hernia repair is maximal on the day of surgery23 (the Lichtenstein mesh repair method was included in another trial of Callesen et al.24). Both regional anesthesia methods (TAP and ILI+IHG) have been previously recommended as components of multimodal analgesia in the post-op period to induce a significant reduction in pain during the initial 24 hours following open hernia repair.2–6 This raises the question as to which block should be used in specific cases. In contrast to Aveline et al.,8 comparing US-guided TAP with landmark-guided ILI+IHG, we compared post-op pain relief and morphine consumption immediately after surgery and during the 24 hours post-op between two US-guided techniques: posterior TAP versus ILI+IHG. The TAP and ILI+IHG approaches represent a significant difference between investigations, with subsequent differences in the results because visualization of target structures enabled the possibility of accurately injecting the local anesthetic into the transversus abdominis plane (in TAP group) or around the ilioinguinal and iliohypogastric nerves (ILI+IHG group). Sasaoka et al.15 in 2005 and Asad et al. in 200925 have indicated that the addition of a genitofemoral nerve block (genital branch of the genitofemoral nerve) to ILI+IHG provides hemodynamic stability during sac traction in children following hernia repair, i.e. a combination of the three nerve blocks provides better pain relief compared with ILI+IHG alone. For appropriate comparisons of the post-op pain relief and morphine consumption, we blocked the genital branch of the genitofemoral nerve in all patients in all groups. This approach enables neutralization of painful impulses originating from structures innervated by the genital branch of the genitofemoral nerve. This is a significant strength of our trial and a critical difference between our investigation and that of Aveline et al.8 Our inclusion and exclusion criteria enabled the selection and comparison of similar groups of patients with regard to the pre-op data, gender (all patients were men), and patients suffering from primary inguinal hernia (but not from recurrent, sliding, or inguinoscrotal huge hernia). Primary open hernioplasty via the Lichtenstein tension-free technique was used in all patients. The same concentration and volume of local anesthetic was used for all patients in the two pre-op block groups. All pre-op blocks were performed by the same experienced anesthesiologists; an intraoperative block of the genital branch of the genitofemoral nerve was performed or controlled by the same experienced surgeons. A very high level of similarity among all three groups is a strong advantage of our trial as compared to other similar studies.8–10,13,15,25 In our prospective randomized controlled and observer-blinded trial, pain intensity and morphine consumption immediately post-op (in the PACU) were significantly less in the TAP and ILI+IHG groups compared with the control group; however, there were no differences between the TAP and ILI+IHG groups themselves. This finding indicates that both methods of US-guided blocks produce significant post-op pain relief immediately after surgery, with no significant difference between the procedures. Twenty-four hours post-op, when the action of bupivacaine has decreased significantly or has ceased, no differences in pain intensity were observed among the all groups. Total morphine and dipyrone consumption in the ILI+IHG group was significantly decreased compared with the TAP and control groups; however, the TAP and control groups were similar. This finding indicates that the ILI+IHG produces superior pain relief compared to TAP during the 24 hours post-op, but not immediately after surgery. Our results may be explained by the hypothesis that the spread of local anesthetic during and after injection differs following TAP and ILI+IHG. Posterior TAP is a compartment block, and the spread of local anesthetic is relatively large, i.e. from Th7 to L1.5,26 Following ILI+IHG, the local anesthetic only spreads around the target nerves. We compared the groups with the same injected volume and concentration of bupivacaine (20 mL of 0.5%); thus, we can assume that following ILI+IHG a relatively high quantity of bupivacaine had spread around the target of the two nerves compared with the TAP block. It is possible that the quality of the compared blocks is similar immediately post-op (in the PACU); however, when the quantity of the local anesthetic around and in the nerves decreases, the quality of the blocks changes, i.e. the quality of the ILI+IHG improves. This is one potential explanation for the difference in morphine consumption over 24 hours between the TAP and ILI+IHG groups. Our study has several limitations as follows: There were strong inclusion criteria: limited to men, primary inguinal hernia, and hernioplasty using the Lichtenstein tension-free method. Thus, the conclusions of the trial are relevant for a specific pathology, gender, and method of operation. We used two scales for the pain intensity measurement, including the modified NP scale20 and the VAS.18 Readers should note this difference in the presented results. Lornoxicam 8 mg IV was injected 30 min before the hernioplasty ended; dipyrone was used as needed to treat mild pain or discomfort during the post-op period; those agents may have influenced pain intensity and morphine consumption. It is impossible to examine the individual anatomical variations of the ilioinguinal and iliohypogastric nerves pre-op. These potential variations have been examined and described by Klaassen et al.27 In this work the authors included the observation of the spinal nerve contribution, as well as the communications with other nerves (such as the accessory nerve, subcostal nerves, and lateral femoral cutaneous nerve). Anatomical variations may explain the partial sensory block following partial analgesia in the evaluated patients after both the TAP and ILI+IHG nerve block, as well as after surgery. CONCLUSIONS The data from this prospective, randomized, controlled, and observer-blinded clinical study lead to several important conclusions. Both regional anesthesia methodologies (US-guided TAP and US-guided ILI+IHG) may be used as components of multimodal anesthesia following a primary unilateral open Lichtenstein patch tension-free inguinal hernioplasty in adult men. Both blocks provided comparative analgesia, as well as a similar morphine-sparing effect immediately after the surgery, which was performed under general anesthesia. During the 24 hours after surgery, US-guided ILI+IHG (a truncal block) was superior to US-guided TAP (a compartment block) in terms of the morphine-sparing effect, i.e. ILI+IHG provided significantly better analgesia. Prophylaxis of PONV with 8 mg ondansetron IV prior to anesthesia induction was effective. The duration of PACU stay was similar in all three groups. There was no correlation between the time of block performance and BMI. Potential anatomical variations in the ilioinguinal and iliohypogastric nerves, which are impossible to diagnose prior to surgery, could play a role in the intensity of post-op acute pain. This was a significant limitation of the trial. Acknowledgments The authors would like to acknowledge the PACU nursing stuff of Hillel Yaffe Medical Center for practical assistance in performing the blocks, and the residents of the Department of Anesthesiology for their practical assistance of data collection, especially intraoperative data collection. * A poster presentation with partial results of this study was presented at the 26th Congress of the European Federation of Societies for Ultrasound in Medicine and Biology (EFSUMB) EUROSON 2014, Tel Aviv, Israel. Conflict of interest: No potential conflict of interest relevant to this article was reported. Abbreviations BMIbody mass index BWbody weight ILI+IHGilioinguinal and iliohypogastric nerve block IVintravenously NP scalenurses’ assessment of postoperative pain scale PACUpostanesthesia care unit PCApatient-controlled analgesia with intravenous morphine PONVpost-op nausea and vomiting pre-oppreoperative or preoperatively post-oppostoperative or postoperatively PROSPECTprocedure-specific post-op pain management TAPposterior transversus abdominis plane block USultrasound ==== Refs REFERENCES 1 Amid PK Shulman AG Lichtenstein IL Open “tension-free” repair of inguinal hernias: the Lichtenstein technique Eur J Surg 1996 162 447 53 8817221 2 White PF Kehlet H Improving post-op pain management: what are the unresolved issues Anesthesiology 2010 112 220 5 10.1097/ALN.0b013e3181c6316e 20010418 3 Joshi GP Rawal N Kehlet H PROSPECT collaboration. Evidence-based management of postoperative pain in adults undergoing open inguinal hernia surgery Br J Surg 2012 99 168 85 10.1002/bjs.7660 21928388 4 Chou R Gordon DB de Leon-Casasola OA Management of postoperative pain: a clinical practice guideline from the American Pain Society, the American Society of Regional Anesthesia and Pain Medicine, and the American Society of Anesthesiologists’ Committee on Regional Anesthesia, Executive Committee, and Administrative Council J Pain 2016 17 131 57 10.1016/j.jpain.2015.12.008 26827847 5 Tran TM Ivanusic JJ Hebbard P Barrington MJ Determination of spread of injectate after ultrasound-guided transversus abdominis plane block: a cadaveric study Br J Anaesth 2009 102 123 7 10.1093/bja/aen344 19059922 6 Mukhtar K Transversus abdominis plane (TAP) block Journal of the New York School of Regional Anesthesia 2009 12 28 33 Available at: http://bit.ly/2a9M8lc accessed July 17, 2016 7 Jankovic Z Transversus abdominis plane block: the holy grail of anaesthesia for (lower) abdominal surgery Period Biol 2009 111 203 8 8 Aveline C Le Hetet H Le Roux A Comparison between ultrasound-guided transversus abdominis plane and conventional ilioinguinal/iliohypogastric nerve blocks for day-case open inguinal hernia repair Br J Anaesth 2011 106 380 6 10.1093/bja/aeq363 21177284 9 Fredrickson MJ Paine C Hamill J Improved analgesia with the ilioinguinal block compared to the transversus abdominis plane block after pediatric inguinal surgery: a prospective randomized trial Pediatr Anesth 2010 20 1022 7 10.1111/j.1460-9592.2010.03432.x 10 Wang Y Wu T Terry MJ Improved perioperative analgesia with ultrasound-guided ilioinguinal/iliohypogastric nerve or transversus abdominis plane block for open inguinal surgery: a systematic review and meta-analysis of randomized controlled trials J Phys Ther Sci 2016 28 1055 60 10.1589/jpts.28.1055 27134411 11 Willschke H Marhofer P Bösenberg A Ultrasonography for ilioinguinal/iliohypogastric nerve blocks in children Br J Anaesth 2005 95 226 30 10.1093/bja/aei157 15923270 12 Nan Y Zhou J Ma Q Li T Lian QQ Li J Application of ultrasound guidance for ilioinguinal or iliohypogastric nerve block in pediatric inguinal surgery Zhonghua Yi Xue Za Zhi 2012 92 873 7 [Article in Chinese, Abstract in English] 22781525 13 Demirci A Efe EM Türker G Iliohypogastric/ilioinguinal nerve block in inguinal hernia repair for postoperative pain management: comparison of the anatomical landmark and ultrasound guided techniques Rev Bras Anestesiol 2014 64 350 6 [Article in Portuguese, Abstract in English] 25168440 14 Moore DC Block of the Inguinal Region Moore DC Regional Block 4th ed Springfield, IL Charles C Thomas 1981 167 73 15 Sasaoka N Kawaguchi M Yoshitani K Kato H Suzuki A Furuya H Evaluation of genitofemoral nerve block, in addition to ilioinguinal and iliohypogastric nerve block, during inguinal hernia repair in children Br J Anaesth 2005 94 243 6 10.1093/bja/aei031 15567812 16 Hadzic A Common Ultrasound-Guided Truncal and Cutaneous Blocks Hadzic A Hadzic’s Peripheral Nerve Blocks and Anatomy for Ultrasound-Guided Regional Anesthesia 2nd ed New York McGraw-Hill Companies, Inc. 2012 459 65 17 Liu WC Chen TH Shyu JF Applied anatomy of the genital branch of the genitofemoral nerve in open inguinal herniorrhaphy Eur J Surg 2002 168 145 9 10.1080/110241502320127748 12182238 18 Wewers ME Lowe NK A critical review of visual analogue scales in the measurement of clinical phenomena Res Nurs Health 1990 13 227 36 10.1002/nur.4770130405 2197679 19 Suehiro K Okutai R Duration of cerebral desaturation time during single-lung ventilation correlates with mini mental state examination score J Anesth 2011 25 345 9 10.1007/s00540-011-1136-1 21484501 20 Chung IS Sim WS Kim GS Nurses’ assessment of postoperative pain: can it be alternative to patients’ self-report? J Korean Med Sci 2001 16 784 8 10.3346/jkms.2001.16.6.784 11748363 21 Aubrun F Monsel S Langeron O Coriat P Riou B Postoperative titration of intravenous morphine Eur J Anaesthesiol 2001 18 159 65 10.1097/00003643-200103000-00004 11298174 22 Birnbaum A Esses D Bijur PE Holden L Gallagher EJ Randomized double-blind placebo-controlled trial of two intravenous morphine dosages (0.10 mg/kg and 0.15 mg/kg) in emergency department patients with moderate to severe acute pain Ann Emerg Med 2007 49 445 53 10.1016/j.annemergmed.2006.06.030 16978739 23 Callesen T Bech K Nielsen R Pain after groin hernia repair Br J Surg 1998 85 1412 14 10.1046/j.1365-2168.1998.00864.x 9782027 24 Callesen T Bech K Andersen J Nielsen R Roikjaer O Kehlet H Pain after primary inguinal herniorrhaphy: influence of surgical technique J Am Coll Surg 1999 188 355 9 10.1016/S1072-7515(98)00316-0 10195718 25 Asad N Afzal F Khawar AA Hanif AA Saleem H Hussain SM Genitofemoral nerve block and intraoperative analgesia in children during inguinal hernia repair Biomedica 2009 25 146 9 26 McDonnell JG O’Donnell B Curley G Heffernan A Power C Laffey JG The analgesic efficacy of transversus abdominis plane block after abdominal surgery: a prospective randomized controlled trial Anesth Analg 2007 104 193 7 10.1213/01.ane.0000250223.49963.0f 17179269 27 Klaassen Z Marshall E Tubbs RS Louis RG Wartmann CT Loukas M Anatomy of the ilioinguinal and iliohypogastric nerves with observations of their spinal nerve contributions Clin Anat 2011 24 454 61 10.1002/ca.21098 21509811

PMC005xxxxxx/PMC5001794.txt

==== Front Rambam Maimonides Med JRambam Maimonides Med JRMMJRambam Maimonides Medical Journal2076-9172Rambam Health Care Campus 10.5041/RMMJ.10249rmmj-7-3-e0022Original ResearchSafety of Completion Thyroidectomy for Initially Misdiagnosed Thyroid Carcinoma Kranthikumar Gangiti M.S.Syed Nusrath M.S., D.N.B.Nemade Hemantkumar M.S., F.H.N.S.O.*Pawar Satish M.S., D.N.B.Chandra Sekhara Rao L. M. S.M.S., D.N.B., P.D.C.C., M.Ch.Subramanyeshwar Rao T. M.S., M.Ch.Department of Surgical Oncology, Basavatarakam Indo-American Cancer Hospital and Research Institute, Hyderabad, India* To whom correspondence should be addressed. E-mail:drhemantnemade@gmail.com7 2016 28 7 2016 7 3 e0022Copyright: © 2016 Kranthikumar et al.2016This is an open-access article. All its content, except where otherwise noted, is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.Introduction Completion thyroidectomy is defined as the surgical removal of the remnant thyroid tissue following procedures of less than total or near-total thyroidectomy. Whether thyroid reoperations are associated with an increased complication risk is controversial. Objective A retrospective analysis was done of patients undergoing completion thyroidectomy for cancer of the thyroid who had undergone surgery elsewhere for solitary thyroid nodule. The incidence of surgical complications in these patients after reoperation was investigated in this study. Material and methods The study included a total of 53 patients who had undergone thyroid lobectomy for a solitary nodule as initial surgery elsewhere and were referred to our institute for completion thyroidectomy when the histopathology revealed malignancy. Results There were 53 patients, 43 females and 10 males. Their mean age was 34.7±12.12 years (range 19–65 years). After initial surgery, the histopathology revealed papillary carcinoma in 46 patients (86.8%), follicular carcinoma in 7 (13.2%). Fourteen out of 53 patients had recurrent laryngeal nerve palsy after initial surgery (26.4%). None of the patients had clinical hypocalcemia after the first surgery. One or more parathyroid glands were identified and preserved in 52 patients (98.1%) in the process of completion thyroidectomy. No patient had additional recurrent nerve injury at the second surgery. The mean serum calcium value preoperatively was 8.96±0.39 mg/dL, and six months after surgery serum calcium was 8.74±0.56 mg/dL. Mean follow-up was 18 months. Transient hypoparathyroidism occurred in 24.5% patients. Five patients were lost to follow-up. Permanent and symptomatic hyperparathyroidism occurred in eight patients (16.67%). Conclusions Completion thyroidectomy is a safe and appropriate option in the management of well-differentiated thyroid cancer. It removes disease on the ipsilateral and contralateral side of the thyroid and carries a low risk of recurrent laryngeal nerve damage, but a higher risk of permanent hypoparathyroidism. Completion thyroidectomyhypoparathyroidismrecurrent laryngeal nerve palsyrevision thyroid surgerythyroid cancer ==== Body INTRODUCTION Solitary thyroid nodule is one of the common presentations in a thyroid clinic. The usual evaluation for the solitary thyroid nodule includes fine-needle aspiration cytology (FNAC), which in many cases is inconclusive. This mandates hemithyroidectomy followed by completion thyroidectomy in the majority of the cases of well-differentiated carcinoma. Ultrasonography and FNAC has led to a 35%–75% decrease in the number of patients requiring surgery.1–4 As a result, there has been a substantial increase in the proportion of malignant lesions in surgical cases. Completion thyroidectomy is traditionally believed to increase complication rates, but many studies contradict this belief.5–13 We analyzed our data to discover whether completion thyroidectomy increased the incidence of recurrent nerve palsy and parathyroid gland-related complications. Furthermore, we investigated the data in terms of ipsilateral and contralateral sides (i.e. the previously operated side and the side untouched in the previous surgery, respectively). MATERIAL AND METHODS Data were obtained for patients who underwent surgery for solitary thyroid nodule elsewhere, including from nursing homes with surgeons not trained in head and neck oncology, and were referred to our institute for completion thyroidectomy after a postoperative histological diagnosis of cancer. Some patients were referred specifically because of the complications during previous surgery, such as recurrent laryngeal nerve palsy. Medical records and details of 53 such patients who met the above criteria and underwent completion thyroidectomy between June 2012 and June 2014, with a mean follow-up of 18 months and a minimum of one year, were collected. All patients underwent thorough clinical examination and neck ultrasonography, thyroid hormone assessment, and indirect laryngoscopy before the second operation. All the histological slides and blocks of the first surgery were reviewed by the pathology experts. Only patients with confirmed diagnosis of papillary thyroid carcinoma with tumor >2 cm, and all follicular carcinomas, were enrolled in the study of patients who underwent completion thyroidectomy. Serum calcium levels were determined preoperatively and postoperatively after 48 hours. All symptomatic patients were started on oral and intravenous calcium supplementation, and asymptomatic patients with hypocalcemia signs were started on oral calcium supplementation. All patients underwent vocal cord assessment after the second surgery at the time of discharge. Postoperatively all patients with papillary and follicular carcinomas were referred for radioiodine ablation. Thereafter patients were followed every six months with serum thyroid-stimulating hormone, serum thyroglobulin, serum antithyroglobulin, and serum calcium. Surgical Technique A single dose of antibiotic was administered just before the incision. The previous incision was revised and extended if necessary. After elevating subplatysmal flaps and retracting strap muscles in the midline, both sides of the thyroid or thyroid bed and the isthmus were carefully evaluated for any thyroid tissue. Meticulous dissection was performed, sometimes approaching the thyroid gland laterally from the strap muscles. Surgical loupes were used during surgery. A nerve monitor was not used. The superior laryngeal nerve, recurrent laryngeal nerve (RLN), and parathyroid glands were identified before resecting thyroid tissue. Where required, appropriate neck dissection was done. The central compartment was dissected only for locally advanced disease or when preoperative ultrasonography or intraoperative assessment revealed suspicious nodes in that region. At the end of the procedure if the parathyroid tissue appeared devascularized it was implanted in the sternocleidomastoid muscle. RESULTS There were 53 patients, 43 women and 10 men. Their mean age was 34.7 years (range 19–65 years) (Table 1). The period between initial and second surgery ranged from 10 days to 150 days, with a median of 60 days. Mean operating time was 68 min. After initial surgery, the histopathology revealed papillary carcinoma in 47 patients (86.8%) and follicular carcinoma in 6 (13.2%). Fourteen out of 53 patients had recurrent laryngeal nerve palsy after initial surgery (26.4%). One or more parathyroid glands were identified and preserved in 52 patients (98%); parathyroids could not be identified in one patient. No patient had additional recurrent nerve injury at the second surgery. The mean serum calcium value preoperatively was 8.96±0.39, and postoperative serum calcium was 8.74±0.56 after six months. No patient required readmission. Mean follow-up was 18 months. Transient hypoparathyroidism occurred in 13 (24.5%) patients. Five patients were lost to follow-up. Permanent and symptomatic hypoparathyroidism occurred in eight patients (8/48 patients, 16.67%) (Tables 2 and 3). Table 1 Patient Demographics. Patient and Surgery Details n Age Group  <20 2  20–40 38  40–60 10  >60 3 Sex  Male 10  Female 43 Surgery  Subtotal thyroidectomy, B/L MRND [including B/L paratracheal (central compartment) node dissection] 12  Subtotal thyroidectomy 13  Lobectomy, B/L paratracheal (central compartment) node dissection 10  Hemithyroidectomy only 4  Lobectomy only 14  Pathology  Papillary carcinoma 47  Follicular carcinoma 6 B/L, bilateral; MRND, modified radical neck dissection. Table 2 Complications. Complications n (%) After First Surgery  RLN palsy 14 (26.4%)  Hypocalcemia 0 After Second Surgery  Additional RLN palsy 0  Lost to follow-up 5 (9.4%)  Transient hypoparathyroidism 13/53 (24.5%)  Permanent hypoparathyroidism 8/48 (16.67%)  Mortality 0 RLN, recurrent laryngeal nerve. Table 3 Second Surgery Type and Hypothyroidism. Surgery Cases of Permanent Hypoparathyroidism (n) Subtotal thyroidectomy, B/L MRND [including B/L paratracheal (central compartment) node dissection] 2 Subtotal thyroidectomy 1 Lobectomy, B/L paratracheal (central compartment) node dissection 3 Hemithyroidectomy only 1 Lobectomy only 1 B/L, bilateral; MRND, modified radical neck dissection. There was no mortality and no chylous leak. There was no additional RLN injury, and the mean pre- and post-completion thyroidectomy serum calcium levels were 8.96±0.39 mg/dL and 8.74±0.56 mg/dL, respectively. No patient developed wound infection or hematoma requiring exploration. Thirteen patients had temporary hypocalcemia, whereas eight patients had permanent and symptomatic hypoparathyroidism. Five patients in the postoperative course of six months regained a normal serum calcium level without need of supplementation. Eight patients had persistent hypocalcemia and remained symptomatic, requiring oral calcium supplementation. Five of these patients had undergone bilateral paratracheal nodal dissection. DISCUSSION A completion thyroidectomy is required when papillary or follicular thyroid carcinoma is diagnosed postoperatively in patients undergoing partial thyroidectomy. It facilitates removal of the residual thyroid tissue/disease, permits screening and ablation of metastatic disease with radioactive iodine, and allows for thyroglobulin level monitoring, thereby eliminating the risk of recurrence of the tumor and prolonging survival. In addition, it is reported that presence of multifocal tumor is associated with a high risk of lymph node metastasis.11 Studies have reported residual tissue malignancy rates of 33% to 44% after the completion surgery.5–10 It was reported to be as high as 78% in Levin’s series.12 Several authors have concluded that completion thyroidectomy is safe; associated morbidity and timing had no impact on development of complications.5–13 The rate of the recurrent laryngeal nerve palsy varies from 0% to 4%.5–13 Although some studies recommend that completion thyroidectomy should be performed either before scar tissue development or after clinical remission of scar tissue, edema, and inflammation, recent evidence shows that timing of surgery has no effect on morbidity.10 Recurrent laryngeal nerve (RLN) injury can occur after thyroid surgery. The rate of RLN injury, mostly transient, ranges from 0.5% to 5%.14 The risk is more important in patients who undergo reoperative thyroid surgery and in patients with thyroid cancer or hyperthyroidism.12 A meticulous surgical technique can lower the postoperative morbidity if precise operative rules are respected. Though the literature describes very low rates of recurrent laryngeal nerve injury during the primary thyroid surgery, in our series it was higher, probably because the majority of the cases were performed in nursing home setups by surgeons not trained in head neck oncology. In our study there was no additional RLN injury in completion thyroidectomy. We used surgical loupes for meticulous dissection and identification of the nerves and parathyroid. A nerve monitor was not used in this study. Completion thyroidectomy was shown to be a fairly safe procedure, which carries a low incidence of complications and also facilitates further management and follow-up with radioactive iodine. There was no recurrent laryngeal nerve palsy in the study by El-Zohairy and Zaher.15 Similar results were noted with a 0% incidence of RLN injury when completion thyroidectomy was necessary for the treatment of thyroid malignancy, and it was effective for diagnosing and removing occult disease in the remaining thyroid by Kupferman et al.5 In another study of 647 patients, conducted by Gulcelik et al. who compared groups for complications, there were no significant differences except temporary hypocalcemia between completion thyroidectomy and total thyroidectomy for differentiated thyroid carcinoma.16 The routine use of intraoperative neuromonitoring seems not to reduce the incidence of RLN during revision thyroid surgery, at least in the setting of a tertiary referral center.17 Hypocalcemia after thyroidectomy is the most common postoperative complication, with a reported incidence from 0.5% to even 50% of the operated patients.18 Transient hypoparathyroidism is said to be present when oral/intravenous calcium supplements are required for less than six months after surgery, and permanent hypoparathyroidism if hypocalcemia symptoms last more than six months or when the requirement for oral calcium/intravenous supplements lasts longer than six months. In a large series of 5,000 patients undergoing bilateral thyroid surgeries the overall incidence of transient and permanent hypoparathyroidism was 7.3% and 1.5%, respectively. Extent of resection and surgical technique had a greater impact on the rates of permanent postoperative hypoparathyroidism than thyroid pathologic condition.19 Other series have reported an incidence of temporary hypoparathyroidism of 0% to 14%, and a permanent hypoparathyroidism incidence around 2% to 8%.5–10,13,20–22 With improvements in surgical technique and experience, complication rates of thyroidectomy performed for benign or malignant diseases are reduced. In spite of the improvement in surgical experience, temporary RLN palsy and hypoparathyroidism are the main complications in completion thyroidectomies, which need special attention.16 Hypocalcemia occurred more frequently when neck dissection was combined with total thyroidectomy (60%) than without it (17%) (P<0.005). The incidence of hypocalcemia was higher after central, than lateral, neck dissection.23 In a series by Roh et al. the incidence of temporary hypocalcemia was 46.3%, and permanent hypocalcemia 4.9%, in patients undergoing central compartment reoperation for recurrent/persistent differentiated thyroid cancer.24 Other series did not reveal increased morbidity following central compartment dissection.25,26 In our study, transient hypoparathyroidism occurred in 13 (24.5%) patients. Five patients were lost to follow-up. Permanent and symptomatic hypoparathyroidism occurred in eight patients (8/48 patients, 16.67%). We considered all patients for radioiodine ablation as the tumor size in previous histopathology reports could not be verified, and in a significant number of patients we found residual thyroid tissue in the previously operated thyroid bed. CONCLUSION Completion thyroidectomy is a safe and appropriate option in the management of well-differentiated thyroid cancer. It removes disease on ipsilateral and contralateral sides of the thyroid and carries a low risk of recurrent laryngeal nerve damage, but a higher risk of permanent hypoparathyroidism. The incidence of hypocalcemia was higher after central than lateral neck dissection. With improvements in surgical technique and knowledge, complication rates of completion or revision thyroidectomy are reduced. Conflict of interest: No potential conflict of interest relevant to this article was reported. Abbreviations FNACfine-needle aspiration cytology RLNrecurrent laryngeal nerve ==== Refs REFERENCES 1 Al-Sayer H Krukowski Z Williams V Fine needle aspiration cytology in isolated thyroid swellings: a prospective two year evaluation Br Med J 1985 290 1490 2 10.1136/bmj.290.6480.1490 3922546 2 Bisi H Camargo R Filho A Role of fine needle aspiration cytology in the management of thyroid nodules: review of experience with 1925 cases Diagn Cytopathol 1991 8 504 10 10.1002/dc.2840080510 3 Caplan R Kisken W Strutt P Fine needle aspiration biopsy of thyroid nodules: a cost effective diagnostic plan Wis Med J 1991 90 183 90 10.1080/00325481.1991.11700990 4 Miller I Hamburger J Kim S Diagnosis of thyroid nodules: use of fine needle aspiration and needle biopsy JAMA 1979 242 481 10.1001/jama.1979.03290310021006 5 Kupferman ME Mandel SJ DiDonato L Wolf P Weber RS Safety of completion thyroidectomy following unilateral lobectomy for well-differentiated thyroid cancer Laryngoscope 2002 112 1209 12 10.1097/00005537-200207000-00013 12169901 6 Kim ES Kim TY Koh JM Completion thyroidectomy in patients with thyroid cancer who initially underwent unilateral operation Clin Endocrinol (Oxf) 2004 61 145 8 10.1111/j.1365-2265.2004.02065.x 15212657 7 DeGroot LJ Kaplan EL Second operations for “completion” of thyroidectomy in treatment of differentiated thyroid cancer Surgery 1991 110 936 40 1745981 8 DeJong SA Demeter JG Lawrence AM Paloyan E Necessity and safety of completion thyroidectomy for differentiated thyroid carcinoma Surgery 1992 112 734 9 1411945 9 Rosário PW Fagundes TA Borges MA Completion thyroidectomy in patients with thyroid carcinoma initially submitted to lobectomy Clin Endocrinol 2004 61 652 3 10.1111/j.1365-2265.2004.02148.x 10 Auguste LJ Attie JN Completion thyroidectomy for initially misdiagnosed thyroid cancer Otolaryngol Clin North Am 1990 23 429 39 2195432 11 Chow SM Law SC Chan JK Au SK Yau S Lau WH Papillary microcarcinoma of the thyroid prognostic significance of lymph node metastasis and multifocality Cancer 2003 98 31 40 10.1002/cncr.11442 12833452 12 Levin KE Clark AH Duh QY Demeure M Siperstein AE Clark OH Reoperative thyroid surgery Surgery 1992 111 604 9 1595056 13 Makay O Unalp O Icoz G Akyildiz M Yetkin E Completion thyroidectomy for thyroid cancer Acta Chir Belg 2006 106 528 31 10.1080/00015458.2006.11679945 17168263 14 Tresallet C Chigot JP Menegaux F How to prevent recurrent nerve palsy during thyroid surgery? Ann Chir 2006 131 149 53 16216215 15 El-Zohairy M Zaher AJ Re-operation for the treatment of well differentiated thyroid cancer: necessity, safety and impaction on further management J Egypt Natl Canc Inst 2004 16 130 6 15959545 16 Gulcelik MA Kuru B Dincer H Complications of completion versus total thyroidectomy Asian Pac J Cancer Prev 2012 13 5225 8 10.7314/APJCP.2012.13.10.5225 23244139 17 Alesina PF Rolfs T Hommeltenberg S Hinrichs J Meier B Mohmand W Intraoperative neuromonitoring does not reduce the incidence of recurrent laryngeal nerve palsy in thyroid reoperations: results of a retrospective comparative analysis World J Surg 2012 36 1348 53 10.1007/s00268-012-1548-6 22411090 18 Proczko-Markuszewska M Kobiela J Stefaniak T Lachinski AJ Sledzinski Z Postoperative PTH measurement as a predictor of hypocalcaemia after thyroidectomy Acta Chir Belg 2010 110 40 4 10.1080/00015458.2010.11680563 20306908 19 Thomusch O Machens A Sekulla C Ukkat J Brauckhoff M Dralle H The impact of surgical technique on postoperative hypoparathyroidism in bilateral thyroid surgery: a multivariate analysis of 5846 consecutive patients Surgery 2003 133 180 5 10.1067/msy.2003.61 12605179 20 Kisaoglu A Ozogul B Akçay MN Completion thyroidectomy in differentiated thyroid cancer. When to perform? Ulus Cerrhai Derg 2014 30 18 21 10.5152/ucd.2014.2486 21 Lefevre JH Tresallet C Leenhardt L Jublanc C Chigot JP Menegaux F Reoperative surgery for thyroid disease Langenbecks Arch Surg 2007 392 685 91 10.1007/s00423-007-0201-6 17593385 22 Rafferty MA Goldstein DP Rotstein L Completion thyroidectomy versus total thyroidectomy: is there a difference in complication rates? An analysis of 350 patients J Am Coll Surg 2007 205 602 7 10.1016/j.jamcollsurg.2007.05.030 17903736 23 Cheah WK Arici C Ituarte PH Siperstein AE Duh QY Clark OH Complications of neck dissection for thyroid cancer World J Surg 2002 26 1013 16 10.1007/s00268-002-6670-4 12045861 24 Roh JL Kim JM Park CI Central compartment reoperation for recurrent/persistent differentiated thyroid cancer: patterns of recurrence, morbidity, and prediction of postoperative hypocalcemia Ann Surg Oncol 2011 18 1312 18 10.1245/s10434-010-1470-9 21140230 25 Palestini N Borasi A Cestino L Freddi M Odasso C Robecchi A Is central neck dissection a safe procedure in the treatment of papillary thyroid cancer? Our experience Langenbecks Arch Surg 2008 393 693 8 10.1007/s00423-008-0360-0 18592264 26 Kim MK Mandel SH Baloch Z Morbidity following central compartment reoperation for recurrent or persistent thyroid cancer Arch Otolaryngol Head Neck Surg 2004 130 1214 16 10.1001/archotol.130.10.1214 15492172

PMC005xxxxxx/PMC5001795.txt

==== Front Rambam Maimonides Med JRambam Maimonides Med JRMMJRambam Maimonides Medical Journal2076-9172Rambam Health Care Campus 10.5041/RMMJ.10250rmmj-7-3-e0023Medicine and Jewish LawShould Sabbath Prohibitions Be Overridden to Provide Emotional Support to a Sick Relative? Greenberger Chaya Ph.D.12*Mor Pnina Ph.D.34 1 Dean, Faculty of Life and Health Sciences, Jerusalem College of Technology, Jerusalem, Israel 2 Chair, Department of Nursing, Jerusalem College of Technology, Jerusalem, Israel 3 Department of Nursing, Jerusalem College of Technology, Jerusalem, Israel 4 Shaare Zedek Medical Center, Jerusalem, Israel* To whom correspondence should be addressed. E-mail:greenber@jct.ac.il7 2016 28 7 2016 7 3 e0023Copyright: © 2016 Greenberger and Mor.2016This is an open-access article. All its content, except where otherwise noted, is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.Background There is a consensus among the halachic authorities that life-saving actions override Sabbath prohibitions. They are painstaking in securing that the sanctity of the Sabbath is maintained but that not a single life be lost. Objective This manuscript examines if and when a relative’s presence at the bedside of a seriously ill individual is potentially life-saving against the backdrop of the scientific literature. It specifically addresses the permissibility of traveling in a motorized vehicle, generally prohibited on the Sabbath, to be with one’s relative in hospital for the provision of emotional support. Methods Discourse of the halachic issues in the context of the scientific literature. Results Stress, mental or physical, has been determined as a potentially life-threatening condition in many disease entities. The literature attests to both the patient’s and the professionals’ perception of the curative potential of the presence of loved ones by advocating for the patient and relieving stress in the hospital experience. Emotional support from a loved one is perceived by some patients as vital to survival. There is halachic consensus that a patient’s perception of the emotional need for a relative’s presence is sufficient to permit overriding rabbinic prohibitions. Torah prohibitions, which may be overridden for medical needs, may be overridden for emotional support, providing a health professional or family member attests to the fulfilment of this specific need as diminishing the danger to the patient’s life. In certain cases, the latter contingency is unnecessary. Conclusions Emotional support has an impact on the patient’s health status; the degree to which its impact is strong enough to save life is still being studied. As more data from scientific studies emerge, they may be relevant to sharpening the halachic rulings with respect to the issue at hand. Emotional supporthalachalife-savingoverriding Sabbath laws ==== Body INTRODUCTION There is consensus among halachic authorities that life-saving actions override Sabbath prohibitions. This manuscript examines if and when a relative’s presence at the bedside of a person in a life-threatening condition (choleh shyesh bo sakana) is considered potentially life-saving. Specifically, it addresses the permissibility of traveling in a motorized vehicle, generally prohibited on the Sabbath, to be with one’s seriously ill relative in hospital for the provision of emotional support. The relevant scientific literature prefaces the halachic discourse in order to illustrate its reflection of the halachic rulings. It is remarkable that, traditionally, hospitals restricted or even barred visiting severely ill patients. Reasons for these prohibitions included the fear that visitors presented both a threat to patients (via the risk of infection and the increased stress of “hosting”) and that medical staff considered them a hindrance to patient care.1–3 It is now known that the presence of relatives by the bedside of a patient contributes to maintaining and even improving the physical and mental health of the patient due to the support, particularly the emotional support, that they provide.4–12 In this regard, clinical practice guidelines of the American Association of Critical Care recommend open visitation for family members,13 and the American Association of Critical Care Nurses delineates around-the-clock support by kin as expected practice in intensive care units13,14 as it diminishes anxiety, enhances safety and security, and minimizes complications. THE ILLNESS EXPERIENCE AND THE DETRIMENTAL EFFECTS OF PSYCHOLOGICAL STRESS Acute illness is an assault on physiological homeostasis, but also an existential threat, as an individual finds himself/herself in strange surroundings, often helpless, in pain, and tense with respect to the unknown future. Anxiety and depression are manifestations of this experience.15 The combination of these factors creates stress, a condition in which strain can exceed the ability of the individual to adapt, causing distress.16,17 Hans Selye identified a three-stage common physiological response to stress—physical or psychological—coining the term “general adaptation syndrome” (GAS): alarm, resistance, and exhaustion.18 Lipp (in Lucinda et al.) identified an additional stage of semi-exhaustion.19 In the alarm stage, stress launches the secretion of adrenaline, accompanied by psychological arousal. This typically triggers a therapeutic increase in blood pressure and pulse rate and stimulates immune activity. However, prolonged stress or short-lived stress of a large magnitude is detrimental. As the individual enters the stage of resistance, cortisol is secreted in an attempt to re-establish homeostasis, and anxiety is pervasive. This may increase vulnerability to infection, prevent or delay surgical healing, and continue to tax the heart. The quasi-exhaustive stage is characterized by the beginnings of general organ deterioration, with the exhaustive stage resulting in depression and organ failure. Research conducted by Lucinda et al. (n=42) found that 72% of patients were experiencing stress four days post-acute myocardial infarction, 71% in the resistance stage.19 A large body of research has confirmed the potentially negative impact of stress. As it is beyond the scope of this article to address all the research linking psychological stress to negative outcomes in illness, studies most relevant to the subject at hand have been chosen. Moser et al.,20 for example, found the degree of anxiety in the first few hours following an acute myocardial infarction to be a significant predictor of complications such as fatal arrhythmia and excessive life-threatening clotting, after controlling for other variables (e.g. the size of the infarction, side effects, and previous infarctions). This is remarkable as it lends evidence to the detrimental effects of stress on the cardiac muscle in the initial “constructive” stage of “alarm reaction.” The heart has unique vulnerability as it labors hard from the start to nourish the stress response. In this regard, Krantz et al.21 reported extreme anger as an immediate trigger of myocardial infarction and found general psychological stress a trigger for acute cardiac events in illness situations, including arrhythmias and sudden death. Similarly, Huffman et al.22 reported a correlation between severe anxiety and depression and sudden onset of excessive clotting in cardiac patients. An acute life-threatening stress response prevalent in hospitalized patients is delirium. The condition develops through the combined presence of physiological factors such as infection and fluid/electrolyte imbalance, coupled with environmental ones such as stimulation overload and isolation from loved ones.23 Patients display delusionary confusion and disorientation, becoming either hyper- or hypo-actively unco-operative, dysfunctional, and potentially harmful to themselves and their surroundings. Ryan et al.24 reported the point prevalence of delirium in hospitalized patients at 20%, with some studies reporting percentages of delirium in intensive care units to be as high as 70%. It is the leading complication of hospitalization for older adults.25–27 Delirium is associated with serious negative consequences, including increased morbidity and mortality, with an 11% increase in mortality for every additional 48 delirious hours.24 Stressful psychological concomitants of acute illness can remain a threat to life beyond the acute phase of illness, resulting in varied morbidities generally and post-traumatic stress specifically.28 The Diagnostic and Statistical Manual of Mental Disorders (DSM IV) now categorizes acute illness as a potential antecedent of post-traumatic stress disorder (PTSD)29,30 in light of the accumulated evidence regarding the possible delayed life-threatening sequels of illness on both physical and psychological health. In a prospective study, 6 (14%) out of 43 patients mechanically ventilated in intensive care units developed severe PTSD at 6 months’ follow-up.31 In a meta-analysis conducted by Edmonson et al.,32 PTSD was found to be both prevalent after acute cardiac illness and associated with future cardiac events and higher subsequent mortality, with a 55% increase in risk, after controlling for other risk factors. Spindler and Pedersen30 reported perceived severity of the illness, rather than the objective illness severity, to be a better predictor of PTSD; Guler et al. found feelings of helplessness to be especially predictive. 33 Symptoms of post-traumatic stress (PTS) can also manifest themselves while an individual is still in the early recuperative stages of illness. Talisayon et al.34 reported that PTS symptoms were present in more than one-third of the critically ill within 1 week post-hospitalization. In intensive care unit (ICU)-ventilated patients, rates of PTS symptoms were reported at 27%, 24%, and 12% in different reports,34 whereas for post-operative cardiac surgery patients the percentage was reported to be 14.7%.35 SOCIAL SUPPORT MODERATES STRESS AND ITS OUTCOMES Lazarus and Folkman36 identified social support as one of the critical resources available for enduring stress, including that of illness and post-trauma. Research over the last few decades points to social support as a significant factor in decreasing morbidity and mortality. The mechanisms include a direct decrease in physiological reactivity of the cardiovascular and neuroendocrine systems37,38 and an indirect positive impact on coping.39 Social support is often divided into emotional, informational, and practical support, the latter two facilitating decision-making and effective health behaviors.17 Evidence, however, points to perceived emotional support—best provided by loved ones—as most influential in stress reduction by reassuring a person that he or she is a valuable individual about whom people care. Meaning, purpose, and a sense of worth and belonging which goes to the core of human existence are nourished by emotional support.17,40 In this vein, Herlitz et al.41 reported that, among 1,290 patients who underwent coronary artery bypass surgery, ratings of the statement “I feel lonely” predicted survival at 30 days, even after controlling for preoperative factors known to increase mortality. FAMILY PRESENCE AT THE BEDSIDE The critical care environment, intensively technological, diminishes personhood and arouses feelings of alienation.42,43 Families provide identity, security, and comfort, while significantly reducing anxiety in this intimidating environment.44–52 Patients reported that physical and verbal contact with family members during invasive and resuscitative procedures was a “healing force” that enabled them to cope more effectively with stressful experiences.47,53,54 In a study of patients undergoing liver transplants, family presence was cited as the most important source of support,55 with 35% requesting their presence during the actual transplant procedure. Fredriksen and Ringsberg’s review,56 moreover, points to separation from loved ones as itself a cause of stress. With regard to specific stress responses to illness, family integration into the continuous care of patients has been found to be protective against the development and exacerbation of delirium.57,58 Martinez et al.59 found that family intervention reduced incidence of delirium by 58%, and it is now part and parcel of the National Institute for Health and Care Excellence (NICE) treatment guidelines for delirium. Similarly, a significant body of research has pointed to the presence of significant others as a possible buffer against the development of PTSD subsequent to traumatic stress.61–63 Many qualitative studies reflect the critical importance of family presence. In research by Mylen et al.64 of former neurosurgical intensive care unit patients, respondents reported how family members infused them with a feeling of security, sense of person, purpose, and motivation. “It was a lot of energy like … you know healing, in the words of one of the patients … to have family around” (p. 45). It was a “reminder” of belonging that gave patients motivation to recover. In a similar vein, Alpers et al.65 found the family to have great impact on bolstering the patient’s inner strength “to go on living” (p. 155), an expression also used by patients in Nygren’s study.66 Another patient put it this way: “Just lying there … not moving … I wouldn’t know how I would have been today, if she [her mother] hadn’t been there.”67 In the study undertaken by Bergbom and Askwall,44 although acknowledging the importance of instrumental assistance given by relatives, patients singled out the moral support they received as “restoring to life.” Similarly, Wang et al.68 interviewed patients shortly before they were released from the ICU; one of the most momentous statements was: “My family gave me courage to persist; I might have given up without their backup” (p. 187). The latter two statements clearly indicate that in the patient’s mind a relative’s emotional support is life-sustaining. It is remarkable that mere presence has been singled out in various studies as the most critical component in family support.45 One patient in the study by Twibell et al.69 said the following: “It’s important for my family to just be there: They don’t have to do anything. We can just look over at each other … I knew when I saw them that I mattered to them. They hadn’t forgotten or given up on me. I want them to be here with me” (p. 111). A patient interviewed by Wahlin et al.70 similarly commented: “It feels safe when you’re lying there, to have someone from the family with you … I don’t have the energy to talk, but he understands that. He just sat there and held my hand” (p. 374). In terms of quantitative research, Rotondi et al.71 interviewed 150 patients following their hospitalization in an intensive care unit, where they were connected to a respirator for more than 24 hours. Of the 41 patients who recalled missing their relatives, 31 reported that this affected them significantly. Out of 38 patients who remembered a feeling of isolation, 28 reported that this noticeably distressed them. Novaes et al.72 interviewed 50 intensive care unit patients in order to determine their sources of stress. The data were collected via the Intensive Care Unit Environment Stressor Scale, a 40-item Likert scale evaluating physical and mental stress. The patients recorded severance from the family as a source of stress. Cornock73 interviewed 71 intensive care unit patients who had been connected to a respirator, as well as 71 nurses from the unit. The two groups were asked to report on three characteristics of the 50 characteristics in the Environment Stressor Scale that constituted the most significant sources of patients’ stress. Eight of the patients included missing their spouse among one of the first three choices, and seven of the patients included time limitation on visits as one of the top three. Nurses similarly graded these two characteristics at the same level as patients, lending professional validity to patients’ perceptions. In Williams’s study,48 a total of 67 nurses reported their observations of patients and their respective families. One nurse related her attempt to wean a patient off a respirator in the presence of a relative. She noted that the patient relaxed more quickly in his presence. Another described her success in weaning down pressure delivered by CPAP (continuous positive airway pressure to support breathing) in the presence of a relative as the patient’s breathing became more effective. In a randomized trial conducted regarding restricted visitation policy (RVP), Fumagalli et al.74 studied the influence of relatives’ visits on the medical state of patients. The researchers compared two similar socio-demographic groups of patients (n=111 and 115) with comparable clinical characteristics, who were hospitalized in the intensive care unit for an extended period of time. The number and length of visits was restricted for one group, while the second group benefited from an unrestricted visitation policy (UVP). Compared with the unrestricted group, the patients with restricted visitations had a 2-fold greater risk of major cardiovascular complications, particularly of pulmonary edema or shock, but also, although not significantly, of arrhythmias and cardiac rupture. The unrestricted group was associated with a greater reduction in anxiety score and a significantly lower increase in thyroid-stimulating hormone from admission to discharge. Furthermore, the mortality rate among those whose visits were not restricted was 1.8% compared to 5.2% in the groups whose visits were restricted. In another analysis of 156 patients (RVP, n=80; UVP, n=76) being treated for myocardial infarction these researchers compared the Killip class distribution (a stratifying of risk criteria) between admission to and discharge from the ICU. The Killip level improved by 58.8% among those whose visits were unrestricted, while only 3.4% of them deteriorated. In those patients whose visits were restricted, only 26.7% improved and 6.7% deteriorated. The clinical differences between the groups were attributed to the reduction in stress and anxiety arising from the unrestricted visits. Researchers using intracranial pressure to measure changes in stress reported a decrease in pressure following relatives’ visits.74,75 Others, using pulse rate and blood pressure, similarly observed a decrease in both indicators following visitation.10,48 RESEARCH SUMMARY Ample evidence points to the importance of family support and presence in alleviating stress in illness and preventing or diminishing its negative sequels. Relationships between emotional support and stress/illness factors have been assessed both qualitatively, by interviewing patients and professional caregivers, and quantitatively, via objective measures. In the former, interviewees in different studies have repeatedly used words with roots “life” and “heal” in describing what family presence means. In the latter, many objective measures—vital signs, cardiac and neurological indicators, psychological and physiological morbidity, as well as mortality—vary positively to different degrees with emotional support. THE HALACHIC DISCOURSE: TENDING TO THE MEDICAL AND EMOTIONAL NEEDS OF THE SERIOUSLY ILL ON THE SABBATH Against the backdrop of this research, we address the following case: Rabbi Zalman Nehemiah Goldberg76 was asked (in 1986) to render a halachic decision with respect to an individual hospitalized after open-heart surgery who suddenly felt unwell and requested that his son travel on the Sabbath in order to be at his bedside. As this involved overriding a Torah prohibition, Rabbi Goldberg permitted fulfilling the request only if the son were certain that his visit would have life-saving implications. It is implicit in this ruling that the presence of a loved one for emotional support may, in certain cases, be potentially life-saving; however, the patient himself is not relied upon to be the judge. Rabbi Goldberg’s stipulation is surprising in the light of other cases in which the patient’s subjective appraisal of his condition as being potentially life-saving suffices to override Torah prohibitions, since “the mind knows the suffering of the soul” (Proverbs 14:1). Accordingly, Rabbi David ben Solomon ibn Zimra (1479–1573), also called Radbaz, ruled that we comply with a patient who claims that he or she needs certain medications on the Sabbath even if the doctor considers that there is no need, as long as the doctor confirms the medication will do no harm. This responsum of Radbaz is cited by the Tzitz Eliezer77 in connection with the halachic question addressed in this manuscript. The latter distinguishes between a patient’s request for medical treatment on the Sabbath, in which case one may override Torah prohibitions even without the doctor’s consent, and the patient’s request for a relative to come to stay by his or her beside, for which purpose these prohibitions may not be overridden. The first impacts directly on the healing process, whereas the second only improves the patient’s emotional state. However, continues the Tzitz Eliezer, if a doctor, an authoritative professional, were to stipulate that the absence of the relative could potentially endanger the patient (as Rabbi Goldberg ruled with respect to the relative), the prohibitions need be overridden as with regard to any action related to healing. Rabbi Epstein, the author of Aruch Hashulchan,78 and Rabbi Hadaya, author of Yaskil Avdi,79 likewise rule that if a doctor affirms that not granting a sick individual’s request to send for his or her relative would put his or her life in danger, this must be regarded as equivalent to medical treatment, and the Torah prohibitions of the Sabbath must similarly be overridden. With respect to the impact of perceptions on healing, we turn to Maimonides’ Hilchot Avodah Zarah80 in which he permits so-called “whispering” (a technical term for a type of sorcery alleged to cure). While this is, in his opinion, Torah-forbidden as a superstitious practice (and hence akin to idolatry), at the insistent request of a dangerously ill individual, it is permitted even on the Sabbath, in order to prevent extreme mental anguish. This ruling is made despite the fact that Maimonides himself is convinced there is no cure in this. Halachic standing is given to the patient’s belief in the healing power of certain actions, even when the belief is mistaken. The author of Nefesh Hayyah81 cites Maimonides’ attribution of halachic status to subjective perceptions as support for his position regarding a case similar to ours. In the case of a dangerously ill individual who expresses a longing to see his relative, Nefesh Hayyah was asked whether a relative may override Torah prohibitions in order to be at the patient’s bedside. The author gives standing to the request but iterates that overriding these prohibitions would, in addition, necessitate some objective evidence regarding the curative potential of the relative’s presence. The assumption may be that the heightened emotional state of the sick individual might bring him to request his relative’s presence, even if he himself does not truly perceive the latter’s absence to be life-threatening. Regarding treatment, however, his perceptions, as we have seen, are assumed to be genuine. A more straightforward possibility is that not fulfilling the individual’s request regarding treatment is deemed by Nefesh Hayyah to be more detrimental than a respective decision regarding the request for a relative’s presence. Rabbi Shmuel Wosner82(Part 8:65) concurs with this ruling, differentiating between a seriously ill individual and a woman in the post-partum state. With respect to the latter, because the birth experience is uniquely laden with emotion, no additional attestation is needed to confirm that emotional well-being has life-saving implications. Therefore, Rabbi Jacob ben Asher, the author of Orach Chayim,83 stipulates that a light may be lit for her even if she is blind, so that she should not be afraid. With regard to the sick individual, the rabbi ruled that a doctor must attest to the life-threatening potential of the emotional stress (and thus the vital need for emotional support). Torah Versus Rabbinical Prohibitions on the Sabbath According to Rabbi Elijah ben Shlomo Zalman, known as the Vilna Gaon (Genius of Vilnius),84 one may send a non-Jew on the Sabbath to arrange for the relative of a seriously ill individual to travel immediately after the Sabbath, to be at his bedside. Here again, only a minor rabbinic prohibition is being overridden, making it permissible in order to relieve mental anguish. Mishna Berurah85 extends the removal of rabbinic prohibitions beyond that of just asking a non-Jew to be an informant to hiring a non-Jewish runner. Shulchan Aruch Shel HaRav,86 however, rules that as the presence of a relative does not affect any real medical recovery, but merely eases emotional suffering, rabbinically forbidden actions may only be undertaken by a non-Jew. A somewhat different case is presented in the responsa of the Shoel U’Meshiv,87 cited also by Rabbi Yisrael Matisyahu Auerbach,88 in which a man hears that his sick wife has become stricken with acute anxiety and is in a village where nobody knows her. The Shoel U’Meshiv rules that the husband may ride there by horse on the Sabbath (a major rabbinic prohibition, as it is being performed by a Jew). He reasons that the wife will certainly benefit from her husband’s arrival, and this is a case of “possible life-saving” (safek pikuach nefesh) which overrides the laws of the Sabbath. The Shoel U’Meshiv does not explain his position. The author of Helkat Yaakov89 perceives the life-saving elements of the husband’s presence as nested in the overall benefit that the woman receives; apart from easing her mind he will provide practical assistance (i.e. safety measures, hygiene) which justifies overriding major rabbinic prohibitions (and, as we have seen earlier, Torah prohibitions). If this is an accurate interpretation of the Shoel U’Meshiv, it cannot be deduced from this that mere emotional support alone would be sufficient to permit a Jew to ride a horse on the Sabbath. Furthermore, a patient is not alone in a hospital as staff tends to both practical and emotional needs. The responsum of Migdal Hashen,90 cited by Rabbi Waldenberg,77 relates specifically to the distinction between emotional support and attendance to the patient’s practical needs. He discusses the case of a sick individual who sends a letter to another town urgently requesting a doctor as he is in danger. He rules that a Jew may travel (on a wagon, a minor rabbinic prohibition) on the Sabbath with the doctor in order to ensure that the doctor arrives as soon as possible. He raises the possibility that the Jew may even be permitted to travel alone (a major rabbinic prohibition) as it is permissible to light a lamp for a woman after childbirth, even if she is blind, to settle her mind in case she is afraid, and that her fear may endanger life. Unlike the Shevet Halevi,82(Part 8:65) who attributes permission to light a lamp for a blind woman post-partum to ease her mind overall, Migdal Hashen attributes it specifically to allaying her fear regarding the impact of the darkness on the quality of the treatment and therefore equates her with the seriously ill individual. For concerns of treatment, even a Torah prohibition is overridden for both these cases. Migdal Hashen equates this case to permitting a relative to travel with the doctor to ease the patient’s fear that the doctor may not look after him properly. It must be stressed that the action required is directly connected with medical needs. On the other hand, the presence of a relative to ease the emotional distress of being alone is not designated a priori by the author of Migdal Hashen as potentially life-saving. It is remarkable that there is a consensus among halachic authorities regarding the obligation to override Torah prohibitions in order to provide information to the health care provider and advocate for the patient. Shevet Halevi (8:68) stipulates that it is always mandatory for a family member to accompany an unconscious patient, as he is certainly incapable of human interaction. Mental Anguish May Be Life-threatening Rabbi Wosner82(Part 50:71) cites examples of situations that are life-threatening in and of themselves, specifically because they cause extreme mental anguish. The Babylonian Talmud states: “If a child is locked behind a door on the Sabbath the door may be broken to bring him out.”91 Rabbi Wosner similarly rules that Torah prohibitions may be overridden to free a trapped, panic-stricken adult. Rabbi Neuwirth92(Part 32:15;Part 41:27) likewise perceives relieving intense fear as a sufficient reason for overriding Torah prohibitions for a seriously ill individual who is afraid of the dark.92(Part 32:63) In contradistinction to these cases, halacha could argue that a distraught patient who calls for his relative is not always in an acute state of panic and, what is more, the relative’s presence does not neutralize the fear as the illness is ever-present. A number of additional halachic authorities, however, consider mental anguish as potentially life-threatening with respect to the Torah prohibitions. The author of Pri Megadim93 categorizes extreme mental anguish as life-threatening, for which even Torah prohibitions may be overridden. Minchat Yitzhak94 cites further authorities, namely the Levush,95 Tosefot Shabbat,96 and Levushi Srad,97 who regard mental anguish as a potentially life-threatening situation and also contend that it may be eased by the presence of a significant other. Rabbi Ovadia Yosef98 ruled that a seriously wounded soldier who requests a relative’s presence for the sole purpose of easing his mind is similar to a woman after childbirth; Torah prohibitions may be overridden in order to fulfill his request, and there is no need to obtain anyone else’s opinion regarding the matter’s urgency. In addition to establishing mental anguish as life-threatening in certain cases, Rabbi Yosef also establishes that the presence of a loved one is potentially life-saving through easing the anguish. Perhaps a victim of terror would also fall into this category. It is possible that Rabbi Neuwirth may have modified his stand regarding the permissibility of a relative accompanying an individual to hospital on the Sabbath. In the second edition of Shemirath Shabbath,99 he permitted riding in the vehicle which is transporting a loved one to hospital, a minor rabbinic violation (if at all). In the second edition of Nishmat Avraham,100 Rabbi Neuwirth is quoted as permitting driving even in a separate vehicle, but only for the purpose of providing practical assistance in hospital, such as giving a medical history. In the third edition of Shemirath Shabbath,92(Part 40:72) Rabbi Neuwirth clearly states that riding in a separate vehicle is permissible, both in order to be present for emotional support and to provide information for his/her relative upon arrival at the hospital. This ruling implies equal life-saving potential in both roles—practical and emotional. Additional Special Cases Shemirath Shabbath92(Part 32:26) rules that one may be lenient even regarding Torah prohibitions with respect to a patient whose chance of recovery depends on his or her emotional state. The example offered is of an individual predisposed to depression who might behave dangerously with regard to himself or others if he perceived that he was not properly being cared for. No specific halachic ruling has been found by these authors regarding relieving emotional stress levels by traveling to the patient’s bedside for individuals who are suffering from acute medical conditions especially sensitive to emotional status, such as a myocardial infarction. Such stress can, as the literature points out, be immediately life-threatening, finding expression in potentially fatal arrhythmias, excessive blood clotting, spiking high blood pressure, and respiratory distress. They also harbor the seeds of potential threat to life at a future time, which, according to some halachic authorities, may warrant overriding even Torah prohibitions.101 Barring fatality, there is also the real possibility of permanent mental deterioration, seriously impinging upon the ability to live a Torah-observant life. In this case, the principle of overriding one Sabbath in order to enable the observance of many more in the future might become applicable. Although emotional support of significant others is part of the preventative protocol for potentially fatal conditions of delirium and post-traumatic stress, it must be kept in mind, however, that social support is only one interventionary measure within a complex treatment protocol, and its therapeutic weight is not readily assessable. Beit Yehudah102 cites a case of a dying individual who lay in a dark house and ruled it permissible to light a lamp on the Sabbath (a Torah prohibition) so that he might see his relatives, thereby soothing his mental anguish. Rabbi Mordecai Gutman103 perceives this ruling as being based on respect for a human being who is made in the Divine image, a supreme need for which Sabbath prohibitions may be overridden. One might argue, more simply, that in this case seeing his relatives might allay his anxiety and thus lengthen his life even if only for a short period. The professional literature points out that fear of dying “alone” can cause worse distress than the fear of death itself. The question of traveling to be beside an individual on his deathbed on the Sabbath was not herein specifically addressed. Nevertheless, it would seem to be no less important than lighting a lamp to enable the individual to see relatives known to be present. What Are Considered “Needs” of a Seriously Ill Individual? Shulchan Aruch104 and Maggid Mishneh105 are of the opinion that all the needs of a seriously ill individual may be met on the Sabbath in a fashion similar to a weekday (e.g. overriding Sabbath prohibitions) even if they are not essentially life-saving. This has relevance to our discourse since if all needs may be met, this would also include the presence of a relative at the bedside, even if it were not, per se, a life-saving action. Other authorities such as Rashi and the Geonim permit only those actions that actually mitigate danger. Perhaps there is no real disagreement. Radbaz (cited by Rabbi Waldenberg77) delineates that the type of need of a sick individual which is permitted is any need that has a life-saving aspect to it, even if only indirectly. Since the patient is already in danger, the range of needs should be expanded to include those with even a remote possibility of impacting on life-saving. Clearly, however, as Rabbi Shlomo Zalman Auerbach remarks (as cited by Rabbi Avraham S. Abraham106), the line of demarcation would not include delivering a newspaper or turning on a radio, which would certainly not be permissible. Rabbi Wosner82(Part 8:71) holds a similar position: regarding a choleh shyesh bo sakana, it can never be fully known what can have a detrimental impact on his condition. Even if refraining from the fulfillment of a need does not immediately increase danger, it might possibly weaken the individual over time and decrease his ability to overcome his illness. This seems to expand the time frame of pikuach nefesh; even future danger warrants overriding Torah prohibitions. Despite this categorization, it will be recalled that Rabbi Wosner forbids relatives from breaching Torah laws in order to be at the side of choleh shyesh bo sakana, the initial assumption still being the lack of correlation between a relative’s presence and pikuach nefesh, unless proven otherwise. Rabbi Asher Weiss107 goes a step further. While it is not possible to assess what will cause a seriously ill individual to succumb to death, anything that is related to a cure, affects healing, or provides an improved feeling of well-being is to be considered a life-saving act, similar to easing the mind of the woman in confinement. Although no specific ruling has been given regarding our case, it is possible that Rabbi Weiss would permit it. Rabbi Moshe Farbstein’s approach is similar.108 With respect to the seriously ill individual, the assessment of what is considered to be life-saving is made at a different level. It is clear from the medical literature that a patient who has a life-threatening condition does not have the mental and physical reserves that non-threatened patients have. Therefore, when considering his or her needs, even those remotely related to healing must be met. Rabbi Farbstein relates to another element that affects the definition of life-saving, namely public opinion. That which, in the opinion of the public, is considered necessary for life-saving, even if in fact the connection is far-fetched, must be considered as life-saving for halachic purposes, and Rabbi Auerbach109 comments likewise. Rabbenu Tam110 considers that a dog bite is, objectively, very far from dangerous to life, but since public opinion considers it dangerous, it must be considered as such, and the Sabbath laws may be overridden in such a case. CONCLUSION There is clear evidence in the literature regarding the detrimental effects of stress and the positive impact of a relative’s presence on the process of recovery through alleviating stress. This has spurred widespread policy changes regarding visitation. Although there is a dearth of randomized controlled trials, there are empirical studies that lend substantial evidence to stress reduction in the presence of relatives, with subsequent decreases in potentially fatal complications in unstable patients. From a subjective perspective, patients report the importance of a relative’s presence using terms relating to life-saving and survival. According to some halachic authorities, patients are not solely reliable reporters when it comes to their emotional needs. However, relatives and nurses have also attested to the importance of the relative’s presence for such instances as being weaned off ventilating devices and reducing anxiety. Halachic authorities refer to physicians as the authoritative health professional; perhaps as nurses continue to become more autonomous they will also be considered authoritative in this regard, especially as they are often the health care providers most attuned of all to the patient’s emotional state and needs. The public’s perception regarding what constitutes danger also has halachic validity, as Rabbi Farbstein has pointed out.104 It is therefore important to continue to follow the professional literature and public opinion regarding the impact of stress, the impact of family presence, and the connection between the two. Further studies regarding these phenomena may affect future halachic rulings. Halachic authorities are painstaking in their rulings in order that the sanctity of the Sabbath may be maintained, but that not a single life should be lost as a result. There is a delicate balance to maintain, and we have seen shades of opinions. With respect to traveling on the Sabbath in order to be with a hospitalized loved one for the sole purpose of giving emotional support, most authorities only permit overriding rabbinic prohibitions if a doctor attests to it being a matter of pikuach nefesh, although as we have seen, there are some important exceptions regarding the place of family support in illness as reflected in the literature. These are special cases in which emotions categorically play a dominant role in life-saving. In reality, however, when a relative is summoned to a patient’s bedside on the Sabbath, his/her arrival may be vital for both medical and emotional needs. In this regard, Rabbi Mordechai Halpern,76 after surveying a broad range of relevant halachic opinions, concludes that, when actually confronted with the situation, a loved one must travel to the scene without hesitation and without speculating which of the two needs the presence is apt to meet and to what degree. The overall situation, he iterates, is clearly one of safek pikuach nefesh for which “one who responds speedily is to be praised and one who hesitates should be rebuked.”111 Conflict of interest: No potential conflict of interest relevant to this article was reported. GLOSSARY HalachaThe corpus of Jewish religious law rooted in the Bible and continually being expanded by its designated authorities Choleh shyesh bo sakanaAn individual whose state of health endangers his life Safek pikuach nefeshA situation in which there is a potential danger to human life, which necessitates taking immediate action Abbreviations ICUintensive care unit PTSpost-traumatic stress PTSDpost-traumatic stress disorder RVPrestricted visitation policy UVPunrestricted visitation policy ==== Refs REFERENCES 1 Berti D Ferdinande P Moons P Beliefs and attitudes of intensive care nurses toward visits and open visiting policy Intensive Care Med 2007 33 1060 5 10.1007/s00134-007-0599-x 17384930 2 Engström Å Söderberg S Receiving power through confirmation: the meaning of close relatives for people who have been critically ill J Adv Nurs 2007 59 569 76 10.1111/j.1365-2648.2007.04336.x 17672848 3 Bishop SM Walker MD Spivak I Family presence in the adult burn intensive care unit during dressing changes Crit Care Nurse 2013 33 14 24 10.4037/ccn2013116 23377154 4 Agard A Lomborg K Flexible family visitation in the intensive care unit: nurses’ decision-making J Clin Nurs 2011 20 1106 14 10.1111/j.1365-2702.2010.03360.x 21029226 5 Falk J Wongsa S Dang J Comer L LoBiondo-Wood G Using an evidence-based practice process to change child visitation guidelines Clin J Onc Nurs 2012 16 21 3 10.1188/12.CJON.21-23 6 Liu V Read JL Scruth E Cheng E Visitation policies and practices in US ICUs Crit Care 2013 17 R71 10.1186/cc12677 23591058 7 Fisher C Lindhors H Matthews T Munroe DJ Paulin D Scott D Nursing staff attitudes and behaviours regarding family presence in the hospital setting J Adv Nurs 2008 64 615 24 10.1111/j.1365-2648.2008.04828.x 19120576 8 Gray H Adam J Brown D McLaughlin P Hill V Wilson L Visiting all hours: a focus group study on staff’s views of open visiting in a hospice Int J Pall Nurs 2011 17 552 60 10.12968/ijpn.2011.17.11.552 9 Karlsson C Tisell A Engstrom A Andershed B Family members’ satisfaction with critical care: a pilot study Nurs Crit Care 2011 16 11 18 10.1111/j.1478-5153.2010.00388.x 21199550 10 Jabre P Belpomme V Azoulay E Family presence during cardiopulmonary resuscitation N Engl J Med 2013 368 1008 18 10.1056/NEJMoa1203366 23484827 11 Glaser R Kiecolt-Glaser JK Stress-induced immune dysfunction: implications for health Nat Rev Immunol 2005 5 243 51 10.1038/nri1571 15738954 12 Lundberg U Stress hormones in health and illness: the roles of work and gender Psychoneuroendocrinology 2005 30 1017 21 10.1016/j.psyneuen.2005.03.014 15963652 13 Bell L Family presence: visitation in the adult ICU Am J Crit Care 2016 25 51 10.4037/ajcc2016322 14 American Association of Critical-Care Nurses Practice Alert Family visitation in the adult intensive care unit Critical Care Nurse 2016 36 e15 19 10.4037/ccn2016677 26830189 15 Tsay SL Halstead MT McCrone S Predictors of coping efficacy, negative moods and post-traumatic stress syndrome following major trauma Int J Nurs Pract 2001 7 74 83 10.1046/j.1440-172X.2001.00257.x 11811320 16 Cobb S Social support as a moderator of life stress Psychosom Med 1976 38 300 14 10.1097/00006842-197609000-00003 981490 17 Thoits PA Mechanisms linking social ties and support to physical and mental health J Health Soc Behav 2011 52 145 61 10.1177/0022146510395592 21673143 18 Selye H The Stress of Life Rev. ed New York, NY McGraw Hill Education 1976 19 Lucinda LB Prosdócimo AC de Carvalho KAT Evaluation of the prevalence of stress and its phases in acute myocardial infarction in patients active in the labor market Rev Bras Cir Cardiovasc 2015 30 16 23 25859863 20 Moser DK Riegel B McKinley S Doering LV An K Sheahan S Impact of anxiety and perceived control on in-hospital complications after acute myocardial infarction Psychosom Med 2007 69 10 16 10.1097/01.psy.0000245868.43447.d8 17244843 21 Krantz DS Sheps DS Carney RM Natelson BH Effects of mental stress in patients with coronary artery disease: evidence and clinical implications JAMA 2000 283 1800 2 10.1001/jama.283.14.1800 10770129 22 Huffman JC Celano CM Januzzi JL The relationship between depression, anxiety, and cardiovascular outcomes in patients with acute coronary syndromes Neuropsychiatr Dis Treat 2010 6 123 36 10.2147/NDT.S6880 20505844 23 Van Rompaey B Schuurmans MJ Shortridge-Baggett LM Truijen S Bossaert L Risk factors for intensive care delirium: a systematic review Intensive Crit Care Nurs 2008 24 98 107 10.1016/j.iccn.2007.08.005 17949984 24 Ryan DJ O’Regan NA Caoimh RO Delirium in an adult acute hospital population: predictors, prevalence and detection BMJ Open 2013 3 e001772 10.1136/bmjopen-2012-001772 25 Fong TG Tulebaev SR Inouye SK Delirium in elderly adults: diagnosis, prevention and treatment Nat Rev Neurol 2009 5 210 20 10.1038/nrneurol.2009.24 19347026 26 Milisen K Foreman MD Abraham IL A nurse-led interdisciplinary intervention program for delirium in elderly hip-fracture patients J Am Geriatr Soc 2001 49 523 32 10.1046/j.1532-5415.2001.49109.x 11380743 27 Young J Inouye SK Delirium in older people BMJ 2007 334 842 46 10.1136/bmj.39169.706574.AD 17446616 28 VanItallie TB Stress: a risk factor for serious illness Metabolism 2002 51 40 5 10.1053/meta.2002.33191 12040540 29 Griffiths J Fortune G Barber V Young JD The prevalence of post traumatic stress disorder in survivors of ICU treatment: a systematic review Intensive Care Med 2007 33 1506 18 10.1007/s00134-007-0730-z 17558490 30 Spindler H Pedersen S Posttraumatic stress disorder in the wake of heart disease: prevalence, risk factors, and future research directions Psychosom Med 2005 67 715 23 10.1097/01.psy.0000174995.96183.9b 16204429 31 Girard TD Shintani AK Jackson JC Risk factors for post-traumatic stress disorder symptoms following critical illness requiring mechanical ventilation: a prospective cohort study Crit Care 2007 11 R28 10.1186/cc5708 17316452 32 Edmondson D Richardson S Falzon L Davidson KW Mills MA Neria Y Posttraumatic stress disorder prevalence and risk of recurrence in acute coronary syndrome patients: a meta-analytic review PLoS One 2012 7 e38915 10.1371/journal.pone.0038915 22745687 33 Guler E Schmid J-P Wiedemar L Saner H Schnyder U Von Känel R Clinical diagnosis of posttraumatic stress disorder after myocardial infarction Clin Cardiol 2009 32 125 9 10.1002/clc.20384 19301284 34 Talisayon R Buckley T McKinley S Acute post-traumatic stress in survivors of critical illness who were mechanically ventilated: a mixed methods study Intensive Crit Care Nurs 2011 27 338 46 10.1016/j.iccn.2011.10.001 22055395 35 Arenson BG MacDonald LA Grocott HP Hiebert BM Arora RC Effect of intensive care unit environment on in-hospital delirium after cardiac surgery J Thorac Cardiovasc Surg 2013 146 172 8 10.1016/j.jtcvs.2012.12.042 23317943 36 Lazarus RS Folkman S Stress, Appraisal, and Coping New York, NY Springer 1984 37 Thorsteinsson EB James JE Douglas ME Omodei M Effects of social support on cardiovascular and cortisol reactivity during passive and active behaveioral challenge J Psychiatry Psychol Ment Health 2011 3 1 12 38 Uchino BN Social support and health: a review of physiological processes potentially underlying links to disease outcomes J Behav Med 2006 29 377 87 10.1007/s10865-006-9056-5 16758315 39 Kim J Han JY Shaw B McTavish F Gustafson D The roles of social support and coping strategies in predicting breast cancer patients’ emotional well-being testing mediation and moderation models J Health Psychol 2010 15 543 52 10.1177/1359105309355338 20460411 40 Lyyra T-M Heikkinen RL Perceived social support and mortality in older people J Gerontol B Psychol Sci Soc Sci 2006 61 S147 52 10.1093/geronb/61.3.S147 16670192 41 Herlitz J Wiklund I Caidahl K The feeling of loneliness prior to coronary artery bypass grafting might be a predictor of short-and long-term postoperative mortality Eur J Vasc Endovasc Surg 1998 16 120 5 10.1016/S1078-5884(98)80152-4 9728430 42 Berwick DM Kotagal M Restricted visiting hours in ICUs: time to change JAMA 2004 292 736 7 10.1001/jama.292.6.736 15304472 43 Rubert RL Long D Hutchinson ML Creating a Healing Environment in the ICU Kaplow R Hardin SR Critical Care Nursing: Synergy For Optimal Outcomes Burlington, MA Jones and Bartlett Learning 2007 27 39 44 Bergbom I Askwall A The nearest and dearest: a lifeline for ICU patients Intensive Crit Care Nurs 2000 16 384 95 10.1054/iccn.2000.1520 11091470 45 Hupcey JE Looking out for the patient and ourselves—the process of family integration into the ICU J Clin Nurs 1999 8 253 62 10.1046/j.1365-2702.1999.00244.x 10578747 46 Hupcey JE The meaning of social support for the critically ill patient Intensive Crit Care Nurs 2001 17 206 12 10.1054/iccn.2000.1568 11868728 47 McKinley S Nagy S Stein-Parbury J Bramwell M Hudson J Vulnerability and security in seriously ill patients in intensive care Intensive Crit Care Nurs 2002 18 27 36 10.1054/iccn.2002.1611 12008875 48 Williams C The identification of family members’ contribution to patients’ care in the intensive care unit: a naturalistic inquiry Nurs Crit Care 2005 10 6 14 10.1111/j.1362-1017.2005.00092.x 15739634 49 Gonzalez CE Carroll DL Elliott JS Fitzgerald PA Vallent HJ Visiting preferences of patients in the intensive care unit and in a complex care medical unit Am J Crit Care 2004 3 194 8 50 Abuatiq A How healthcare providers manage intensive care patients’ stressors? International Journal of Nursing (IJN) 2014 3 2 51 Celik GK Keleş A Demircan A Evaluation of patients’ families’ attitudes to witnessing invasive procedures in the emergency department J Acad Emerg Med 2013 12 63 4 52 Dougal R Anderson J Reavy K Shirazi C Family presence during resuscitation and invasive procedures in the emergency department: one size does not fit all J Emerg Nurs 2011 37 152 7 10.1016/j.jen.2010.02.016 21397129 53 Petterson M Family presence protocol can be a powerful healing force Crit Care Nurs 1999 19 104 54 Eichhorn DJ Meyers TA Guzzetta CE Family presence during invasive procedures and resuscitation: hearing the voice of the patient Am J Nurs 2001 101 48 55 10.1097/00000446-200105000-00020 55 Geary PA Formella LA Tringali R Significance of the insignificant Crit Care Nurs 1994 17 51 9 10.1097/00002727-199411000-00009 56 Fredriksen ST Ringsberg KC Living the situation stress-experiences among intensive care patients Intensive Crit Care Nurs 2007 23 124 31 10.1016/j.iccn.2006.09.002 17088063 57 Rosenbloom-Brunton DA Henneman EA Inouye SK Feasibility of family participation in a delirium prevention program for hospitalized older adults J Gerontol Nurs 2010 36 22 33 10.3928/00989134-20100330-02 58 Jirong Y Hshieh TT Inouye SK Hospital Elder Life Program (HELP) Malone ML Capezuti E Palmer RM Geriatrics Models of Care Switzerland Springer International Publishing 2015 25 37 59 Martinez FT Tobar C Beddings CI Vallejo G Fuentes P Preventing delirium in an acute hospital using a non-pharmacological intervention Age Ageing 2012 41 629 34 10.1093/ageing/afs060 22589080 60 Young J Murthy L Westby M Akunne A O’Mahony R Diagnosis, prevention, and management of delirium: summary of NICE guidance BMJ 2010 341 c3704 10.1136/bmj.c3704 20667955 61 Brewin CR Andrews B Valentine JD Meta-analysis of risk factors for posttraumatic stress disorder in trauma-exposed adults J Consult Clin Psychol 2000 68 748 10.1037/0022-006X.68.5.748 11068961 62 Guay S Billette V Marchand A Exploring the links between posttraumatic stress disorder and social support: processes and potential research avenues J Trauma Stress 2006 19 327 38 10.1002/jts.20124 16788995 63 Ozer EJ Best SR Lipsey TL Weiss DS Predictors of posttraumatic stress disorder and symptoms in adults: a meta-analysis Psychol Bull 2003 129 52 73 10.1037/0033-2909.129.1.52 12555794 64 Mylén J Nilsson M Berterö C To feel strong in an unfamiliar situation: patients’ lived experiences of neurosurgical intensive care. A qualitative study Intensive Crit Care Nurs 2016 32 42 8 10.1016/j.iccn.2015.08.001 26552575 65 Alpers LM Helseth S Bergbom I Experiences of inner strength in critically ill patients—a hermeneutical approach Intensive Crit Care Nurs 2012 28 150 8 10.1016/j.iccn.2011.10.004 22225817 66 Nygren B Norberg A Lundman B Inner strength as disclosed in narratives of the oldest old Qual Health Res 2007 17 1060 73 10.1177/1049732307306922 17928479 67 Hafsteinsdóttir TB Grypdonck M Being a stroke patient: a review of the literature J Adv Nurs 1997 26 580 8 10.1046/j.1365-2648.1997.t01-19-00999.x 9378880 68 Wang K Zhang B Li C Wang C Patients and perspectives: qualitative analysis of patients’ intensive care experience during mechanical ventilation J Clin Nurs 2008 18 183 90 10.1111/j.1365-2702.2008.02518.x 69 Twibell RS Craig S Siela D Simmonds S Thomas C Being there: inpatients’ perceptions of family presence during resuscitation and invasive cardiac procedures Am J Crit Care 2015 24 e108 15 10.4037/ajcc2015470 26523015 70 Wåhlin I Ek AC Idvall E Patient empowerment in intensive care—an interview study Intensive Crit Care Nurs 2006 22 370 7 10.1016/j.iccn.2006.05.003 16890438 71 Rotondi AJ Chelluri L Sirio C Patients’ recollections of stressful experiences while receiving prolonged mechanical ventilation in an intensive care unit Crit Care Med 2002 30 746 52 10.1097/00003246-200204000-00004 11940739 72 Novaes FP Knobel E Bork E Pavao OF Nogueira-Martins LA Ferraz M Stressors in ICU: perception of the patient, relatives and health care team Intensive Care Med 1999 25 1421 26 10.1007/s001340051091 10660851 73 Cornock MA Stress and the intensive care patient: perceptions of patients and nurses J Adv Nurs 1998 27 518 27 10.1046/j.1365-2648.1998.00555.x 9543037 74 Fumagalli S Boncinelli L Lo Nostro A Reduced cardiocirculatory complications with unrestrictive visiting policy in an intensive care unit results from a pilot, randomized trial Circulation 2006 113 946 52 10.1161/CIRCULATIONAHA.105.572537 16490836 75 Mellott KG Sharp PB Anderson LM Biobehavioral measures in a critical-care healing environment J Holist Nurs 2008 26 128 35 10.1177/0898010107306690 18539880 76 Rabbi Goldberg ZM Travelling to the bedside of a father with heart disease on the Sabbath Rulings of the Ariel Educational Institute of the Academy of Torah and Education Ariel—United Israel Institutes 13 Available at: http://bit.ly/29NTNX6 [Hebrew] 77 Rabbi Waldenberg EY (1915–2006). Responsa Tzitz Eliezer Meshivat Nefesh, Part 9 #8 15 [Hebrew] 78 Rabbi Epstein YM (1829–1908 Aruch Hashulchan) Orach Chayim New Square, NY Oz Vehadar 2006 306 20 [Hebrew] 79 Rabbi Hadaya O (1889–1969) Responsa 7 22 [Hebrew] 80 Rabbi ben Maimon M (1138–1204) Hilchot Avodah Zarah (Laws of Idolatry) 11 [Hebrew] 81 Rabbi Margaliot R (1889–1971). Nefesh Hayyah Novellae on Shulchan Aruch 278 Lvov Zohar 1932 [Hebrew] 82 Rabbi Wosner S (1913–2015) Shevet Halevi Bnei Brak Zikron Meir 2002 Part 8 65 [Hebrew] 83 Rabbi Jacob ben Asher (1269–1343) Orach Chayim 330 1 [Hebrew] 84 Rabbi Elijah ben Shlomo Zalman (1720–1797). Shulchan Aruch, Orach Chayim 328:9, on Babylonian Talmud, Tractate Baba Batra 156b; Shulchan Aruch, Orach Chayim 306:9 [Hebrew]. 85 Rabbi Cohen YM (1838–1933) Mishna Berurah Warsaw 1884 306 41 [Hebrew] 86 Rabbi Shneur Zalman of Liady (1745–1812) Shulchan Aruch HaRav Orach Chayim Kfar Habad Kehut 1989 306 20 [Hebrew] 87 Rabbi Nathansohn JS (1808–1875) Responsa Shoel U’Meshiv 3rd ed Brooklyn Klilat Yofi 1999 2 180 [Hebrew] 88 Rabbi Auerbach YM (1839–1900) Netzer Yisrael, Likutei Rima Lvov 1878 25 74 [Hebrew] 89 Rabbi Breisch MJ (1895–1976) Helkat Yaakov, Orach Chayim Tel Aviv Seder-Kol 1992 108 Available at: http://bit.ly/2a4LylJ [Hebrew] 90 Rabbi Gesenbauer SN Migdal Hashen Lvov, Austro-Hungarian Empire UW Salat Publishing 1884 [Hebrew] 91 Babylonian Talmud, Tractate Yoma 83a. 92 Rabbi Neuwirth Y Shemirath Shabbath: A Guide to the Practical Observance of Shabbath 3rd ed Jerusalem Feldheim 2010 32 15 41 27 [Hebrew] 93 Rabbi Teomim J (1727–1792). Pri Megadim: Eshel Avraham on Rabbi Gumbiner A (1637–1682) Magen Avraham on Shulchan Aruch Orach Chayim 306 18 [Hebrew] 94 Rabbi Weiss YY (1902–1983) Minchat Yitzhak Jerusalem Minchat Yitzhak 1993 4 8 [Hebrew] 95 Rabbi Jaffe M (1530–1612) On Shulchan Aruch Orach Chayim 306 3 [Hebrew] 96 Rabbi Meisels R (c.1700–c.1778) Tosefot Shabbat on Shulchan Aruch Orach Chayim 330 25 [Hebrew] 97 Rabbi Eibeschitz DS (1755–1813) Levushi Srad on Magen Avraham 306 18 [Hebrew] 98 Rabbi Yosef O (1920–2013) Responsa Yabia Omer, Orach Chayim Jerusalem Machon Maor 2004 10 29 [Hebrew] 99 Rabbi Neuwirth Y Shemirath Shabbath: A Guide to the Practical Observance of Shabbath 2nd ed Jerusalem Feldheim 1989 40 70 [Hebrew] 100 Rabbi Abraham AS (1960–) Nishmat Avraham 2nd ed Jerusalem Schlesinger Institute 2007 306 4 [Hebrew] 101 Rabbi Unterman IY (1886–1976) Shevet M’Yehuda. Vol 8. Supplement to chapters 8–10 Jerusalem Mosad Harav Kook 1983 [Hebrew] 102 Rabbi Eiesh YM (1700–1760) Responsa Beit Yehuda Leverno Stamperia & Medola 1746 Orach Chayim 59 [Hebrew] 103 Rabbi Gutman M Uniting families on the Sabbath after a terrorist attack Tehumin 2004 34 359 69 [Hebrew] 104 Rabbi Karo J (1488–1575) Shulchan Aruch: Orach Chayim 328 4 Tel Aviv Talman 1977 [Hebrew] 105 Rabbi DiTulus, Vidal of Tolosa (1284–1360) Maggid Mishneh on Maimonides, Hilchot Shabbat 2 14 [Hebrew] 106 Rabbi Abraham AS Nishmat Avraham Orach Chayim 306 9 Jerusalem Falk Schlesinger Institute 1983 [Hebrew] 107 Rabbi Weiss A Holeh shyesh bo sakana: is overriding the Sabbath permissible? Shvilai Harefuah 2006 8 59 108 Rabbi Farbstein M The borders of pikuah nefesh: overriding Sabbath prohibitions for a sick individual whose life is in danger ASSIA 2003 9 106 87 [Hebrew] 109 Rabbi Auerbach SZ Rav Neuwirth YY Shemirath Shabbath: A Guide to the Practical Observance of Shabbath 3rd ed Jerusalem Feldheim 2010 32 72 110 Rabbi ben Meir J (1100–1170, best known as Rabbeinu Tam) Tosefot on Babylonian Talmud, Tractate Avodah Zarah 28 2 111 Talmud Yerushalmi Tractate Yuma 8 5

PMC005xxxxxx/PMC5001796.txt

==== Front Rambam Maimonides Med JRambam Maimonides Med JRMMJRambam Maimonides Medical Journal2076-9172Rambam Health Care Campus 10.5041/RMMJ.10251rmmj-7-3-e0024Review ArticleAcrolein—an α,β-Unsaturated Aldehyde: A Review of Oral Cavity Exposure and Oral Pathology Effects Aizenbud Dror D.M.D. M.Sc1*Aizenbud Itay 2Reznick Abraham Z. Ph.D.3Avezov Katia D.M.D., M.Sc.1 1 Department of Orthodontics and Craniofacial Anomalies, School of Graduate Dentistry, Rambam Health Care Campus, Oral Biology Research Laboratory, Technion–Ruth and Bruce Rappaport Faculty of Medicine, Haifa, Israel 2 Hebrew University, Hadassah, School of Dental Medicine, Jerusalem, Israel 3 Department of Anatomy and Cell Biology, Ruth and Bruce Rappaport Faculty of Medicine, Technion, Haifa, Israel* To whom correspondence should be addressed. E-mail:aizenbud@ortho.co.il7 2016 28 7 2016 7 3 e0024Copyright: © 2016 Aizenbud et al.2016This is an open-access article. All its content, except where otherwise noted, is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.Acrolein is a highly reactive unsaturated aldehyde widely present in the environment, particularly as a product of tobacco smoke. Our previous studies indicated the adverse consequences of even short-term acrolein exposure and proposed a molecular mechanism of its potential harmful effect on oral cavity keratinocytic cells. In this paper we chose to review the broad spectrum of acrolein sources such as pollution, food, and smoking. Consequently, in this paper we consider a high level of oral exposure to acrolein through these sources and discuss the noxious effects it has on the oral cavity including on salivary quality and contents, oral resistance to oxidative stress, and stress mechanism activation in a variety of oral cells. Acroleincigarette smokeoxidative stressα,β-unsaturated aldehydes ==== Body INTRODUCTION Environmental Distribution of Acrolein and the Oral Cavity The oral cavity is a unique anatomic structure, characterized by the juxtaposition of soft and hard tissues which are constantly exposed to various endogenic and environmental factors influenced by personal priorities and habits. Its inner mucosal surface is a dynamic habitat that is prone to various physical, inflammatory, immunologic, and malignant injuries. Acrolein, the simplest α,β-unsaturated aldehyde (systematic name: propenal), is a clear colorless-to-yellowish flammable liquid (at room temperature), with a molecular formula of C3H4O (CH2=CH-CHO). It is extremely acrid with a burnt, sweet, pungent, choking odor that occurs naturally in the environment as a result of combustion of wood. According to Environmental Protection Agency classification, acrolein is a high-priority air and water toxin which irritates mucous membranes.1 The general population is exposed to acrolein primarily by inhalation of air, especially indoor air containing environmental tobacco smoke.2 Acrolein is a highly reactive compound that rapidly binds to the sulfhydryl groups of proteins and other cellular molecules. Human and animal studies indicate that the primary target of acrolein toxicity following inhalation and oral and dermal exposure is the tissue at the site of contact, as demonstrated by irritation in the respiratory and gastrointestinal tracts, eyes, and skin. α,β-Unsaturated aldehydes are present in healthy subjects’ saliva and airway secretions in low-micromolar concentrations and have been found to elevate up to 10-fold in heavy smokers.3,4 Owing to its ubiquitous presence in the environment and high reactivity, toxic effects of acrolein on various cells and organs have been extensively studied. The effects of acrolein have been most extensively investigated following exposure by inhalation. Oral cavity tissues are the first to encounter the inhaled air and cigarette smoke, and their responses to harmful stimuli are critical in maintaining local homeostasis. Hence, considering the highest level of acrolein exposure through food substances and smoking and its potential chronic toxicity,5,6 we aimed to review further the oral cavity exposure and the noxious effect of acrolein in oral pathology, presented in our group’s recently published studies. ACROLEIN SOURCES Environmental Pollutant Acrolein is an important pollutant widely distributed in the environment as a major by-product of incomplete combustion of organic matter. It is also produced by photochemical oxidation of hydrocarbons of organic pollutants in the atmosphere.7 An estimated minimum of 218 tons of acrolein is released yearly to the atmosphere from anthropogenic sources involving fermentation and ripening processes, industrial waste incinerators, furnaces, fireplaces, power plants, combustion of polyethylene plastics, cigarette smoke, overheated cooking of food and oils, and release of volatile components in forest manufacturing processes.8,9 As a broad-band biocide, acrolein is intentionally released into the environment as a herbicide and an algaecide in water circuits, irrigation canals, drainage ditches, processing waters, cooling water towers, and water treatment basins to control the growth process of aquatic plants. The main microbiocidal “non-pesticidal” use of acrolein is the active ingredient in a product used by oil companies to cleanse hydrogen sulfide in order to control sulfide-producing bacteria and remove hydrogen sulfide and iron sulfide from oil production and injection wells. Acrolein can also solubilize ferrous sulfide deposits that obstruct wells, tanks, and barrels. Furthermore, acrolein has been known to be used as a component in military poison gas mixtures.2 The production of acrolein and its use as biocide result in its release to the environment, with the vast majority emitted to the air. In air, acrolein primarily undergoes degradation by means of reactions with photochemically generated hydroxyl radicals in the troposphere, where its half-life is 15–20 hours.7 In water and soil, volatilization and degradation are presumed to be the main removal process. In water, acrolein undergoes reversible hydration to form 3-hydroxypropanal, which undergoes aerobic biodegradation that occurs optimally in acclimated cultures.10 The estimated half-life in natural unsterilized water is 30–50 hours.11 The degradation of acrolein in soil is believed to occur through hydration, biodegradation, and irreversible binding to organic components in soil. The overall reactivity-based half-life of acrolein in soil is estimated to be between 30 and 100 hours.12 Exposure of the general population to acrolein occurs primarily by way of atmospheric contact.1 Humans’ predominant environmental exposure is by means of inhalation of cigarette smoke or automotive exhaust. The US Environmental Protection Agency reported mean ambient acrolein concentrations of 14.3 μg/m3 (6.2 ppb), ranging from 8.2 to 24.6 μg/m3 (3.6 to 10.7 ppb), based upon data from 1961 to 1980.1 Concentrations in indoor air may exceed outdoor levels 2- to 20-fold.2 Populations residing near waste disposal sites or municipal landfills might be subject to higher than average levels of acrolein in the air since acrolein is volatile, and also via drinking water obtained from groundwater wells. However, exposure to acrolein in drinking water is not considered to be a grave problem for these populations due to the volatilization of acrolein from soils and the retention or degradation of acrolein in soils.7 Acrolein Derived from Cigarette Smoke Human exposure to a large amount of acrolein is through cigarette smoke. Acrolein is present in very high concentrations in the vapor phase of all cigarettes, and its levels vary up to 10-fold between high-tar and ultralow-tar cigarette smoke extracts. It constitutes 50%–60% of the total vapor phase electrophiles.13 The International Agency for Research on Cancer14 noted that acrolein concentrations in smoke from various cigarettes ranged from 3 to 220 μg/cigarette. Jones15 reported concentrations of acrolein in mainstream smoke (defined as smoke that is directly exhaled from the smoker) ranging from 10 to 140 μg/cigarette and estimated concentrations in sidestream smoke (defined as smoke emitted from the smoldering tobacco between puffs) in the range of 100–1,700 μg/cigarette. Acrolein levels between 2.3 and 275 μg/m3 have been reported in smoky indoor environments such as bars and restaurants.14 The respiratory tract is commonly exposed to a range of α,β-unsaturated aldehydes from cigarette smoke exposure. It was estimated that, during cigarette smoking, acrolein concentrations on the airway surface may be as high as 80 μM.16 α,β-Unsaturated aldehydes are present in saliva and airway secretions in low-micromolar concentrations in healthy subjects and are elevated up to 10-fold in heavy smokers.17,18 It has been reported that inhalation of acrolein as a result of smoking may be a significant contributor to serious lung injury and death.7 Inhaled acrolein can be absorbed and may induce systemic effects by increasing platelet activation as a contributing factor to the prothrombotic risk in humans and may induce atherosclerosis and coronary artery disease.19 Moreover, human endothelial cells are particularly sensitive to acrolein. Acrolein adducts may cause systemic endothelial cell dysfunction and atherosclerosis.20 Acrolein Derived from Food Humans are exposed to acrolein by the ingestion of many foods.21,22 Such foods include fruits (raspberries, grapes, strawberries, and blackberries), vegetables, fish, caviar, doughnuts, cocoa beans, and some types of cheeses. Acrolein can be formed in the different food-processing stages from amino acids, animal fats, or carbohydrates. In fermentation, ripening, and heating and/or overheating, acrolein is formed during the Maillard reaction as a result of the conversion of amino acids and the oxidative deamination of polyamines.23–25 Acrolein was detected in non-alcoholic beverages such as coffee and tea.26 It also may be produced as an unwanted by-product during alcoholic fermentation or during the storage and maturation of alcoholic products and therefore might be detected in spirits and wines.27 Acrolein was detected in the emissions of varieties of heated cooking oils, e.g. during frying and deep frying, and therefore French fries present significant contents of acrolein.28 In addition acrolein is produced from thermal degradation of cellophane and polystyrene thermoplastics, which are used for food packaging.29 The World Health Organization (WHO) working group determined a tolerable oral acrolein intake of 7.5 μg/kg bodyweight/day.21 However, a reliable estimate of the acrolein exposure via food and water is virtually impossible. Endogenous Acrolein Production Acrolein is produced endogenously as well. One important intrinsic source is as a by-product of fragmentation of polyunsaturated fatty acids during lipid peroxidation (LPO). Lipid peroxidation is an oxidative degradation of lipids, such as following free radical damage to cell membranes. Numerous studies have revealed that LPO products are associated with the development of inflammation-related diseases, such as chronic obstructive pulmonary disease (COPD), and vascular diseases including atherosclerosis, Alzheimer’s disease, and stroke.30 The end-products of LPO are primarily reactive aldehydes, such as malondialdehyde (MDA) and 4-hydroxynonenal (HNE). In addition, the excretion of HPMA (S-(3-hydroxypropyl) mercapturic acid), a significant LPO degradation product and acrolein metabolite, is elevated in the urine of smokers to a level about twice that of non-smoking subjects.31 Another source of intrinsic acrolein is the metabolism of the amino acids methionine and threonine. The conversion of threonine into acrolein was found to be mediated by myeloperoxidase (MPO), an enzyme present in human neutrophils that plays a role in killing bacteria and other pathogens. In humans MPO-mediated oxidation of threonine can yield acrolein under conditions of acute oxidative stress, such as myocardial infarction and stroke.32 Polyamines, which are derived from arginine, are also a source of endogenous acrolein.33 CHEMICAL REACTIVITY—ACROLEIN CYTOTOXICITY Among all α,β-unsaturated aldehydes, acrolein is the strongest electrophile, which accounts for its high reactivity with nucleophiles.34 Unsaturated aldehyde reaction with cysteine, lysine, and histidine residues of proteins is by means of the Michael addition reaction. This reaction is referred to as “protein carbonylation” as a result of the addition of the aldehydic carbonyl group to the protein, resulting in a change in its chemical structure. Protein carbonyl content is the most general and well-used biomarker of severe oxidative protein damage.35 Acrolein adducts created in the thiol side chain of cysteine are considerably more stable than adducts formed by all other α,β-unsaturated aldehydes.13 Because acrolein readily reacts by making covalent adducts with sulfhydryl and amino groups of proteins, it is unlikely to distribute systemically, and thus its adverse effects are characterized in terms of cytotoxicity at the site of entry. It has been suggested that carbonylated proteins are involved in vital cellular functions including energy metabolism, protein synthesis, tissue damages, cytoskeletal integrity, and neuronal plasticity.36 Carbonylation at various nucleophilic sites implicated in acrolein cytotoxicity causes oxidative injury and disturbs the cell redox balance following an increased generation of reactive oxygen species (ROS) and glutathione (GSH) depletion.37,38 Thus the acrolein adduct mediates oxidative damage to cells and tissues as it impairs the structure and function of biomolecules that are considered to be involved in many pathological conditions.9 Moreover, it dysregulates major cellular pathways in the process of apoptosis, transcription, cell cycle control, protein biosynthesis, and cell signaling.39 Target organs of toxicity of acrolein are primarily the local tissues affected.21 Inhalation of acrolein causes irritation and inflammation of the respiratory tract followed by hyperplasia and metaplasia of the respiratory epithelium. Oral exposure causes gastric ulcers or bleeding.40 However, the oral cavity as a local anatomic structure is directly exposed to acrolein deleterious toxicity effects by way of increased proteins and DNA adducts, decreased GSH levels, and the impact on cell signaling pathways. THE ROLE OF ACROLEIN IN THE PATHOGENESIS OF ORAL INJURIES Oxidative Stress Cellular oxidative stress is implicated in the pathogenesis of many diseases, including strokes, myocardial infarction, and atherosclerosis. In the oral cavity, oxidative stress has been linked with impaired inflammatory response and the onset of periodontal (tooth supporting tissue) destruction.41 This is particularly important since there is overwhelming evidence that chronic periodontitis induces a state of systemic inflammation.42 Additionally, impaired redox balance is involved in many inflammatory disorders that have oral manifestations, such as lichen planus and pemphigus vulgaris.43,44 Acrolein changes the redox state of cells by triggering ROS formation or acts as an oxidant in a variety of cells, for instance in retinal cells, bronchial epithelial cells, as well as T cells.45–47 The reason for the redox change caused by acrolein is probably cellular GSH consumption due to the generation of GSH–acrolein adducts, which prevents its oxidation to glutathione disulfide (GSSG) and subsequent regeneration by the glutathione reductase enzyme.37 This has been observed among others in cardiac cells,18 pulmonary epithelia,46 and oral fibroblasts.48 Glutathione depletion could therefore be the mechanism behind cigarette smoke-induced cytotoxicity and could be correlated with the reduced reparative and regenerative activity of gingival and periodontal tissues previously reported in smokers. Systemic elimination of acrolein is performed in the liver by conjugation with GSH,31 but in the oral environment, where the exposure to acrolein is direct due to its solubility in saliva, the local elimination by conjugation with glutathione in the cells and the saliva may also be significant. In a unique smoking simulator apparatus direct exposure to cigarette smoke and acrolein caused an increase in cellular oxidative stress in a keratinocytic model of oral exposure.49 This was observed concurrently with a decrease in intracellular GSH and ratified that the mechanism of redox change caused by acrolein is probably via cellular GSH consumption in oral epithelial cells as well (Figure 1). Figure 1 Dichlorofluorescein (DCF) Assay for Cellular Total Oxidation State. I: Fluorescence intensity is proportional to reactive oxygen species and free radicals within the cytosol. A: Air subjected control cell culture; B: cellular oxidative status after a single puff of cigarette smoke; C: after exposure to 200 μmol acetaldehyde; D: after exposure to 1 μmol acrolein. II: Average fluorescence analyses of 3–5 different DCF experiments. Total cellular fluorescence of keratinocytes exposed to cigarette smoke with and without pre-incubation with N-Acetyl Cysteine (NAC). (*=Statistically significant). Reprinted from Figure 4 of Avezov K, et al.,49 with permission from Elsevier. Adduction to Amino Acids and Cross-linking of Proteins Another biological effect of acrolein is the result of its high reactivity with nucleophilic sites in proteins. In the oral cavity there is a large number of potential protein targets such as enzymes, glycoproteins, and immunoglobulins. In the saliva, inhibition of salivary enzymes such as lactate dehydrogenase (LDH), aspartate aminotransferase (AST), acid phosphatase, and amylase by cigarette smoke was attributed to its unsaturated aldehydic constituents, such as acrolein and crotonaldehyde.50,51 Dose-dependent carbonyl modifications in salivary proteins were observed when subjected to cigarette smoke and acrolein (Figure 2) but not to acetaldehyde, thus explaining the inhibition in their activities.52 Loss of function of salivary enzymes can cause serious damage to their digestive and antimicrobial role. More oral infections are observed in smokers. These include for instance increased likelihood of subgingival infection with certain periodontal pathogens53 and oral opportunistic infections such as Candida albicans colonization.54 Figure 2 Salivary Protein Carbonylation Assay. I (upper panel): A representative western blot (WB) analysis of total saliva proteins exposed to cigarette smoke. A: untreated control; B: after single puff of cigarette smoke; C: after three puffs of cigarette smoke; D: after nine puffs of cigarette smoke. I (lower panel): Average of densitometric analyses of three to five different WB assays of the same experiment. II (upper panel): A representative WB analysis of the total saliva proteins exposed to aldehydes. A: untreated control; B: acrolein content present in one cigarette (1 μmol); C: acrolein (10 μmol); D: acetaldehyde content present in one cigarette (20 μmol). II (lower panel): Average of densitometric analyses of three to five different WB assays of the same experiment. Reprinted from Figure 3 of Avezov K, et al.,52 with permission from Elsevier. The antioxidant properties of the saliva are dependent on the presence of some enzymatic and non-enzymatic antioxidant systems, such as superoxide dismutase enzyme (SOD), catalase, oral peroxidase enzymes, uric acid, and GSH. Salivary GSH content are also depleted due to acrolein exposure55 and could further reduce the protective role of saliva against oral oxidative stress. Carbonyl protein modifications were observed not only in the salivary fluid, but intracellularly as well, in two types of prevailing oral tissues: fibroblasts48 and epithelium (Figure 3). Aldehydes which are readily dissolved in aqueous saliva can easily penetrate cellular membranes and affect cellular contents. A number of stress, cytoskeletal, and redox signal proteins were identified as protein targets in proteomic analysis of bronchial epithelial cells exposed to acrolein.56 Figure 3 Intracellular Protein Carbonylation Assay. Left: A representative WB analysis of total intracellular proteins exposed to aldehydes. A: Air-exposed control; B: 1 μmol of acrolein (equal to 1 cigarette); C: 10 μmol of acrolein (equal to 10 cigarettes); D: 20 μmol of acetaldehyde (equal to 1 cigarette); E: 200 μmol of acetaldehyde (equal to 10 cigarettes). Right: Intracellular protein carbonyl ratio. Average densitometric analyses of three to five different WB assays of the same experiment. Reprinted from Figure 3, Panels II and III of Avezov K, et al.,49 with permission from Elsevier. Adduction of Acrolein to DNA The high chemical reactivity of acrolein allows it to react with a variety of macromolecules, including DNA. The reaction of acrolein with a guanosine analog in vitro was first described in 1984 by Chung et al.57 Since then, the damage by unsaturated aldehydes to DNA was reported in many tissues, especially in lung cells of smokers, where it was suggested to represent a major etiological agent for cigarette smoke-related lung cancer.58 A pattern of DNA damage in the p53 tumor suppressor gene produced from acrolein exposure resembles the p53 mutation patterns found in lung cancer. Cigarette smokers are at substantially greater mortality risk from oral cancer than are non-smokers. Although estimates vary, most studies have reported mortality ratios of about 5–10:1 for never smokers versus smokers. Furthermore, the risk of death from oral cancer is cigarette smoke consumption-related; the more cigarettes consumed daily and the more years one has smoked, the greater the risk. In salivary analyses of oral squamous cell carcinoma patients, the levels of 8-oxoguanine DNA glycosylase, an enzyme involved in base excision repair, were found to be low;59 8-hydroxy-deoxyguanosine (8-OHdG), a widely used indicator of DNA oxidation was found to be higher by 65% along with reduced salivary antioxidants.60 Therefore, there is a foundation for the belief that the same mechanisms of oxidative stress induction and DNA damage by acrolein might incite oral cancer as well as lung cancer. Very little is known about the reactivity of acrolein with DNA in oral cells. One of the future objectives of our study group is to investigate and prove this deleterious connection. Modulation of Gene Activation by Acrolein A few transcription factors are regulated by the redox state of the cells. These include oxidative stress-sensing nuclear factor kB (NF-kB) and nuclear erythroid-2 related factor 2 (Nrf2). Nuclear factor kB is involved in the expression of more than 400 genes responsible for the regulation of cellular responses to stimuli such as stress, inflammation, free radicals, radiation, and bacterial or viral antigens. The Nrf2 signaling pathway regulates the cellular redox homeostasis and enzymatic protection against oxidative and electrophilic stress, including the production of enzymes involved in the GSH biosynthesis pathway. Both NF-kB and Nrf2 are present in the cell cytosol in an inactive state, and their release by one of the above-mentioned stimuli permits their translocation to the nucleus, where they regulate gene expression. Lambert et al.47 reported the immunosuppressing effects of cigarette smoke on human T cells. They showed that cigarette smoke extract inhibited transcription of cytokine genes and the production of proinflammatory cytokines, including interleukin-2 (IL-2), IL-10, granulocyte-macrophage colony-stimulating factor, interferon-c, and tumor necrosis factor-a (TNF-a). The cigarette smoke compound responsible for this inhibition appeared to be acrolein, through a direct modification of the DNA-binding domain of the NF-kB pathway.31 This NF-kB-mediated gene suppression might be responsible for the reduced periodontal immunologic response to infection found in smokers.53 The Nrf2 signaling pathway is also affected by acrolein. It is involved in the cellular protection after exposure to cigarette smoke and aldehydes in the respiratory system and ocular epithelium.61 Wu et al.62 revealed increased expression of the antioxidant Nrf2-activated enzyme HO-1 in endothelial cells that were exposed to acrolein. Exposure of human lung epithelial (A549) cells to acrolein first depleted 80% of the intracellular GSH and then increased the transcription of c glutamylcysteine synthetase, Nrf2-activated enzyme, resulting in normalization of GSH levels.63 The activation of the transcription factors such as Nrf2 pathway in oral cells following cigarette smoke and acrolein exposure has not yet been studied and is currently under extensive investigation by our study group. IN CONCLUSION The effect of acrolein in the oral cavity is summarized in Figure 4. Due to its high reactivity, acrolein quickly degrades in water, air, and soil. Thus, the direct pathologic damage to the general population from environmental sources is very difficult to establish. However, in oral tissues of smokers, constantly exposed to high levels of acrolein, the toxic effects of acrolein could be the cause of additive damage especially because of its high reactivity at the site of contact. The current paper summarizes for the first time the diverse sources of acrolein and its mechanisms of toxicity in the oral cavity. Further studies should focus on protective pathways, which to date have only been partially revealed. Figure 4 The Summary of Acrolein Effect on the Oral Cavity: The Impact on Saliva and Oral Cells. Conflict of interest: No potential conflict of interest relevant to this article was reported. Abbreviations GSHglutathione LPOlipid peroxidation MPOmyeloperoxidase NF-kBnuclear factor kB ROSreactive oxygen species ==== Refs REFERENCES 1 DeWoskin RS Greenberg M Pepelko W Strickland J Toxicological review of acrolein Support of Summary Information on the Integrated Risk Information System (IRIS) (CAS No. 107-02-8) Washington, DC US Environmental Protection Agency 2003 2 Gomes R Meek ME Eggleton E Concise International Chemical Assessment Document 43 Acrolein Geneva World Health Organization 2002 3 Andreoli R Manini P Corradi M Mutti A Niessen WM Determination of patterns of biologically relevant aldehydes in exhaled breath condensate of healthy subjects by liquid chromatography/atmospheric chemical ionization tandem mass spectrometry Rapid Commun Mass Spectrom 2003 17 637 45 10.1002/rcm.960 12661015 4 Annovazzi L Cattaneo V Viglio S High-performance liquid chromatography and capillary electrophoresis: methodological challenges for the determination of biologically relevant low-aliphatic aldehydes in human saliva Electrophoresis 2004 25 1255 63 10.1002/elps.200305843 15174046 5 Ismahil MA Hamid T Haberzettl P Chronic oral exposure to the aldehyde pollutant acrolein induces dilated cardiomyopathy Am J Physiol Heart Circ Physiol 2011 301 H2050 60 10.1152/ajpheart.00120.2011 21908791 6 Weinhold B Acrolein and neuro disorders Environ Health Perspect 2011 119 A68 10.1289/ehp.119-a68a 21285012 7 Ghilarducci DP Tjeerdema RS Fate and effects of acrolein Rev Environ Contam Toxicol 1995 144 95 146 10.1007/978-1-4612-2550-8_2 8599034 8 Lipari F Dasch JM Scruggs WF Aldehyde emissions from wood-burning fireplaces Environ Sci Technol 1984 18 326 30 10.1021/es00123a007 22280078 9 Aldini G Orioli M Carini M Protein modification by acrolein: relevance to pathological conditions and inhibition by aldehyde sequestering agents Mol Nutr Food Res 2011 55 1301 19 10.1002/mnfr.201100182 21805620 10 Bowmer KH Higgins ML Some aspects of the persistence and fate of acrolein herbicide in water Arch Environ Contam Toxicol 1976 5 87 96 10.1007/BF02220892 13740 11 Lijinsky W Reuber MD Chronic carcinogenesis of acrolein and related compounds Toxicol Ind Health 1987 3 337 45 10.1177/074823378700300306 3686537 12 Mackay JM Dose selection in in vivo genetic toxicology assays Environ Mol Mutagen 1995 25 323 7 10.1002/em.2850250410 7607188 13 Esterbauer H Schaur RJ Zollner H Chemistry and biochemistry of 4-hydroxynonenal, malonaldehyde and related aldehydes Free Radic Biol Med 1991 11 81 128 10.1016/0891-5849(91)90192-6 1937131 14 Working Group of the International Agency for Research on Cancer IARC monographs on the evaluation of the carcinogenic risk of chemicals to humans: some monomers, plastics and synthetic elastomers, and acrolein IARC Monogr Eval Carcinog Risk Chem Hum 1979 19 1 513 285915 15 Jones AP Indoor air quality and health Atmos Environ 1999 33 4535 64 10.1016/S1352-2310(99)00272-1 16 Rahman I MacNee W Lung glutathione and oxidative stress: implications in cigarette smoke-induced airway disease Am J Physiol 1999 277 6 Pt 1 L1067 88 10600876 17 Yao H Rahman I Current concepts on oxidative/carbonyl stress, inflammation and epigenetics in pathogenesis of chronic obstructive pulmonary disease Toxicol Appl Pharmacol 2011 254 72 85 10.1016/j.taap.2009.10.022 21296096 18 Luo J Hill BG Gu Y Mechanisms of acrolein-induced myocardial dysfunction: implications for environmental and endogenous aldehyde exposure Am J Physiol Hear Circ Physiol 2007 293 H3673 84 10.1152/ajpheart.00284.2007 19 Sithu SD Srivastava S Siddiqui MA Exposure to acrolein by inhalation causes platelet activation Toxicol Appl Pharmacol 2010 248 100 10 10.1016/j.taap.2010.07.013 20678513 20 Uchida K Role of reactive aldehyde in cardiovascular diseases Free Radic Biol Med 2000 28 1685 96 10.1016/S0891-5849(00)00226-4 10946210 21 Abraham K Andres S Palavinskas R Berg K Appel KE Lampen A Toxicology and risk assessment of acrolein in food Mol Nutr Food Res 2011 55 1277 90 10.1002/mnfr.201100481 21898908 22 Kuper H Adami HO Boffetta P Tobacco use, cancer causation and public health impact J Intern Med 2002 251 455 66 10.1046/j.1365-2796.2002.00993.x 12028500 23 Beauchamp RO Jr Andjelkovich DA Kligerman AD Morgan KT Heck HD A critical review of the literature on acrolein toxicity Crit Rev Toxicol 1985 14 309 80 10.3109/10408448509037461 3902372 24 Hirayama T A large-scale cohort study on risk factors for primary liver cancer, with special reference to the role of cigarette smoking Cancer Chemother Pharmacol 1989 23 S114 17 2924384 25 Lane RH Smathers JL Monitoring aldehyde production during frying by reversed-phase liquid chromatography J Assoc Off Anal Chem 1991 74 957 60 1757421 26 Feron VJ Til HP de Vrijer F Woutersen RA Cassee FR van Bladeren PJ Aldehydes: occurrence, carcinogenic potential, mechanism of action and risk assessment Mutat Res 1991 259 363 85 10.1016/0165-1218(91)90128-9 2017217 27 Moghe A Ghare S Lamoreau B Molecular mechanisms of acrolein toxicity: relevance to human disease Toxicol Sci 2015 143 242 55 10.1093/toxsci/kfu233 25628402 28 Srivastava S Sithu SD Vladykovskaya E Oral exposure to acrolein exacerbates atherosclerosis in apoE-null mice Atherosclerosis 2011 215 301 8 10.1016/j.atherosclerosis.2011.01.001 21371710 29 Zitting A Heinonen T Decrease of reduced glutathione in isolated rat hepatocytes caused by acrolein, acrylonitrile, and the thermal degradation products of styrene copolymers Toxicology 1980 17 333 41 10.1016/0300-483X(80)90014-1 7210018 30 Lee SE Park YS Role of lipid peroxidation-derived alpha, beta-unsaturated aldehydes in vascular dysfunction Oxid Med Cell Longev 2013 2013 629028 10.1155/2013/629028 23819013 31 Stevens JF Maier CS Acrolein: sources, metabolism, and biomolecular interactions relevant to human health and disease Mol Nutr Food Res 2008 52 7 25 10.1002/mnfr.200700412 18203133 32 Baker J Arey J Atkinson R Formation and reaction of hydroxycarbonyls from the reaction of OH radicals with 1,3-butadiene and isoprene Environ Sci Technol 2005 39 4091 9 10.1021/es047930t 15984787 33 Sakata K Kashiwagi K Sharmin S Ueda S Igarashi K Acrolein produced from polyamines as one of the uraemic toxins Biochem Soc Trans 2003 31 371 4 10.1042/bst0310371 12653641 34 Ishizuka T Hayashi T Nakazawa H Serum vascular endothelial growth factor is a candidate biomarker of renal cell carcinoma in hemodialysis patients Nephron 2000 84 83 4 10.1159/000045545 10644915 35 Grimard V Li C Ramjeesingh M Bear CE Goormaghtigh E Ruysschaert JM Phosphorylation-induced conformational changes of cystic fibrosis transmembrane conductance regulator monitored by attenuated total reflection-Fourier transform IR spectroscopy and fluorescence spectroscopy J Biol Chem 2004 279 5528 36 10.1074/jbc.M311014200 14660584 36 Mello CF Sultana R Piroddi M Acrolein induces selective protein carbonylation in synaptosomes Neuroscience 2007 147 674 9 10.1016/j.neuroscience.2007.04.003 17570602 37 Kehrer JP Biswal SS The molecular effects of acrolein Toxicol Sci 2000 57 6 15 10.1093/toxsci/57.1.6 10966506 38 Uchida K Kanematsu M Sakai K Protein-bound acrolein: potential markers for oxidative stress Proc Natl Acad Sci U S A 1998 95 4882 7 10.1073/pnas.95.9.4882 9560197 39 Thompson CA Burcham PC Genome-wide transcriptional responses to acrolein Chem Res Toxicol 2008 21 2245 56 10.1021/tx8001934 19548348 40 Faroon O Roney N Taylor J Ashizawa A Lumpkin MH Plewak DJ Acrolein health effects Toxicol Ind Health 2008 24 447 90 10.1177/0748233708094188 19028774 41 Chapple ILC Matthews JB The role of reactive oxygen and antioxidant species in periodontal tissue destructtion Periodontol 2000 2007 43 160 232 10.1111/j.1600-0757.2006.00178.x 17214840 42 D’Aiuto F Nibali L Parkar M Patel K Suvan J Donos N Oxidative stress, systemic inflammation, and severe periodontitis J Dent Res 2010 89 1241 6 10.1177/0022034510375830 20739696 43 Sezer E Ozugurlu F Ozyurt H Sahin S Etikan I Lipid peroxidation and antioxidant status in lichen planus Clin Exp Dermatol 2007 32 430 4 10.1111/j.1365-2230.2007.02436.x 17459065 44 Yesilova Y Ucmak D Selek S Oxidative stress index may play a key role in patients with pemphigus vulgaris J Eur Acad Dermatol Venereol 2013 27 465 7 10.1111/j.1468-3083.2012.04463.x 22324759 45 Li X Liu Z Luo C Lipoamide protects retinal pigment epithelial cells from oxidative stress and mitochondrial dysfunction Free Radic Biol Med 2008 44 1465 74 10.1016/j.freeradbiomed.2008.01.004 18258206 46 Grafstrom RC Dypbukt JM Willey JC Pathobiological effects of acrolein in cultured human bronchial epithelial cells Cancer Res 1988 48 1717 21 3349453 47 Lambert C McCue J Portas M Acrolein in cigarette smoke inhibits T-cell responses J Allergy Clin Immunol 2005 116 916 22 10.1016/j.jaci.2005.05.046 16210070 48 Colombo G Dalle-Donne I Orioli M Oxidative damage in human gingival fibroblasts exposed to cigarette smoke Free Radic Biol Med 2012 52 1584 96 10.1016/j.freeradbiomed.2012.02.030 22387198 49 Avezov K Reznick AZ Aizenbud D Oxidative damage in keratinocytes exposed to cigarette smoke and aldehydes Toxicol In Vitr 2014 28 485 91 http://www.journals.elsevier.com/toxicology-in-vitro 50 Zappacosta B Persichilli S Mordente A Inhibition of salivary enzymes by cigarette smoke and the protective role of glutathione Hum Exp Toxicol 2002 21 7 11 10.1191/0960327102ht202oa 12046726 51 Nagler R Lischinsky S Diamond E Drigues N Klein I Reznick AZ Effect of cigarette smoke on salivary proteins and enzyme activities Arch Biochem Biophys 2000 379 229 36 10.1006/abbi.2000.1877 10898939 52 Avezov K Reznick AZ Aizenbud D LDH enzyme activity in human saliva: the effect of exposure to cigarette smoke and its different components Arch Oral Biol 2014 59 142 8 http://www.journals.elsevier.com/archives-of-oral-biology/ 24370185 53 Zambon JJ Grossi SG Machtei EE Ho AW Dunford R Genco RJ Cigarette smoking increases the risk for subgingival infection with periodontal pathogens J Periodontol 1996 67 1050 4 10.1902/jop.1996.67.10s.1050 54 Soysa NS Ellepola ANB The impact of cigarette/tobacco smoking on oral candidosis: an overview Oral Dis 2005 11 268 73 10.1111/j.1601-0825.2005.01115.x 16120112 55 Zappacosta B Persichilli S De Sole P Mordente A Giardina B Effect of smoking one cigarette on antioxidant metabolites in the saliva of healthy smokers Arch Oral Biol 1999 44 485 8 10.1016/S0003-9969(99)00025-4 10401526 56 Spiess PC Deng B Hondal RJ Matthews DE van der Vliet A Proteomic profiling of acrolein adducts in human lung epithelial cells J Proteomics 2011 74 2380 94 10.1016/j.jprot.2011.05.039 21704744 57 Chung FL Young R Hecht SS Formation of cyclic 1,N2-propanodeoxyguanosine adducts in DNA upon reaction with acrolein or crotonaldehyde Cancer Res 1984 44 990 5 6318992 58 Feng Z Hu W Hu Y Tang M Acrolein is a major cigarette-related lung cancer agent: preferential binding at p53 mutational hotspots and inhibition of DNA repair Proc Natl Acad Sci U S A 2006 103 15404 9 10.1073/pnas.0607031103 17030796 59 Shpitzer T Hamzany Y Bahar G Salivary analysis of oral cancer biomarkers Br J Cancer 2009 101 1194 8 10.1038/sj.bjc.6605290 19789535 60 Nagler RM Klein I Zarzhevsky N Drigues N Reznick AZ Characterization of the differentiated antioxidant profile of human saliva Free Radic Biol Med 2002 32 268 77 10.1016/S0891-5849(01)00806-1 11827752 61 Kosmider B Messier EM Chu HW Mason RJ Human alveolar epithelial cell injury induced by cigarette smoke PLoS One 2011 6 e26059 10.1371/journal.pone.0026059 22163265 62 Wu CC Hsieh CW Lai PH Lin JB Liu YC Wung BS Upregulation of endothelial heme oxygenase-1 expression through the activation of the JNK pathway by sublethal concentrations of acrolein Toxicol Appl Pharmacol 2006 214 244 52 10.1016/j.taap.2005.12.013 16480751 63 Tirumalai R Rajesh Kumar T Mai KH Biswal S Acrolein causes transcriptional induction of phase II genes by activation of Nrf2 in human lung type II epithelial (A549) cells Toxicol Lett 2002 132 27 36 10.1016/S0378-4274(02)00055-3 12084617

PMC005xxxxxx/PMC5001797.txt

==== Front Rambam Maimonides Med JRambam Maimonides Med JRMMJRambam Maimonides Medical Journal2076-9172Rambam Health Care Campus 10.5041/RMMJ.10252rmmj-7-3-e0025Rambam Grand RoundsCancer of the Throat: A Physician’s Experience as a Patient Brook Itzhak M.D., M.Sc.*Professor of Pediatrics, Georgetown University School of Medicine, Washington DC, USA* E-mail:ib6@georgetown.edu.7 2016 28 7 2016 7 3 e0025Copyright: © 2016 Brook I.2016This is an open-access article. All its content, except where otherwise noted, is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.The author, a practicing physician, was diagnosed with throat cancer and lost his vocal cords. He endured the side effects of radiation, repeated surgeries, and the effects of prolonged hospitalizations; confronted medical mistakes and discrimination after losing his vocal cords; and struggled to regain his speech and find new meaning and purpose for his life. Facing the hardship and trials of becoming a laryngectomee illustrated to him how dependent and helpless a patient can become. Being unable to speak, eat, and breathe normally, while dealing with a potentially terminal illness, makes the patient very vulnerable, both physically and emotionally. A skillful, competent, error-free, empathetic, and caring approach that recognizes what the patient is experiencing can expedite recovery and well-being and help the patient return to a productive and meaningful life. Cancerdepressionlaryngectomymedical errorsthroat ==== Body INRODUCTION This manuscript describes my personal experiences enduring several surgical procedures, including radical neck dissection and laryngectomy for removal of pyriform sinus squamous cell carcinoma. As an infectious diseases clinician with an interest in head and neck infections, I had little knowledge or experience in the management of head and neck cancer. I was suddenly thrust into an unfamiliar and challenging reality where I had to cope with a life-threatening malignancy as a patient rather than in the role of a clinician. In addition to experiencing the untoward effects of radiation treatments, surgery, and prolonged hospitalizations, I suffered from diagnostic and therapeutic errors related to my care, as well as discrimination. It was a struggle to learn to speak again and a challenge to find new purpose for my life. I share these experiences with the hope that medical providers will better recognize and appreciate the hardships faced by their patients diagnosed with cancer. My goal is to encourage proficient, competent, error-free, and compassionate care, and to facilitate patient recovery and well-being. FACING THROAT CANCER I was shaken to the core upon being diagnosed with hypo-pharyngeal cancer. I refused to believe my diagnosis until I had seen the pathological specimens with my own eyes.1 I had to accept facts that I had previously avoided—I was vulnerable and my life might be shortened due to this illness. I had always believed I would stay in good health and be free of any serious diseases. As I grew older I began to accept the inevitable reality that illness might ultimately lead to my demise; however, such thoughts were for a far distant future. Without warning I faced an unforeseen danger and the potential of imminent death. Although my cancer had been detected and removed at an early stage (T1, N0, M0) and had not spread, my prognosis was unclear. I was left with a lingering fear that the cancer would return or metastasize. Recovering from the side effects of radiotherapy (altered taste, inflamed mucosa, reduced saliva production, hypothyroidism) was a lengthy process. As time passed I began to believe I might be cancer-free. However, a year and a half after completing radiotherapy my symptoms reoccurred—pain and difficulty swallowing. Although I hoped otherwise, a biopsy revealed the return of the malignancy. I did not share my diagnosis with many others at first—only family and work colleagues. I did not want to appear weak or helpless or be stigmatized and discriminated against. However, with the cancer’s return I could no longer hide my illness—future surgeries might lead to my disfigurement. I was surprised to find that as I began sharing my diagnosis with others, the feelings of isolation and embarrassment passed and I greatly benefited from their much-needed support and sympathy. SELECTING THE APPROPRIATE TREATMENT With the recurrence of the cancer, two treatment options were presented to me: removal of my entire larynx versus removal of only a portion of my larynx. After further consideration my physicians proposed preserving my vocal cords by using only laser surgery to remove the cancerous tissue. However, the final decision was up to me. I therefore consulted with several head and neck cancer specialists from the US and abroad. I asked each one about the advantages and disadvantages of the different surgical approaches for my specific cancer (T2, N0, M0). Some believed in performing total laryngectomy while others believed that laser, when used by experienced operators, would work. Wishing to save my vocal cords and to avoid major surgery, I finally chose laser surgery. I now had to select the hospital where the surgery would be performed. One hospital was out of town; although they had a great deal of experience with trans-oral laser microsurgery for head and neck cancer, there was a long waiting time. The other hospital was close to home and could schedule the procedure much earlier. It was the hospital where I practiced, and I knew the physicians personally. However, they had minimal experience with this surgery. In considering my options, the earlier surgical date, familiarity with the staff, and the fact that I felt too emotionally and physically drained to travel were the deciding factors: I selected the local hospital. To my great disappointment, despite three separate attempts within three weeks, the surgeons were unable to remove the tumor. Initially they had been optimistic. However, after each procedure the final pathological report returned showing residual cancer. Over those three weeks my life turned into a draining emotional up-and-down ride from initial feelings of elation to disappointment and despair. Having failed to remove the cancer, my surgeons began planning a partial or complete laryngectomy. However, I was beginning to doubt their competence and sought the opinion of a surgeon who was more experienced in treating my kind of cancer; he practiced at a large medical center in a different city. My new surgeon recommended radical surgery that evolved into a total laryngectomy. He explained that this surgery offered the highest chances of survival and ultimate cure for someone in my condition. I agreed. I had to wait an additional four weeks, but the surgery was successful, and the malignancy was excised in its entirety. To date the cancer has not returned locally or systemically. HOSPITALIZATION FOLLOWING THE SURGERIES I had no way of knowing what I would feel or endure following the operations. I experienced sedation, frailty, dizziness, and clouded sensations. Being unable to eat or drink by mouth and relying on tube feedings only grew more difficult over time. I began to have dreams where I was eating and drinking normally again, and was terrified to be caught since it was forbidden to me. I lay helpless in bed, linked up to catheters, suction and oxygen tubes, intravenous and arterial lines, and connected to monitoring equipment. Blood samples were taken multiple times a day, and an intravenous line had to be reinserted more than once. Over time, finding venous access became increasingly difficult. I had to be suctioned multiple times a day to remove the accumulated mucus and sputum. Swallowing saliva was so painful that I spit it out or let it drool down my face. Unable to speak, the only way I could communicate was by writing on a chalkboard or in a notebook. Adjusting to this new living reality was more than difficult. At this point in my treatment, I began to understand why some individuals elected to forgo life-extending interventions. This realization was new to me—I had always supported prolonging patients’ lives. For the first time it dawned on me that prolonging life might not always be the optimal choice. The constant presence of family members and friends became increasingly difficult and taxing. Yes, even though they were supportive, and ensured that I received appropriate care, sometimes I wanted to be left alone, longed for privacy, and just wanted to rest. Dealing with my visitors’ needs for reassurance and support was also difficult. They meant well, but at times their extended stays and conversations were an additional burden to me. PHYSICIANS’ ATTITUDE I encountered two different kinds of physicians throughout my fight with cancer: optimists and pessimists. Being a medical professional made it easier for me to navigate the medical system, obtain treatment, and get answers to questions from the majority of my doctors. I preferred clinicians who were willing to discuss all the possible risks, complications, and my long-term prognosis, even if it was not a good one. I was aware of their professional constraints and was grateful whenever they were open, sincere, and accepted responsibility for their errors. Several of my doctors probably concluded that I understood and knew more about my illness and treatment options than I did. I repeatedly had to ask them to explain everything to me and not to assume that I knew everything. Colleagues with whom I had previously worked gave me greater individual attention than those with whom I had not. Being a clinician myself, I always wanted the full details about my condition including my test results and differential diagnoses. Expressing my opinions on familiar issues was at times helpful. I also tried to keep my physicians apprised regarding any new symptoms and signs that might help in their care for me. Doing this helped them recognize that my cancer had returned. However, after losing the ability to speak, being only able to write made communication much more difficult. Some of my surgeons had no patience to “listen” in this manner. They always seemed to be in a rush to continue their medical rounds and get back to the operating room or clinic. They spent as little time as possible by my bedside, often two to three minutes, and only examined my upper airways and surgical site. Most did not even carry a stethoscope, and I had to insist that they listen to my chest. Many of the nurses were inconsistent in their examinations and often ignored important physical systems. Frustrated by my surgeons’ superficial approach and unable to get the input I needed from them, I began writing down all my inquiries and concerns before the medical rounds. This ensured that I got some responses. However, I rarely had an opportunity to ask new or follow-up questions during the rounds. All too often they left me with numerous unanswered inquiries and I felt upset and ignored. Thankfully, some surgeons and residents were more attentive and showed greater sensitivity and compassion. However, I also observed how the impatient, insensitive, and rushed attitude of some senior surgeons served as an inappropriate role model to many physicians in training. On several occasions I actually faced very rude clinicians. Once, my tracheotomy tube became obstructed and I urgently needed it cleaned. I asked a senior resident to perform the task. He did so grudgingly, rinsing it out with tap water instead of the proper sterile cleaning kit and water. As he finished, I could still feel that the tracheotomy tube was clogged. However, he refused to clean it again with the appropriate cleaning kit. He harshly uttered: “We call the shots here,” and exited the room. This humiliating experience again left me feeling helpless and quite upset. When I told the attending surgeon about the incident he seemed uninterested took no action to reprimand the resident. Despite these upsetting events, I am deeply grateful to the clinicians, nurses, and technicians who cared for me. The removal of my larynx and the reconstruction of my upper airways were impeccable. The majority of the professionals who cared for me were empathetic and personable and conveyed their compassion and sympathy. THE POWER OF A HUG In some situations, the best thing your medical professional can do is to give you a hug. I first experienced this after I informed my internist that my throat biopsy revealed the presence of cancer. Without saying a word, my physician approached me and gave me a warm hug. I was distraught after learning about the cancer’s recurrence, and his hug conveyed that he felt my pain and distress. I realized that I would not be alone in fighting the cancer and that I would be supported by medical professionals who sincerely cared for me. The hug was a genuine act of sustenance and sympathy and was given at the moment I needed it most. This spontaneous gesture of human contact was a refreshing return to the fundamentals of medicine that has been forgotten in modern medicine; all too often technology and tests replace human contact. A “hug” can be manifested in a variety of ways. It can be expressed through a pat on the back, a heartfelt handshake, or by eye contact. Indeed, there is scientific confirmation that human contact can lead to the production of chemicals such as oxytocin and endorphins that can reduce pain and generate calmness and happiness.2,3 MEDICAL MISTAKES The care I received from medical professionals at clinics and hospitals was generally very good. However, errors were made by physicians, nurses, medical technicians, and clerks, at all levels of my care. As a medical professional, I recognized many errors and was able to prevent some of them. Unfortunately, patients without a medical background may not be able to recognize and abort mistakes and are more likely to suffer their consequences. Realizing how prevalent medical errors are was shocking to me. Although I have been practicing medicine for over 40 years, I had never observed so many mistakes in patient care as while I myself was a patient. This was probably because I had never before suffered from a serious illness nor been hospitalized for a serious illness. After becoming a laryngectomee I had difficulty preventing and reporting many of these errors. However, I rapidly learned that it was up to me to ensure that mistakes did not occur. The first mistake occurred when my otolaryngologist attempted to remove my hypopharyngeal tumor via laser surgery, but instead excised scar tissue. The error was not recognized until a week later because the excised tissues were not inspected using frozen sections in the operating room. Two additional surgeries followed, but the tumor was not removed in its entirety. Post-surgical edema contributed to the difficulty in removing the entire malignancy during the repeat surgeries. I was exposed to several hazardous situations because of nurses’ errors. While still in the surgical intensive care unit a mucous plug suddenly blocked my airways. I wanted to alert the nurses but my call button had fallen to the floor. I attempted to get the nurses’ attention by unplugging my monitors one by one. Although my room was in front of the nurses’ station, no one came to my aid despite the sounding of the alarm signals from all my monitors. No help came for at least 10 minutes, and then only because my wife entered my room and saw that I was having trouble breathing. Multiple mistakes were made by the hospital staff, including failing to wash hands or use gloves when needed; taking oral temperature without covering the thermometer with protective plastic; taking blood pressure with cuffs that were too small or too large; administering incorrect medication doses or delivering drugs by mouth instead of through the nasogastric tube; improper suctioning of mucous and cleaning of the tracheal tube; failing to connect the call button; forgetting to enter physician orders to my chart, and entering a different patient’s orders to my chart. The result of the last-mentioned error was that oral feeding was started prematurely, 7 days after my laryngectomy instead of 14 days later. I only suspected that this was inappropriate because my surgeon had previously informed me when oral feeding would be resumed. Despite my repeated request to get a confirmation from a senior surgeon, oral feeding was continued for almost 18 hours until the error was recognized. The error apparently occurred because an order to start oral feeding for another patient was entered to my chart. Miraculously these incidents did not cause long-term harm. However, once I appreciated the prevalence and the magnitude of these errors, I started questioning every order and procedure made by the medical professionals. This made my hospital stay very unpleasant and stressful. Conveying my concerns whenever I observed errors was very difficult because I was unable to speak, was very weak, and was receiving medications that blurred my sensations. I also suppressed my desire to question or criticize the medical staff because of my dependence on them; I did not want to anger or upset them. I realized that the only way to prevent mistakes was by fending for myself and to become my own advocate without fear of repercussions. I tried to get the assistance of the hospital’s patient advocates several times but was disappointed at their reluctance to intervene or inability to help. I realized how important it is for a patient to have a family member or friend act as their advocate. I was fortunate that my family members were able to report some of the errors; this led to improvement in the care I received. The way to prevent medical mistakes is for the medical team to talk honestly about them with their patients. Medical errors reduce patients’ trust in the system and the people who care for them. Admitting and accepting responsibility for a mistake, when appropriate, and having an open dialogue about it can restore trust and contribute to patients’ well-being. Once a patient realizes that his/her concerns are acknowledged and steps are taken to prevent future errors they can start relaxing and concentrate on their recovery.4 LIFE AFTER LARYNGECTOMY Even though my surgeons attempted to prepare me for total laryngectomy and its aftermath, it was difficult for me fully to absorb the information I received during the difficult days prior to surgery. Hence, I was not fully able to deal with life afterward. No words could convey the feelings and consequences of becoming a laryngectomee until I actually became one. It was impossible to appreciate the reality of my losses, which included losing my voice, all sensations in parts of my face and neck, and the ability to smell. My daily life changed in unanticipated ways, and many things that I took for granted were gone. I had to contend with reflux of food and liquid when bending; the inability to swallow dry food or to converse during meal times; reduced neck and shoulder range of motion; no sensation over the skin donor site on my left arm; having a stoma in my neck, edema of my face and neck, and losing my Adam’s apple; repeated bouts of uncontrolled coughing; the constant need to clean my voice prosthesis and trachea; obtaining supplies needed to speak with a voice prosthesis and carrying them with me at all times; ensuring that my stoma was covered at all times with a base plate properly glued to my skin; learning to speak using a voice prosthesis and facing technical problems in its maintenance; losing my voice when the voice prosthesis was blocked; being misunderstood when speaking in noisy places or over the phone; difficulties in expressing feelings or laughing; and facing discrimination. The above obstacles and unforeseen hardships made my life challenging—to say the least. I had to find a way to eat and speak again and find a meaningful and productive way of living. I constantly reminded myself that dealing with these difficulties was better than the alternative of dying from cancer. OVERCOMING THE FEAR OF RECURRENCE AND DEPRESSION I faced depression, hopelessness, and fear of cancer recurrence following my surgeries. Initially, feeling depressed had enabled me to face being newly diagnosed with cancer and feeling hopeless had a paradoxical calming effect. Fortunately, I quickly realized that giving up on life would send the wrong message to my children and I would set the wrong example for them. However, after becoming a laryngectomee, battling depression became a constant fight when confronted with new norms and restrictions. My low levels of thyroid hormone, anemia, and physical exhaustion further contributed to the down feeling. I learned that there were some things I could do to reverse the depression. Slowly returning to enjoyable and gratifying tasks such as teaching, caring for my own patients, and writing, were all very helpful. My family and most of my friends were supportive of my endeavors, and my colleagues and patients accepted me without reservation. Also instrumental to overcoming the feelings of depression and hopelessness were a local laryngectomee support group, a psychologist, a caring and supportive head and neck surgeon, and a speech and language pathologist (SLP). I also took advantage of the advice and camaraderie available over the Internet. There are several laryngectomee and head and neck patient support group sites where head and neck cancer survivors from across the globe help each other with advice and support. This is a treasured resource for individuals traveling a similar road. Fear of the cancer’s return was particularly acute before taking tests to monitor my condition, such as computed tomography and positron emission tomography scans. I slowly learned to accept living with the continuous uncertainty of a cancer diagnosis; as time passed my anxiety subsided. LEARNING TO SPEAK AGAIN Relearning how to speak was an uphill struggle full of frustration and pitfalls. Speech was impossible for more than nine weeks following my laryngectomy. An electrical larynx would have helped me to speak sooner, but that option was not offered to me. The guidance of an experienced and dedicated SLP immediately after losing my vocal cords was most helpful. However, the SLP who worked with me after I was discharged from the hospital was not as knowledgeable and did not recognize my need for more guidance and follow-up. I quickly learned that the field of speech pathology is both an art and a science; each SLP has her or his unique approach to solving problems, and I could benefit from all of them. Finding what worked best for me was a trial and error process.5 I had to learn how to cope with voice prosthesis leakage, find out the optimal method for placing and sealing the baseplate for the heat–moisture exchanging filter, and perfect the way I spoke. My voice came out in a soft rusty whisper after my laryngectomy—it barely resembled my original voice. Speaking was a laborious task that involved straining the muscles of my rib cage and diaphragm to push air forcefully into the voice prosthesis. Unexpected coughing interfered with my speech, and if my prosthesis got obstructed by mucus, speech was impossible. Also challenging was the expression of emotion through my voice or altering vocal intensity. As mentioned above, the help and guidance of a devoted SLP was instrumental in regaining and improving my speech. Teamwork between my SLP and my otolaryngologist was invaluable in finding solutions to many issues. Unfortunately, collaboration between SLPs and otolaryngologists does not always happen, as many ear nose and throat specialists have minimal familiarity with and expertise in the care of voice prostheses and speech restoration. Mastering the ability to speak and especially to give a lecture was difficult. Unfortunately, I received no training on how to speak correctly following my laryngectomy. It was only six years later that a new SLP embarked on an intense four-month speech training program and taught me to speak properly and effortlessly. This was done with the assistance of an expiratory intra-tracheal air pressure manometer. I practiced how to speak slowly, producing minimal air pressure (almost a whisper), to over-articulating words, and learned to take a breath every four to five words. This was a difficult and slow process, since my manner of speaking before laryngectomy had been rapid and flowing. I learned to economize when speaking by delivering the essentials of each topic using fewer words. After mastering these techniques lecturing became much easier. In turn, my self-confidence increased and I no longer dreaded failing when giving lectures. EXPRESSING MYSELF IN WRITING The difficulties in expressing myself through speaking drove me to convey my ideas in writing. I decided to share my personal journey by writing a book6 that describes what happened to me in the three years following my cancer diagnosis. I shared how I faced and dealt with various treatments and how I adjusted to the new reality of life after surgery. I described my worries, doubts, concerns, fears, frustrations, disappointments, and ultimate acceptance of my handicaps, and the adjustment to living with the permanent question as to what will come next. I wrote motivated by the hope that medical professionals would better understand the difficulties and uphill battle a patient with cancer faces, and the critical role they can play in providing their patients with the best care possible. I also created a website7 and wrote a guidebook8 offering useful medical, dental, and psychological information to assist head and neck patients, including laryngectomees. Therein I discuss the side effects of radiation treatment and chemotherapy; speaking methods post-laryngectomy; airway management; stoma and voice prosthesis care; eating and swallowing issues; medical, dental, and psychological challenges; respiratory emergency care; pain relief; and help with traveling. Writing these books helped me turn a trying and challenging period of my life into something helpful for others. I chose to change this lemon into lemonade! PERSONAL REFLECTION AND OUTCOME I know that I underwent the optimal procedure to remove the primary tumor although it meant losing my voice and suffering the lifelong side effects of radiation therapy. I was, and remain, grateful for the additional years I gained through my physicians’ care, and the many joyful new moments experienced such as helping other people, seeing my grandchildren grow, and traveling to new places. Resumption of my old life occurred one step at a time: I began reading professional literature again; I had to learn to shower without getting water in my lungs; I began walking outside in my neighborhood; eventually I resumed bicycling and climbing mountains. One of the most gratifying moments came when I was able to deliver a lecture again. Even though I was very dependent on the microphone because of my weak voice, it was an important milestone. I had to accept the reality that I could not deliver a lecture as I had before, nor speak as clearly and forcefully as presenters with intact vocal cords. However, I have learnt that even with my poor voice I can still address important issues that can impact patient care and generate greater compassion for patients with serious illnesses. Through lecturing I found new meaning and purpose to my life that helps myself as well as others. I am invigorated by turning the loss of my vocal cords into a means for doing good. By sharing my personal experiences with patients and health professionals, others can benefit. In this arena, my imperfect voice has become an advantage rather than a handicap—it sends a powerful message to the audience. Experiencing serious illness as a patient has significantly influenced how I relate to my patients. It has made me more caring, empathetic, and sensitive. I better appreciate what patients and their families endure. I do my best to help them know they are not alone, to provide them with more attention and support, and to be sensitive and understanding of the magnitude of their loss. I strive to follow the example of the health care providers who helped me the most. My personal experiences have taught me the value of caring and supportive attitudes from medical professionals. I aim towards setting the right example for the medical students and residents whom I train. CONCLUSIONS The physical and emotional hardships and trials following laryngectomy render the patient helpless and dependent. Receiving skillful, caring, and compassionate care is invaluable during that period. Patient care can be improved by preparing the patients and their families for the medical, psychological, and social repercussions of the treatments received; by paying greater attention to the patients and spending more time with them; by keeping them updated and cognizant about their medical condition; by applying proper procedures in nursing care; by preventing medical mistakes through greater vigilance; and by training physicians and nurses how to respond appropriately to the unique post-surgical needs of their patients. Supplementary Information Conflict of interest: No potential conflict of interest relevant to this article was reported. Professor Brook kindly wrote this paper following his Grand Rounds Lecture at Rambam Health Care Campus where he personally shared from his experience. For more information about the author, refer to the separate supplementary material. Abbreviations SLPspeech and language pathologist ==== Refs REFERENCES 1 Brook I Neck cancer: a physician’s personal experience Arch Otolaryngol Head Neck Surg 2009 135 118 10.1001/archoto.2008.529 19221236 2 Dunbar RIM The social role of touch in humans and primates: behavioural function and neurobiological mechanisms Neurosci Biobehav Rev 2010 34 260 8 10.1016/j.neubiorev.2008.07.001 18662717 3 Tabatabaee A Tafreshi MZ Rassouli M Aledavood SA AlaviMajd H Farahmand SK Effect of therapeutic touch in patients with cancer: a literature review Med Arch 2016 70 142 7 27194823 4 Liebman CB Hyman CS A mediation skills model to manage disclosure of errors and adverse events to patients Health Aff (Millwood) 2004 23 22 32 10.1377/hlthaff.23.4.22 5 Brook I Rediscovering my voice JAMA 2009 302 981 10.1001/jama.2009.981 6 Brook I My Voice: A Physician’s Personal Experience with Throat Cancer Charleston, SC CreateSpace Publication 2009 1-4392-6386-8 Available at: http://bit.ly/29Fk2wl 7 Brook I My voice [blog] Available at http://bit.ly/29B5UCA accessed on June 26, 2016 8 Brook I The Laryngectomee Guide Charleston, SC CreateSpace Publication 2013 13: 9781483926940 Available at: http://bit.ly/29MUmwt

PMC005xxxxxx/PMC5001798.txt

==== Front Rambam Maimonides Med JRambam Maimonides Med JRMMJRambam Maimonides Medical Journal2076-9172Rambam Health Care Campus 10.5041/RMMJ.10253rmmj-7-3-e0026History of MedicineRemembering More Jewish Physicians Weisz George M. M.D., F.R.A.C.S. (Ortho), M.A.12*Grzybowski Andrzej M.D., Ph.D.345 1 School of Humanities (Program in History of Medicine), University of New South Wales, Sydney, Australia 2 School of Humanities, University of New England, Armidale, New South Wales, Australia 3 Department of Ophthalmology, Poznan City Hospital, Poznań, Poland 4 Department of Ophthalmology, University of Warmia and Mazury, Olsztyn, Poland 5 Editor-in-chief, Archives of the History and Philosophy of Medicine, Poznań, Poland* To whom correspondence should be addressed. E-mail:gmweisz1@aol.com7 2016 28 7 2016 7 3 e0026Copyright: © 2016 Weisz and Grzybowski.2016This is an open-access article. All its content, except where otherwise noted, is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.The history of medicine has been an intriguing topic for both authors. The modern relevance of past discoveries led both authors to take a closer look at the lives and contributions of persecuted physicians. The Jewish physicians who died in the Holocaust stand out as a stark example of those who merit being remembered. Many made important contributions to medicine which remain relevant to this day. Hence, this paper reviews the lives and important contributions of two persecuted Jewish physicians: Arthur Kessler (1903–2000) and Bronislawa Fejgin (1883–1943). BacteriophageHolocaustlathyrismmedical heroism ==== Body INTRODUCTION While in theory we understand the importance of history, in practicality there is a tendency to ignore it. This can lead to important discoveries being overlooked, minimized, or wrongly credited. This is particularly likely to happen when little is known about the person who did the original research. However, the rich tradition of Western medicine, which so proudly dates back to the Hippocratic oath, does itself and the future a great disservice if the past is not remembered. The inspiration for this paper was Dr Israel Milejkowski, Chief Medical Officer of the Warsaw Ghetto. Before his untimely demise on the way to Treblinka in 1943, he said: And you, Jewish physicians, you deserve some words of recognition. What can I say to you, my companions in misfortune?1 Milejkowski’s “companions in misfortune” surely merit being remembered. Indeed, a number of reviews of persecuted physicians have been published, mostly in Israeli journals.1–8 In light of modern advances related to lathyrism in starving populations, and the potential use of biophages for antibiotic-resistant bacteria, two such physician-researchers and their work merit our attention. This paper reviews the lives and work of two physician-researchers whose contributions to the above-mentioned fields have either been underestimated or ignored—Dr Arthur Kessler and Dr Bronislawa Fejgin. Dr Kessler’s impact on medicine came about because of the horrors of the Holocaust, whereas Dr Fejgin’s impact has been nigh forgotten for the same reasons. Nevertheless, their contributions to medical research are of great relevance today and will remain as their personal legacy to modern medicine. ARTHUR KESSLER (1903–2000) Born in Czernowitz in 1903, Arthur Kessler (Figure 1) only became a Romanian citizen after World War I. With a doctorate in medicine from Vienna, Kessler was recruited into the Romanian army. He was transferred to reserve duties only in 1930 and allowed to practice privately. With the Nazi invasion of Romania, his activities were progressively curtailed, and in 1942 Kessler was deported to the province of Transnistria, a narrow strip of land situated between Moldova in the west and the Ukraine in the east, between the rivers Bug and Dniester (Figure 2).9 Figure 1 Portrait of Dr Arthur Kessler. Source: From the private collection of D. Kessler, with permission. Figure 2 Map of Eastern Europe Showing the Location of Transnistria (red). Vapniarka and Grass Pea This region was uncultivated at the time and prone to harsh winters. The passing German forces found large quantities of grass pea horse-fodder left behind by the retreating Soviets. Possibly as a deliberate experiment10 they suggested feeding grass pea to the prisoners in Vapniarka. The Nazis would not have been unaware of the toxicity of grass pea (Lathyrus sativus), which is now known to contain a toxic alanine product known as oxalyldiaminopropionic acid (ODAP). The leadership of Vapniarka was Romanian; hence their willingness to feed their prisoners with bread baked from barley and chopped straw and grass peas boiled in salty water is noteworthy, since without doubt they also knew the grass pea was toxic—grass pea was not included in the diets of the camp personnel at all.11 Consumed in large quantity and over a prolonged period of time, this “nutrient” had been known to be neurotoxic since antiquity. In Europe, consuming grass pea was forbidden by George, Duke of Württemberg, in 1671, an order re-enforced by his successor, Leopold.12 The consumption of grass pea was also forbidden in France and Algeria in the early nineteenth century. This type of food-poisoning all but disappeared in Europe, except during the civil war in Spain and in the Balkans during WWII, but remained endemic in underdeveloped East African and Far East Asian countries. Nevertheless, historically, famine-afflicted populations relied on the grass pea when it was the only source available for nutrition. Detection of Grass Pea Poisoning Kessler ended up in the Vapniarka detention center where he served as a physician to the other incarcerated prisoners. Shortly after his arrival he recorded: … within months, hundreds of young male inmates of the camp began limping and had begun to use stick-crutches to propel themselves about, in some cases inmates had been rapidly reduced to crawling on their back sides to make their ways through the compound.9,13 Kessler collected the data of hundreds of patients who developed spastic paralysis of the lower limbs while in Vapniarka (Figure 3). He connected their condition with a never-seen and non-existent disease in Europe, not taught in medical schools and not appearing in textbooks. The symptoms involved a gradual paralysis ascending from the lower limbs to the thoracic level of the cortico-spinal tract, later on histologically proven to be a spinal cord degenerative process.11 This upper neuron disease, which would later be called lathyrism, appeared clinically as monoplegia, paraparesis or paraplegia, and incontinence; he meticulously recorded his findings on any available piece of paper.15 Kessler eventually determined that the toxic dosage of grass pea was 300 mg/day over a 3-month period (Figure 4). Kessler wrote: “We are eating poison and we will die of it.”15 Figure 3 Dr Kessler’s Drawing of a Paretic Limb, Vapniarka, 1942. Source: D. Kessler,14 with permission. Figure 3 Dr Kessler’s Drawing of the Lathyrus Plant, Vapniarka, 1942. Source: D. Kessler,14 with permission. Armed with the documented poisoning, the inmates organized a hunger strike.16 When the group confronted one of the brutal commandants of the camp, Kessler wrote that he responded, “what makes you think that we are interested in keeping you alive?”16 Kessler courageously persisted in confronting successive visiting authorities, some interested Romanian physicians, the Governor of the province, and a Ukrainian neurologist familiar with the disease. As a result, the toxic feeding program was replaced with a meagre food supply (two slices of bread a day, dried “fruit,” and artificially sweetened tea).17,18 Additional Medical Challenges in Vapniarka Despite the success in ending this toxic diet, there was no lack of patients. Starvation and the freezing Ukrainian winter led to aggravated gangrene, necessitating amputations and prostheses. Kessler also had to treat patients severely debilitated by tuberculosis, the typhus exanthematicus that flourished in the summer, despair-induced depression, and those who suffered the irreversible effects of the grass pea food-poisoning. A crutch support was improvised by the inventive physician, bandages for skin ulcerations and gangrene (angio-lathyrism) were manufactured, and various nappy supports were supplied to incontinent patients. Vascular occlusion occurred in adults, and the softened bone deformity (osteo-lathyrism) appeared in developing skeletons.11,13,15,17–20 Post-Vapniarka The camp was eventually dismantled, and Kessler was transferred to Olgopol, another camp in the province with even worse conditions. The hunger was more severe, the tuberculosis was more advanced, and typhus was more prevalent. When the war turned against the Nazis, the camp was closed, and the surviving prisoners were returned to Romania proper. There were hundreds of camps in that region, yet today almost nothing remains to testify of their existence, and residents of the region deny that Vapniarka ever existed.17 Kessler was fortunate to be amongst the surviving half of the 800,000 Jews who lived in pre-war Romania. His documents were hidden and published after liberation. After the war, Kessler immigrated to Israel and was employed by one of the health maintenance organizations there, practicing allergology. Kessler published numerous articles in local and international medical journals, focusing on lathyrism and pediatric allergology.9 His allergy-related research spanned a number of topics, including bee stings, allergies in infants, sensitivity to medications, bronchial asthma, and allergies to house dust, to mites, and to birds. He warned of the risk of aerosols and dealt with the influence of Israel’s climate on asthma patients, including publications on asthma mortality. Many of his publications remain relevant and continue to be cited today. Kessler’s Legacy Kessler’s research was foundational to the discovery of and subsequent research on the residual neural, vascular, and osteopathic symptoms and findings of lathyrism. Lathyrism would continue to be extensively investigated in Ichilov Hospital, Tel Aviv, particularly in the 200 survivors of Vapniarka.13,15,19,20 Dr Arthur Kessler died in Israel in 2000. However, his research and toxicology tables remain in use to this day. In one renowned case, knowledge of his research led to determining the “mysterious” cause of death of a college student in Alaska.21 Kessler would have been pleased to know that today a new detoxified grass pea that is alanine toxin-free and ODAP-free is being cultivated for use in famine-hit areas.22 BRONISLAWA FEJGIN Bronislawa Fejgin was a physician who was murdered during the Shoah. Despite her cut-short career, the legacy of her research into bacteriology and serology is of ongoing relevance.23 Her brief biography would tell the following short story: Fejgin was born in Warsaw in November 1883, a date that appears on her diploma issued by the Sorbonne Medical School in Paris in 1914, as well as her admission form to the Warsaw Ghetto on October 9, 1940 (Figure 5). There is almost no information about her family or her life prior to university. Figure 5 Dr Bronislawa Fejgin’s Warsaw Ghetto Registration Document. Source: Main Physicians Library, Warsaw, Poland (public domain). After graduation, Fejgin returned to Warsaw where she rapidly advanced and eventually headed the State Institute of Hygiene. Apart from local journals, her discoveries were published in French and English and even some German literature. In 1940, after being incarcerated in the Warsaw Ghetto, she became manager of the Bacteriological Institute there. Fejgin also taught bacteriology in the clandestine Medical School that the Czyste Jewish Hospital staff conducted in the ghetto under the disguise of a technical school for the prevention of infectious diseases. The school was never discovered and lasted some 10 months before mass deportation to Belzec.1 This exceptional and ingenious effort by the Jewish authorities in the ghetto involved organizing nighttime lectures for some 450 students deported to Warsaw from all over Poland. Preclinical lectures were offered, covering topics such as anatomy, histology, physiology, bacteriology, and clinical care.8 The lectures were taught under lantern or candle light, microscopes were smuggled in by sympathetic Polish (former) colleagues, and exams were recorded and final marks assessed. The hope of surviving and continuing medical practice in the Jewish community was the driving force of the teachers. The results of the exams were buried in metal containers and exhumed after the war. The surviving 50 students had their examinations recognized, and they were able to complete their medical studies.8 Scientific Contributions Fejgin’s scientific contributions impacted three different lines of bacteriology and serology, which at times overlapped. They are briefly reviewed below by topic rather than in chronological order. General bacteriology Fejgin’s doctoral thesis in Paris was on bacteriology and vaccinations for the inflamed uterus. She also studied diphtheria and isolated a batch of filterable anti-diphtheria fluid (lytic form) that could be produced in guinea-pigs or rabbits.23 This was the early form of what would later appear as the protective bacteriophage of Twort–D’Herelle.24,25 One of her important epidemiological discoveries related to the transmission of scarlet fever. Fejgin discovered that if fingers that had been infected with streptococcus by saliva were used to turn the pages of books the bacteria remained active on the pages for 4–6 weeks, rendering the pages infectious to the next reader.26 Typhus exanthematicus (spotted fever) The second line of interest is revealed through Fejgin’s numerous published results diagnosing typhus exanthematicus (spotted fever). She was the discoverer of cross-agglutination between Proteus HX 19 bacillus and the typhus “virus,” later identified as Rickettsia prowazeki. By reproducing a filterable form of HX 19 and injecting guinea-pigs intraperitoneally, a milder form of typhus fever was obtained, known in 1922–1925 as Nicolle disease.27,28 Hers was a basic study for future preparation of vaccines. Equally important was a diagnostic test that she developed: Fejgin found that the urine of typhus patients agglutinated a proteus antigen, a test useful for identifying typhus when no serum was available.28 Fejgin stated that in 1908 she established the presence of agglutinants to intestinal bacilli in the serum of typhus fever patients. This essential test was later utilized for serological diagnosis of typhus exanthematicus by Rudolf Weigl in Poland in the 1920s and 1930s and by Ludwik Fleck in the Lwów Ghetto.6 Phage phenomenon The third line of interest in Fejgin’s research related to the recently introduced phage phenomenon, described by Twort in 1915 and by D’Herelle in 1917. Their findings were actually based on Fejgin’s, namely that the phenomenon of bacterial autolysis resulted from an intracellular lytic agent, a “virus,” which was produced by the bacteria itself and would later become known as a bacteriophage.23 Her studies also complemented those of Twort and D’Herelle via her proposed preparation of the lytic agent for Proteus HX 19 and diphtheria.23–27 Fejgin expanded the bacteriophage studies on shigella, diphtheria, typhoid, and proteus, all with cultures, as well as agglutination and antibacterial lytic agent production that led to a new species of bacteria with diminished virulence. In 1922 she concluded that the lytic agent was produced by live bacteria and is lytic only to live bacteria.28 This proved to be an essential and basic concept for phage therapy. There has been a prolonged controversy in the literature regarding the true discoverer of the bacteriophage. Based on the sequence of Fejgin’s publications, she made a very early and essential contribution.29–31 By the end of the 1930s and the early 1940s bacteriophage research had markedly decreased, if not entirely been abandoned, partly because of the emergence of antibiotics, but perhaps as well due to Fejgin’s untimely demise in the Warsaw Ghetto in January of 1943.1,5 Fejgin’s Legacy Today, bacteriophage therapy is emerging as an essential in the fight against antibiotic-resistant bacteria. Despite dozens of publications, of which only a few are cited herein, her name is not to be found in a MEDLINE search, a fact that should be rectified. Clearly Fejgin was an anomaly in the male-dominated world of medical research; this too may be a factor in her contributions, for the most part, going unrecognized.32,33 CONCLUSION In conclusion, it seems appropriate to recall the statement of Dr Milejkowski: You too [Drs Kessler and Fejgin], were part of the whole, You too were menaced by forced labour, starvation, deportation, by all the forms of death that stalked our Ghetto. And you gave the murderers a bold answer with your work: I shall not die in vain.8 Conflict of interest: No potential conflict of interest relevant to this article was reported. Abbreviations ODAPoxalyldiaminopropionic acid ==== Refs REFERENCES 1 Weisz GM Remembering Jewish physicians Isr Med Assoc J 2015 17 203 5 26040043 2 Weisz GM Grzybowski A Medical discoveries in the ghettos, the anti-typhus battle Isr Med Assoc J 2011 13 261 5 21845963 3 Weisz GM Grzybowski A Albury WR Henryk, Thaddeusz and Ludwik Stabholz: three outstanding physicians from a Jewish medical family Korot 2012 21 363 72 4 Weisz GM Rembrandt’s Jewish Physician—Dr Ephraim Bueno (1599–1665): a brief medical history Rambam Maimonides Med J 2013 4 e0010 10.5041/RMMJ.10110 23908860 5 Weisz GM Dr Otto Heinrich Warburg Rambam Maimonides Med J 2015 6 e0008 10.5041/RMMJ.10183 25717390 6 Weisz GM Hitler’s Jewish physicians Rambam Maimonides Med J 2014 5 e0023 10.5041/RMMJ.10157 25120923 7 Weisz GM Dr. Fleck fighting Fleck typhus Soc Stud Sci 2010 40 145 53 10.1177/0306312709348569 20527327 8 Weisz GM Grzybowski A Albury WR The fate of the Warsaw ghetto medical faculty Isr Med Assoc J 2012 14 209 13 22675834 9 Kessler D Dr. Arthur Kessler (1903–2000) Lathyrus Lathyrism Newsletter 2003 3 5 7 10 Weindling P Victims and Survivors of Nazi Human Experiments: Science and Suffering in the Holocaust London, UK Bloomsbury Publishing 2014 11 Garfinkle J Anderman F Shevel MI Neurolathyrism in Vapniarka: medical heroism in a concentration camp Can J Neurol Sci 2011 38 389 44 10.1017/S0317167100012403 12 Enneking D Lathyrus database: Lathyrism <1970 Bibliography 2003 Available at: http://bit.ly/2905PYG Accessed on April 18, 2016 13 Westmore A Weisz GM Medical research undertaken in captivity War Soc 2009 28 89 112 10.1179/072924709791329180 14 Kessler D A Camp Physician (Ein Artzt Im Lager)—Dr. Arthur Kessler and epidemic of Lathyrism in the Vapniarca concentration camp (oral presentation) 16th Conference on Medicine and the Shoah May 17, 2016 Nahariya, Israel 15 Weintraub S Cohen DF Salama M Steifler M Weissman SL Skeletal findings in human neurolathyrism Eur Neurol 1980 19 121 7 7389756 16 Hirsch M Spitzer L There Was Never a Camp Here! Ghosts of Home: The Afterlife of Czernowitz in Jewish Memory Berkeley and Los Angeles, CA University of California 2010 197 231 17 Kuhn A McAllister KE There Was Never a Camp Here: Searching for Vapniarka Kuhn A McAllister KE Locating Memory: Photographic Acts New York, NY Berghahn Books 2006 135 54 18 Enneking D Ein Arzt im Lager—A camp physician. Excerpts of Arthur Kessler’s memoirs relating to the epidemic of neurolathyrism in camp Vapniarka, Transnistria CCDN News 2015 25 3 10 Available at: http://bit.ly/28XaSal Accessed June 6, 2016 19 Cohn DG Steifler M Human neurolathyrism. A follow up study of 200 patients Schweiz Arch Neurol Neurochir Psychiatr 1981 128 151 6 7244573 20 Radovici A An outbreak of lathyrism Bull Acad Med Roum 1945 17 77 86 21 Krakauer J How Chris McCandless died The New Yorker 9 12 2013 Available at: http://bit.ly/28ZF2Kv Last Accessed April 19, 2016 22 Lambein F Lathyrus sativus L. (grass pea) toxic plant or nutraceutical? CCDN News 2015 25 1 2 Available at: http://bit.ly/29ja5lN Accessed April 30, 2016 23 Falstein L Martyrdom of Jewish Physicians in Poland New York, NY Exposition Press 1963 24 Fejgin B Sur le principe lythique anti-diphteriques Compt Rend Soc Biol 1925 93 365 25 Fejgin B Sur la forme filtrante du bacile d’ Eberth Compt Rend Soc Biol 1925 92 1528 26 Fejgin B Sur la nature du phenomene “D’Herelle” Compt Rend Soc Biol 1923 89 1383 27 Fejgin B La toxicite de bacille de Shiga bacteriophage Compt Rend Soc Biol 1936 89 1385 28 Brokman H Fejgin B Hirszfeld H Meyzner M Przesmycki F Caronia G Contribution a l’etude de l’etiologie de la scarlatine Compt Rend Soc Biol 1925 6 944 29 Fejgin B Examen de quelque souches isolee des cobayee inocules avec le bacteriophage anti-HZ19 Compt Rend Soc Biol 1924 90 1202 30 Fejgin B Recherches sur le typhus exanthematique experimental Compt Rend Soc Biol 1924 90 1200 31 Kutter E DeVos D Gvasalia G Phage therapy in clinical practice: treatment of human infections Curr Pharm Biotechnol 2010 11 69 86 10.2174/138920110790725401 20214609 32 Abedon ST Kuhl SJ Blasdel BG Kutter EM Phage treatment of human infections Bacteriophage 2011 1 66 85 10.4161/bact.1.2.15845 22334863 33 Duckworth DH Who discovered the bacteriophage Bacteriol Rev 1976 40 783 802

PMC005xxxxxx/PMC5001804.txt

==== Front Rev Saude PublicaRev Saude PublicarspRevista de Saúde Pública0034-89101518-8787Faculdade de Saúde Pública da Universidade de São Paulo 0024010.1590/S1518-8787.2016050006310Original ArticlesEstimation of HIV incidence in two Brazilian municipalities, 2013 Estimação da incidência de HIV em dois municípios brasileiros, 2013 Szwarcwald Célia Landmann I Ferreira Orlando da Costa Júnior II de Brito Ana Maria III Luhm Karin Regina IV Ribeiro Clea Elisa Lopes V Silva Ana Maria VI Cavalcanti Ana Maria Salustiano VII Ito Tomoko Sasazawa VIII Raboni Sonia Mara IX de Souza Paulo Roberto Borges Júnior I Pereira Gerson Fernando Mendes X I Laboratório de Informações em Saúde. Instituto de Comunicação e Informação Científica e Tecnológica em Saúde. Fundação Oswaldo Cruz. Rio de Janeiro, RJ, BrasilII Departamento de Genética. Universidade Federal do Rio de Janeiro. Rio de Janeiro, RJ, BrasilIII Departamento de Saúde Coletiva. Centro de Pesquisas Aggeu Magalhães. Fundação Oswaldo Cruz. Recife, PE, BrasilIV Departamento de Saúde Comunitária. Setor de Ciências da Saúde. Universidade Federal do Paraná. Curitiba, PR, BrasilV Secretaria Municipal da Saúde de Curitiba. Curitiba, PR, BrasilVI Laboratório Municipal de Saúde Pública. Secretaria Municipal de Saúde de Recife. Recife, PE, BrasilVII Laboratório Central de Saúde Pública. Secretaria Estadual de Saúde de Recife. Recife, PE, BrasilVIII Prefeitura Municipal de Curitiba. Curitiba, PR, BrasilIX Universidade Federal do Paraná. Curitiba, PR, BrasilX Departamento de Doenças Sexualmente Transmissíveis, Aids e Hepatites Virais. Ministério da Saúde. Distrito Federal, BrasilCorrespondence: Célia Landmann Szwarcwald. Instituto de Comunicação e Informação Científica e Tecnológica em Saúde – Fiocruz. Av. Brasil, 4365 sala 225 Manguinhos 21040-360 Rio de Janeiro, RJ, Brasil. E-mail: celia.szwarcwald@icict.fiocruz.br Authors’ Contribution: Conception of the article: CLS, OCFJ, PRBSJ, GFMP. Preparation of the text: CLS, OCFJ, AMB, KRL, SMR, PRBSJ, GFMP. Data analysis: CLS, OCFJ, AMS, AMSC, TSI, SMR, PRBSJ. Results discussion: AMB, KRL, CELR. Conducted laboratory tests: MAS, AMSC, TSI, SMR. Coordinated the research in Curitiba: KRL, CELR. Coordinated the research in Recife: AMB. Conflict of Interest: The authors declare no conflict of interest. 26 8 2016 2016 50 556 4 2015 1 9 2015 This is an Open Access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. ABSTRACT OBJECTIVE To estimate HIV incidence in two Brazilian municipalities, Recife and Curitiba, in the year of 2013. METHODS The method for estimating incidence was based on primary information, resulting from the Lag-Avidity laboratory test for detection of recent HIV infections, applied in a sample of the cases diagnosed in the two cities in 2013. For the estimation of the HIV incidence for the total population of the cities, the recent infections detected in the research were annualized and weighted by the inverse of the probability of HIV testing in 2013 among the infected and not diagnosed cases. After estimating HIV incidence for the total population, the incidence rates were estimated by sex, age group, and exposure category. RESULTS In Recife, 902 individuals aged 13 years and older were diagnosed with HIV infection. From these, 528 were included in the study, and the estimated proportion of recent infections was 13.1%. In Curitiba, 1,013 people aged 13 years and older were diagnosed, 497 participated in the study, and the proportion of recent infections was 10.5%. In Recife, the estimated incidence rate was 53.1/100,000 inhabitants of 13 years and older, while in Curitiba, it was 41.1/100,000, with male-to-female ratio of 3.5 and 2.4, respectively. We observed high rates of HIV incidence among men who have sex with men, of 1.47% in Recife and 0.92% in Curitiba. CONCLUSIONS The results obtained in the two cities showed that the group of men who have sex with men are disproportionately subject to a greater risk of new infections, and indicate that strategies to control the spread of the epidemic in this population subgroup are essential and urgent. RESUMO OBJETIVO Estimar a incidência de HIV em dois municípios brasileiros, Recife e Curitiba, no ano de 2013. MÉTODOS O método de estimação da incidência foi baseado em informações primárias, resultantes do ensaio laboratorial Lag-Avidity para detecção de infecções recentes do HIV, aplicado em uma amostra dos casos diagnosticados nas duas cidades em 2013. Para a estimação da incidência de HIV para a população total das cidades, as infecções recentes detectadas na pesquisa foram anualizadas e ponderadas pelo inverso da probabilidade de teste de HIV no ano de 2013 entre os casos infectados e não diagnosticados. Após a estimação da incidência de HIV para a população total, foram estimadas as taxas de incidência por sexo, faixa de idade e categoria de exposição. RESULTADOS Em Recife, foram diagnosticados 902 indivíduos de 13 anos e mais com infecção de HIV. Desses, 528 foram incluídos no estudo, e a proporção estimada de infecções recentes foi de 13,1%. Em Curitiba, foram diagnosticadas 1.013 pessoas de 13 anos e mais, 497 participaram do estudo, e a proporção de infecções recentes foi de 10,5%. Em Recife, a taxa de incidência estimada foi de 53,1 por 100 mil habitantes de 13 anos e mais, enquanto em Curitiba, de 41,1 por 100 mil, com razão do sexo masculino para o feminino de 3,5 e 2,4, respectivamente. Foram evidenciadas elevadas taxas de incidência de HIV entre homens que fazem sexo com homens, de 1,47% em Recife e 0,92% em Curitiba. CONCLUSÕES Os resultados obtidos nas duas cidades mostraram que o grupo dos homens que fazem sexo com homens está desproporcionalmente sujeito ao maior risco de novas infecções, e indicam que estratégias para controle da disseminação da epidemia nesse subgrupo populacional são essenciais e urgentes. HIV Infections, diagnosisAIDS Serodiagnosis, methodsRisk GroupsAcquired Immunodeficiency Syndrome, epidemiology ==== Body INTRODUCTION Epidemiological surveillance information has been considered essential to evaluate the actions of control of the HIV/AIDS epidemic and to subsidize the planning of intervention strategies. In the last two decades, many countries have adopted HIV prevalence as an important indicator of second generation surveillance, which associates the HIV test result with risk behavior 1 . However, prevalence studies provide the proportion of people infected with HIV, including those with recent infection and with long-term infection. With the expansion of the antiretroviral treatment and, consequently, the survival of individuals infected by HIV 7 , the interpretation of prevalence becomes increasingly difficult, being essential to rely on estimates of HIV incidence for the delineation of the current scenario of the epidemic. The incidence indicates the degree in which the virus transmission is occurring and the groups under the highest risk of HIV transmission, and allows to identify the emergence of new sub-epidemics of HIV in the general population. Because of its recognized importance, HIV incidence estimates are being incorporated, increasingly, to the surveillance activities in many countries 2 , 8 , 20 , 22 . Recently, laboratory tests have been developed to estimate the HIV incidence in cross-sectional studies. The algorithms based on laboratory tests make it possible to identify whether the infection is recent or long-term 9 . The main advantage of this type of study is using only one blood sample, collected at a point of time, such as in studies to estimate prevalence, eliminating the follow up of individuals 19 . In the 2000s, the BED-CEIA test (Calypte HIV-1 BED Incidence Capture EIA) was developed, which is an immunoenzymatic analysis that allows to distinguish cases with recent infection from cases with long-term infection by HIV 21 . The BED-CEIA test was used to estimate the HIV incidence in several countries and different epidemiologic scenarios 5 , 8 , 16 , including between the reported cases of HIV to estimate the HIV incidence in the United States of America (USA) 10 . However, validation studies have shown, consistently, disparate estimates of HIV incidence when using calculations based on the result of the BED-CEIA test, depending on the adopted parameters and method used for the estimate 17 . Additionally, studies show false-recent results of BED between people with very low CD4 count (< 50/μl) or receiving antiretroviral therapy (ART) 15 . Another method that allows to distinguish recent infections from long-term infections is based on the IgG avidity test for HIV-1 antigens, first described in 2002. Our methodology is based on the principle that the avidity of the antibodies produced in the early stage of infection is reduced, in opposition to what is observed in long-term infections 24 . Among the advantages of the avidity test, we highlight: the high sensitivity and specificity in detecting recent infections; the simplicity and automatism of the technique; the good performance of the test regardless of HIV subtype; and the fact the avidity index is not affected by the use of ART 6 . This article aimed to estimate the incidence of HIV in two Brazilian municipalities, Recife and Curitiba, in 2013, by statistical estimation method based on the Lag-Avidity laboratory test results for detection of recent infections, applied to the cases diagnosed with HIV over that year in the two cities. METHODS The employed incidence estimation method used primary information, resulting from laboratory tests to detect recent HIV infections applied in Recife and Curitiba. The two municipalities were chosen for the research because they are located in the Northeast and South, respectively, which have different levels of socioeconomic development and different epidemiological situations regarding HIV infection 4 . The avidity test (SediaTM HIV-1LAg-Avidity EIA, Sedia Biosciences Corporation, Portland, Oregon, USA) for detection of recent HIV infection was applied in a sample of positive HIV tests, diagnosed in Recife and Curitiba in 2013. The study was done in partnership with the Centers for Disease Control and Prevention (CDC) and was approved by the Ethics Committee of the Oswaldo Cruz Foundation (Protocol 485,175) and by the CDC responsible body (Office of the Associate Director for Science of the Division of Global HIV/AIDS, US Centers for Disease Control and Prevention). CDC has licensed the sale of the SediaTM HIV-1LAg-Avidity EIA in the United States only for research use. Samples classified as recent in the SediaTM test have an average duration of seroconversion of 141 days (95%CI 119–150). In Brazil, the product is still not registered by Anvisa. The CDC gave the necessary tests for the conduction of the project and paid all the costs of imports. Two lots were used during the study (EK0201 and FA3001), both in Curitiba and Recife. Individuals who carry out HIV testing in the public sector look for two types of health units: the ones that just collect blood and the ones that collect the blood and do the HIV test. In the first group are the basic health units, which collect the blood and routinely send the samples to municipal laboratories for analysis. In the second are the Centros de Testagem e Aconselhamento (CTA – Testing and Counseling Centers), the “Get To Know” trailers, the family health units, which perform HIV rapid test, and the public hospitals. In health units that collect and perform the HIV test, it was necessary to collect venous blood among individuals with positive result to enable the application of the test of recent infection. The samples of HIV-infected individuals were properly labeled, with their identification, and sent to public laboratories previously selected. As in some health units the positive result of HIV is rare or occurs during non-business hours, as in the case of trailers, the blood samples were collected in gel separator tube, centrifuged, and stored in common refrigerator (from 4°C to 8°C). The dispatch to the reference laboratory was done within 24 hours. The blood samples of HIV-infected individuals were stored in freezer (-20°C) in the public laboratories for further application of the avidity test for detection of recent infections. To include tests conducted in the private sector, we considered the blood samples that had positive result for HIV in selected private laboratories in the two cities. The blood samples were properly transported and stored in the public laboratories previously selected, for application of the recent detection test, concomitantly with the samples from the public sector. Information Bank In each Municipal Secretariat of Health (SMS), we composed an information bank of individuals who performed the HIV test and had positive results, with the following variables: name, city of residence, sex, date of birth, HIV test date, and laboratory test result (recent or long-term infection). The information bank was related to other information systems of the municipal secretariats, such as the electronic records system in Curitiba and the Laboratory Tests Requisition System (SIREX) in Recife, and the Information System for testing and counseling centers, in both cities, to add sociodemographic and exposure category information, when available. The individuals were identified by full name, date of birth, and city of residence, to eliminate duplicates. These are found, for example, between individuals that make the HIV test several times a year, or between those who have received positive results for HIV and repeat the test. After the elimination of duplicates and composition of the information bank in the SMS, the individual’s identification variables were replaced with codes for the statistical analysis of the data. Data Analysis In the proposed model to estimate the HIV incidence in the USA 10 , they assumed that the number of new infections has a Poisson distribution with parameter I, while the number of recent infections detected by the laboratory test has a Poisson distribution with parameter R = IP, where P is the probability that a person infected with HIV is detected with recent infection by the biological marker used in a sample of cases diagnosed in a given year. The probability P can be written by the product: P = P 1 . P 2 . P 3, being: P 1 the probability of an individual infected with HIV performing his first test in a given year; P 2 the probability of a person diagnosed with HIV in that year having positive result of the recent infection test; and P 3 the probability of conduction of recent infection test among people diagnosed in a given year 10 . Thus, if r is the number of cases with recent infections detected in the sample of HIV-infected cases, after setting the number of recent infections in the window period for the annual number of recent infections by a mathematical correction factor, the estimator (i) of I, proposed by McWalter and Welte 17 (2010), is given by: in which, N = number of individuals infected with HIV diagnosed for the first time in 2013 (population); n = number of positive individuals diagnosed in 2013 for which the avidity test was conducted; r = number of recent infections detected in the sample; w = average duration to detect recent infection, expressed as a fraction of the year (equal to 141/365, because the avidity test window period is of 141 days); FRR = false recent rate for the test used (0.2%) 6 ; p 1 = estimator of the probability of a not yet diagnosed positive individual be tested for HIV in a given year. For the application of the equation (1), the probability p 1 was estimated at 0.44, as the proportion of people infected with HIV who performed HIV test in 2013 between the previously undiagnosed ones. The number of people infected with HIV who performed the HIV test in 2013 (64,070) was obtained by the information of the Control System of Laboratory Tests (SISCEL), and the number of people without previous diagnosis (145,000), by the cascade of continuous care of HIV, Brazil, 2013 4 . In each city, we compared the distribution of cases classified as recent and long-term by Chi-square test for homogeneity of distributions, according to sex, age group (13-24; 25-34; 35-49; 50+), and exposure category: women; men who have sex with men (MSM); men classified in other exposure categories (heterosexual; injectable drug user; blood; other; not informed). After the HIV incidence calculation to the total population, we estimated the incidence rates by sex, age group, and exposure category, in each city. To find the number of recent infections in each analysis category, we multiplied the proportion of recent infections in the specific category by the estimated number of people in each category in 2013. The populations by sex and age group were projected for the year 2013 from the information of the Demographic Census, 2010. To calculate the size of the group of men who have sex with men, we multiplied the male population of individuals of 13 years or more estimated for 2013 by 3.5%, estimated proportion of MSM in Brazil 25 . RESULTS In Table 1, we present the results of the laboratory test in Recife and Curitiba. In the first city, the proportion of recent infections was 13.1% and, in the second, 10.5%. Although the punctual value of the proportion of recent infections was higher in Recife, there was no statistically significant difference, with intersection of the confidence intervals. Table 1 Avidity test results. Recife and Curitiba, 2013. Result Recife Curitiba Number of individuals diagnosed (N) 902 1,013 Number of positive individuals for which the avidity test was conducted (n) 528 497 Number of recent infections detected by the avidity test (r) 69 52 Proportion of recent infections in the sample and 95%CI 13.1% (10.2–15.9) 10,5% (7.8–13.2) Yearly proportion of recent infections and 95%CI 33.8% (26.4–41.3) 27.2% (20.1–34.1) Yearly number of recent infections and 95%CI 300 (234–366) 269 (201–338) Estimated probability that an individual infected with HIV was tested in 2013 (p)a 44.0% 44.0% Estimated number of new cases of HIV and 95%CI 683 (532–833) 612 (456–767) Estimated population with 13 years or moreb 1,286,197 1,486,347 Incidence rate (per 100,000 individuals with 13 years or more) and 95%CI 53.1 (41.4–64.8) 41.1 (30.7–51.6) a Estimated by the cascade of continuous care of HIV, Brazil, 20134. b Designed by the demographic information disclosed by the DATASUS website. The application of the equation (1) to the results presented in Table 1 made it possible to estimate the number of new cases of HIV in 2013 in each of the cities. In Recife, the estimated number of new infections was 683, corresponding to an incidence rate of 53.1/100,000 inhabitants of 13 years or more, while in Curitiba, the estimated number of new infections was 612, with an incidence rate of 41.1/100,000 inhabitants of 13 years or more. Table 2 contains the proportions of recent infections by age group, sex, and exposure category. In both cities, the largest proportion of recent infections is found among young people, under 25 old. In Curitiba, there was a decreasing gradient with increasing age and the differences in the proportions were significant (p < 1.0%). In Recife, the youngest age range also stood out, but the differences were not statistically significant at the 5% level. Table 2 Proportion (%) of recent and long-term infections by sex, age group, and exposure category. Recife and Curitiba, 2013. Recife Recent (%) Long-term (%) pa Age group (years) 13-24 19.4 80.6 0.230 25-34 12.1 87.9 35-49 11.1 88,9 50+ 11.8 88.2 Sex F 8.9 91.1 0.073 M 14.5 85.5 Exposure category F 8.9 91.1 < 0.001b MSM 26.0 74.0 Men in other exposure categories 9.2 90,8 Curitiba       Age group (years) 13-24 18.8 81.2 0.001b 25-34 11.2 88.8 35-49 7.5 92.5 50+ 2.7 97.3 Sex F 11.3 88.7 0.313 M 9.7 90.3 Exposure category F 11.3 88.7 0.01b MSM 15.5 84.5 Men in other exposure categories 6.8 93.2 F: female; M: male; MSM: men who have sex with men a descriptive level of significance of the Chi-square test for homogeneity of distributions. b significant at 5% level. Regarding the comparison by sex, the proportions of recent infections were similar in the two cities (Table 2). Regarding the exposure category, in both cities, the proportions of recent infections were significantly higher among men who have sex with men (MSM), both when compared to women and when compared to other men classified in other exposure categories. The results displayed in Table 2 show that the differences by exposure category were significant at the 1% level, in the two studied cities. Table 3 shows the incidence rates (per 100,000 inhabitants) estimated by sex, age group, and exposure category. The highest rate was found in the group of 25 to 34 years old, and the lowest, between people with 50 years or more, in both cities. Concerning the analysis by sex, the sex ratio (3.4) in the city of Recife is more pronounced than in Curitiba (2.4), with higher risk among men. However, the results that most stand out in Table 3 are the high rates of HIV incidence in the MSM group. In Recife, the incidence rate of HIV exceeds 1.0%, while in Curitiba, it is close to 1.0% (0.92%). Additionally, we observe that the incidence rates of HIV for females are similar in both cities. The biggest differences are in the incidence rates for males, MSM group, especially in Recife. Table 3 Incidence rates of HIV (per 100,000 inhabitants) according to age group, sex, and exposure category. Recife and Curitiba, 2013.   Incidence rate of HIV (per 100,000 inhabitants) Recife Curitiba Age group (years) 13-24 58.7 60.2 25-34 78.9 74.1 35-49 57.9 33.9 50+ 22.7 4.4 Sex F 25.2 25.1 M 87.3 59.4 Exposure category F 25.2 25.1 MSM 1,469.0 923.7 Men in other exposure categories 37.2 28.1 Total (13 years or more) 53.1 41.2 F: female; M: male; MSM: men who have sex with men DISCUSSION In Brazil, the serological tests for detecting recent infection have been applied since the late 1990s 19 . However, laboratory tests have never been used in Brazil, previously, to estimate the HIV incidence in the general population. In this study, the avidity test has been applied in a sample of the HIV cases diagnosed in 2013, in Recife and Curitiba, to investigate the possibility of using laboratory tests to detect recent HIV infections, routinely in Brazil, allowing the monitoring of HIV incidence in the major cities of the Country. Our methodology was feasible, especially for HIV cases diagnosed in the public sector. Among the individuals who perform the HIV test in units that only collect blood, the samples are sent to the municipal laboratories for analysis, routinely, being possible to perform the test of recent infection for all positive cases. Among the units that collect and perform the HIV test, the planning of venous blood collection among the positive cases was necessary. In Recife, among the cases diagnosed with HIV, the venous collection is routinely performed for the syphilis test. However, in Curitiba, this procedure is no longer being adopted because of the use of rapid tests for syphilis, and it was necessary to use venous blood samples collected for the first CD4 T lymphocyte count. Although all venous blood samples have been scheduled in a maximum period of 15 days from the date of HIV diagnosis, there was a higher number of losses than in Recife and possible delay in the venous blood collection. Regarding the private sector, the logistics of deploying the laboratory test for detecting recent infection was more complicated. In private laboratories, not all blood samples are tested locally and the confirmatory HIV test is usually performed in another city. To consider individuals diagnosed with HIV in private laboratories, the laboratories that perform, locally, the ELISA tests, needed to store part of the venous blood sample of the positive individuals before the conduction of the confirmatory test. The blood samples stored in the private laboratory were transported to public laboratories for the confirmatory HIV testing and for detecting recent infection among the confirmed cases. The inclusion of private sector cases was made by agreement with selected laboratories in the two cities and only since May 2013. In addition, as we did not consider the positive samples of blood banks and hospitals with less than five positive cases per month, the avidity test has been applied in 58.5% of cases diagnosed in Recife and 49.0% in Curitiba. In countries with concentrated epidemics, the population-based studies by sampling for estimation of the proportion of recent infections are impractical, both by requiring very large sample sizes and by the difficulties of attracting individuals under greater risk to HIV 14 . These countries have chosen to estimate HIV incidence based on reported cases of HIV, considering all individuals infected with HIV diagnosed in a given year as population elements, while those detected as suffering recent infections as sample persons. The selection probabilities are estimated by the probabilities of individuals infected with HIV, not diagnosed earlier, be tested in a given year. Recent infections detected in the research are weighted by the inverse of the probabilities of selection (expansion factors) to produce the number of incident cases in the year 8 , 10 . In this study, for the application of the method in the two Brazilian cities, we used McWalter and Welte 17 calculation, the most recommended among the proposed mathematical formulas, because it is a simplified and updated version and for not presenting statistically significant differences when compared to the other estimates. In both cities, the annualized number of recent infections was weighted by the inverse of the proportion of HIV-infected cases in Brazil, which were tested in 2013 and had no previous diagnosis 4 . Because of the unavailability of specific estimates of this proportion for Recife and Curitiba, we used the weighting factor estimated for all Brazil in the two cities. This was the main limitation of the study, since large differences in the weighting factor may lead to relevant changes in the incidence estimates. The results were consistent with previous Brazilian studies. In Curitiba, the incidence rate of HIV is on the level of 40/100,000 individuals with 13 years or more, while in Recife, it exceeded 50/100,000. According to data of the latest Epidemiological Bulletin, Recife occupies the eighth place among Brazilian capitals because of the high detection and mortality rates, while Curitiba occupies the 16th position 4 . Both in Curitiba and in Recife, the highest prevalences of recent infection have occurred in young people, with the highest incidence rate in the group of 25 to 34 years old. In both cities, the incidence rates were higher among men, but the sex ratio was much more pronounced in Recife, with 3.4 new infections among men for every new infection among women. In a previous study carried out in Recife, the proportion of recent infections for males was 2.4 times greater than for females 11 , equal value to the one found for Curitiba in this study. The increase is probably explained by the increasing number of incident cases among men who have sex with men, which accounted for 43.5% of recent infections, in Recife, and 36.8%, in Curitiba. In fact, the results that most called the attention in the study were the high incidence rates of HIV in the MSM group. In Curitiba, the incidence rate among MSM is close to 1.0%, and, in Recife, it exceeded this value. It is noteworthy that all male cases without reported exposure category were classified as “male in other exposure category”, i.e., the incidence rate among MSM may be even greater than the estimated here. Despite earlier concerns about an increase in heterosexual cases and “feminization” of the epidemic, the findings of this study indicate a predominance of new infections among MSM, corroborating evidence of a resurgence of the epidemic in this population group in other countries 3 , 23 . A research conducted incidence 10 Brazilian cities, using Respondent Driven Sampling as sampling method, showed prevalence of HIV of 14.2% among MSM 11 , approximately 30 times greater than the prevalence of HIV in the heterosexual male population. In an investigation of the composition of recent infections by exposure category, held in testing centers in Rio de Janeiro in 2004-2005, the estimated incidence among MSM was 11 times greater than among heterosexual men 5 . Despite the evidence of increased risk among MSM, the periodic test coverage for HIV in this population subgroup is still low and insufficient to ensure the early detection and immediate treatment 13 . The already widely recognized benefits of early introduction of antiretroviral therapy 18 seem to be neglected by the stigma and fear of being positive 16 . In sum, the results obtained in the two cities showed that the MSM group is disproportionately subject to greater risk of new infections, and indicate that strategies to control the spread of the epidemic in this population subgroup are essential and urgent. Funding: Centers for Disease Control and Prevention (CDC). ==== Refs REFERENCES 1 Ades AE Serial HIV seroprevalence surveys: interpretation, design, and role of HIV/AIDS prediction J Acquir Immune Defic Syndr Hum Retrovirol 1995 9 5 490 499 10.1097/00042560-199509050-00006 7627625 2 Bätzing-Feigenbaum J Loschen S Gohlke-Micknis S Zimmermann R Herrmann A Kamga Wambo O et al Country-wide HIV incidence study complementing HIV surveillance in Germany Euro Surveill 2008 13 36 pii18971 3 Beyrer C Sullivan P Sanchez J Baral SD Collins C Wirtz AL et al The increase in global HIV epidemics in MSM AIDS 2013 27 17 2665 2678 10.1097/01.aids.0000432449.30239.fe 23842129 4 Boletim Epidemiológico HIV-Aids Brasília (DF) Ministério da Saúde, Departamento de DST, Aids e Hepatites 2014 3 1 3 80 5 Castro CA Grinsztejn B Veloso VG Bastos FI Pilotto JH Morgado MG Prevalence, estimated HIV-1 incidence and viral diversity among people seeking voluntary counseling and testing services in Rio de Janeiro, Brazil BMC Infect Dis 2010 10 224 10.1186/1471-2334-10-224 6 Duong YT Qiu M De AK Jackson K Dobbs T Kim AA et al Detection of Recent HIV-1 infection using a new limiting-antigen avidity assay: potential for HIV-1 incidence estimates and avidity maturation studies PLoS One 2012 7 3 55 10.1371/journal.pone.0033328 7 Guibu IA Barros MB Donalísio MR Tayra  Alves MC Survival of AIDS patients in the Southeast and South of Brazil: analysis of the 1998-1999 cohort Cad Saude Publica 2011 27 1 S79 S92 10.1590/S0102-311X2011001300009 21503528 8 Hall HI Song R Rhodes P Prejean J An Q Lee LM et al Estimation of HIV incidence in the United States JAMA 2008 300 520 529 10.1001/jama.300.5.520. 18677024 9 Janssen RS Satten GA Stramer SL Rawal BD O’Brien TR Weiblen BJ et al New testing strategy to detect early HIV-1 infection for use in incidence estimates and for clinical and prevention purposes JAMA 1998 280 1 42 48 10.1001/jama.280.1.42 9660362 10 Karon JM Song R Brookmeyer R Kaplan EH Hall HI Estimating HIV incidence in the United States from HIV/AIDS surveillance data and biomarker HIV test results Stat Med 2008 27 23 4617 4633 10.1002/sim.3144 18833636 11 Kerr LR Mota RS Kendall C AdeA Pinho Mello MB Guimarães MD et al HIV among MSM in a large middle-income country AIDS 2013 27 3 427 435 10.1097/QAD.0b013e32835ad504 23291540 12 Lima KO Salustiano DM Coêlho MR Cavalcanti AM Diaz RS Lacerda HR Incidence of recent human immunodeficiency virus infection at two voluntary counseling testing centers in Pernambuco, Brazil, from 2006 to 2009 J Clin Microbiol 2012 50 6 2145 2146 10.1128/JCM.05107-11 22403429 13 Lippman SA Veloso VG Buchbinder S Fernandes NM Terto V Sullivan PS et al Over-the-counter human immunodeficiency virus self-test kits: time to explore their use for men who have sex with men in Brazil Braz J Infect Dis 2014 18 3 239 244 10.1016/j.bjid.2014.02.002 24704388 14 Magnani R Sabin K Saidel T Heckathorn D Review of sampling hard-to-reach and hidden populations for HIV surveillance AIDS 2005 19 2 S67 S72 10.1097/01.aids.0000172879.20628.e1 15930843 15 Marinda ET Hargrove J Preiser W Slabbert H van Zyl G Levin J et al Significantly diminished long-term specificity of the BED capture enzyme immunoassay among patients with HIV-1 with very low CD4 counts and those on antiretroviral therapy J Acquir Immune Defic Syndr 2010 53 4 496 499 10.1097/QAI.0b013e3181b61938 20306555 16 Marsicano E Dray-Spira R Lert F Aubrière C Spire B Hamelin C et al Multiple discriminations experienced by people living with HIV in France: results from the ANRS-Vespa2 study AIDS Care 2014 26 1 S97 106 10.1080/09540121.2014.907385 24738926 17 McWalter TA Welte A A comparison of biomarker based incidence estimators PLoS One 2009 4 10 55 10.1371/journal.pone.0007368 18 Montaner JS Lima VD Harrigan PR Lourenço L Yip B Nosyk B et al Expansion of HAART coverage is associated with sustained decreases in HIV/AIDS morbidity, mortality and HIV transmission: the “HIV Treatment as Prevention” experience in a Canadian setting PLoS One 2014 9 2 55 10.1371/journal.pone.0087872 19 Morgado MG Bastos FI Estimates of HIV-1 incidence based on serological methods: a brief methodological review Cad Saude Publica 2011 27 1 S7 18 10.1590/S0102-311X2011001300002 21503527 20 Ndawinz JD Costagliola D Supervie V New method for estimating HIV incidence and time from infection to diagnosis using HIV surveillance data: results for France AIDS 2011 25 15 1905 1913 10.1097/QAD.0b013e32834af619 21811147 21 Parekh BS Kennedy MS Dobbs T Pau CP Byers R Green T et al Quantitative detection of increasing HIV type 1 antibodies after seroconversion: a simple assay for detecting recent HIV infection and estimating incidence AIDS Res Hum Retroviruses 2002 18 4 295 307 10.1089/088922202753472874 11860677 22 Rehle T Shisana O Pillay V Zuma K Puren A Parker W National HIV incidence measures: new insights into the South African epidemic S Afr Med J 2007 97 3 194 199 17440667 23 Sighem AV Vidondo B Glass TR Bucher HC Vernazza P Gebhardt M et al Resurgence of HIV infection among men who have sex with men in Switzerland: mathematical modelling study PLoS One 2012 7 9 55 10.1371/journal.pone.0044819 24 Suligoi B Galli C Massi M Di Sora F Sciandra M Pezzotti P et al Precision and accuracy of a procedure for detecting recent human immunodeficiency virus infections by calculating the antibody avidity index by an automated immunoassay-based method J Clin Microbiol 2002 40 11 4015 4020 10.1128/JCM.40.11.4015-4020.2002 12409368 25 Szwarcwald CL Barbosa-Júnior A Pascom AR Souza-Júnior PR Knowledge, practices and behaviours related to HIV transmission among the Brazilian population in the 15-54 years age group, 2004 AIDS 2005 19 4 S51 S58 10.1097/01.aids.0000191491.66736.16 16249655

PMC005xxxxxx/PMC5001805.txt

==== Front Rev Saude PublicaRev Saude PublicarspRevista de Saúde Pública0034-89101518-8787Faculdade de Saúde Pública da Universidade de São Paulo 0023910.1590/S1518-8787.2016050006183Original ArticlesEvaluation of Reference Centers for Special Immunobiologicals implementation Avaliação da implantação dos Centros de Referência para Imunobiológicos Especiais Nóbrega Laura Andrade Lagôa I Novaes Hillegonda Maria Dutilh II Sartori Ana Marli Christovam I I Departamento de Moléstias Infecciosas e Parasitárias. Faculdade de Medicina. Universidade de São Paulo. São Paulo, SP, BrasilII Departamento de Medicina Preventiva. Faculdade Medicina. Universidade de São Paulo. São Paulo, SP, BrasilCorrespondence: Laura Andrade Lagôa Nóbrega. Av. Dr. Enéas de Carvalho Aguiar, 470 1º andar sala 102 Cerqueira César. 05403-000 São Paulo, SP, Brasil. E-mail: lagoalaura@yahoo.com.br Authors’ Contribution: All authors participated in the study design. Preparation of the data collection instrument: LALN, AMCS. Conduction of data collection: LALN. Organization and analysis of the database: LALN. All authors participated in the interpretation of results and in the drafting and critical review of the manuscript. Conflict of Interest: The authors declare no conflict of interest. 26 8 2016 2016 50 5810 2 2015 25 10 2015 This is an Open Access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. ABSTRACT OBJECTIVE To describe the Reference Centers for Special Immunobiologicals and evaluate their implementation considering formal regulations. METHODS We conducted a program evaluation, of evaluative research type. From August 2011 to January 2012, a questionnaire was applied to the 42 Reference Centers for Special Immunobiologicals existing in the Country, approaching the structure, human resources, and developed activities dimensions. We conducted a descriptive analysis of data and used a clustering for binary data with the squared Euclidean distance, by the farthest neighbor method, to aggregate services with similar features. RESULTS We observed great diversity among the services in the three dimensions. The clustering resulted in five service profiles, named according to their characteristics. 1) Best structure: 12 Reference Centers for Special Immunobiologicals with the highest proportion of services with the minimum of rooms recommended, purpose-built vaccine refrigerators, preventive maintenance of the cold chain, and oxygen source. 2) Immunobiologicals distributor: six Reference Centers for Special Immunobiologicals that distributed more than applied immunogens; no doctor present for more than half of the working hours and no purpose-built vaccine refrigerators . 3) Incipient implementation: five Reference Centers for Special Immunobiologicals with inadequate structure, such as absence of purpose-built vaccine refrigerators, preventive maintenance of the cold chain and oxygen source; none had computer. 4) Vaccination rooms: 13 Reference Centers for Special Immunobiologicals, everyone did routine immunization, most participated in vaccination campaigns. 5) Teaching and research: six services, all inserted into teaching hospitals, developed researches and received trainees; most had doctors in more than half of the working hours. CONCLUSIONS The evaluation of the Reference Centers for Special Immunobiologicals implementation was based on the profiles found and considered the official regulations: services categorized as “better structure” and “teaching and research” were considered implemented; “immunobiologicals distributor” and “vaccination room” services, partially implemented, and the ones with the “incipient implementation” profile, not implemented. The results of this evaluation can contribute to the reformulation of the services, considering the current context. RESUMO OBJETIVO Descrever os Centros de Referência para Imunobiológicos Especiais e avaliar sua implantação considerando as regulamentações formais. MÉTODOS Foi realizada uma avaliação de programa, tipo pesquisa avaliativa. De agosto de 2011 a janeiro de 2012, foi aplicado questionário aos responsáveis pelos 42 Centros de Referência para Imunobiológicos Especiais existentes no País, abordando as dimensões estrutura, recursos humanos e atividades desenvolvidas. Foi feita análise descritiva dos dados e utilizado agrupamento para dados binários com uso da distância euclidiana quadrática, pelo método do vizinho mais distante, para agregar serviços com características semelhantes. RESULTADOS Observou-se grande diversidade entre os serviços nas três dimensões. O agrupamento resultou em cinco perfis de serviços, denominados de acordo com suas características. 1) Melhor estrutura: 12 Centros de Referência para Imunobiológicos Especiais com a maior proporção de serviços com o mínimo de salas preconizado, câmaras de vacinas, manutenção preventiva da rede de frio e fonte de oxigênio. 2) Dispensador de imunobiológicos: seis Centros de Referência para Imunobiológicos Especiais que mais dispensavam do que aplicavam imunógenos; sem médico presente por mais da metade do expediente do serviço e sem câmara de vacinas. 3) Implantação incipiente: cinco Centros de Referência para Imunobiológicos Especiais com estrutura inadequada, como ausência de câmaras de vacinas, de manutenção preventiva da rede de frio e de fonte de oxigênio; nenhum possuía computador. 4) Sala de vacinas: 13 Centros de Referência para Imunobiológicos Especiais, todos faziam imunização de rotina, a maioria participava de campanhas de vacinação. 5) Ensino e pesquisa: seis serviços, todos inseridos em hospitais de ensino, desenvolviam pesquisas e recebiam estagiários; a maioria possuía médicos em mais da metade do expediente. CONCLUSÕES A avaliação de implantação dos Centros de Referência para Imunobiológicos Especiais baseou-se nos perfis encontrados e considerou as regulamentações oficiais: os serviços categorizados como “melhor estrutura” e “ensino e pesquisa” foram considerados implantados; os serviços “dispensador de imunobiológicos” e “sala de vacinas”, parcialmente implantados e os do perfil “implantação incipiente”, não implantados. Os resultados dessa avaliação podem contribuir para a reformulação dos serviços, considerando o contexto atual. ImmunizationHealth CentersImmunization Programs, organization & administrationVaccines, supply & distributionProgram EvaluationHealth Services Evaluation ==== Body INTRODUCTION Patients with chronic diseases, such as immune deficiencies (congenital or acquired), neurological, hematological, and metabolic disorders, heart diseases, lung diseases, and others, or with exposure to risk situations, have a higher risk of infection or severe illness by certain pathogens and have recommendations for specific immunizations 15 , 24 , 25 . Several countries, such as United Kingdom, France, Germany, United States, Mexico, and Argentina, have established vaccination calendars for these individuals 1 , 2 , 4 , 7 , 8 , 13 . To meet these special groups in Brazil, the Programa Nacional de Imunizações (PNI – National Immunization Program) created the Reference Centers for Special Immunobiologicals (CRIE), which are public and free vaccination units. These Centers provide vaccines and immunoglobulins not available on the PNI routine for individuals with a higher risk of infection or severe illness, and for those with contraindication of immunobiologicals used routinely. Besides, the CRIE are also responsible for the investigation and follow-up of cases of adverse events following immunization (AEFI) 15 . The first services were created in 1993. Until 2000, 34 CRIE were created and, since 2002, each Brazilian state has at least one of these services 16 . Several publications discuss indications of special immunobiologicals or describe the care to specific groups in the CRIE 3 , 5 , 6 , 21 , 24 , 26 , but we did not find studies that evaluate the implementation of CRIE in national perspective. The aim of this study was to describe the CRIE existing in Brazil, in 2011, and evaluate if their implementation occurred in accordance with the guidelines and regulations established by PNI. METHODS The methodology adopted in the study was evaluation of health program of evaluative research type 19 . All managers of the 42 CRIE in the Country were invited to participate, identified by a list provided by the PNI General Coordination in June 2011. In 21 states, only one CRIE was identified, located in the capital. Six states and the Federal District had more than one service: Para (two), Bahia (two), Federal District (four), Sao Paulo (seven), Rio de Janeiro (three), and Rio Grande do Sul (three). The CRIE are under the coordination of three instances: Brazilian Ministry of Health, by the PNI; State Secretariats of Health (SES); and the institution in which they are located (local level, usually higher complexity hospitals). The functioning and operation of CRIEs follow the Ordinance 48, from July 28, 2004 a , which sets the minimum structure and the necessary human resources, also considering possible emergency care, since the target audience includes individuals with increased risk of presenting adverse events following immunization. The operation must occur on a time that allows the application of immunobiologicals (or their distribution to application in other service) in cases of urgency, as prophylaxis after exposure. The indications of immunobiologicals follow the PNI recommendations, listed in the CRIE Manual, revised and updated periodically by a group of experts 15 . The data source was a semi-structured questionnaire developed for this study, consisting of 170 questions. The following program dimensions were studied: structure (existence of minimal physical area, institution in which it is inserted, equipment and inputs), human resources (number and training of professionals, workload, training for emergency care), and developed activities (application of immunobiologicals and their distribution to be applied in another service, assistance for adverse events following immunization, teaching and research activities). The regulations of the Ordinance that establishes general guidelines for CRIE operation guided the formulation of the questions, and the questionnaire was reviewed by the Technical Advisory of the General Coordination of PNI and by the Technical Management of Support to CRIE Management from the General Coordination of PNI. The questionnaire was available in online platform (virtual environment), from August 2011 to January 2012, and answered by the managers of the CRIE, or their substitutes, after agreement by signing the informed consent form. The project was approved by the Research Ethics Committee of the Hospital das Clinicas of Faculdade de Medicina of Universidade de São Paulo (Protocol of research 0281/10). In the descriptive analysis, we estimated the simple frequencies of the answers for the characterization of the services. To create service profiles with similar characteristics, we selected nuclear issues in the three evaluated dimensions and performed the analysis using the clustering for binary data with the squared Euclidean distance, by the furthest neighbor method (complete linkage). This technique allows one to move successively in the algorithm by clustering small groups in larger ones, according to the values of the squared Euclidean distance, producing a tree of groups named dendrogram. The cutoff level in the dendrogram is defined by the researcher and must represent the most suitable number of groups, in accordance with the research goals 9 . RESULTS All 42 CRIE filled the questionnaire, with varying completeness and quality. Figure 1 presents the number of immunobiologicals doses applied by the CRIE, from 2006 to 2010, by region and year (nine CRIE without information). In 2010, the CRIE accounted for 518,964 immunobiologicals doses applied. The record was made by applied doses and each person could receive more than one immunobiological. Few services (21.0%) recorded the attended cases in which the required immunobiological was not released. The total number of applied doses increased 66.0% in the studied period; but this increase was not uniform, being higher in the Northeast (118.0%), followed by South (87.0%), and Southeast (63.0%). In the North and Midwest there was not increase in the number of applied doses; however, this datum was not informed by 10 (59.0%) services of these regions. Thirty-five CRIE (83.0%) mentioned applying immunobiologicals of the PNI routine calendar and 31 (74.0%) reported participating in vaccination campaigns. Figure 1 Number of immunobiologicals doses applied in the Reference Centers for Special Immunobiologicals (CRIE), by region and year, in the period from 2006 to 2010. CRIE without information: 9 The 40 CRIE (95.0%) reported that they rendered assistance to adverse events following events; 36 (86.0%) mentioned counting on support of experts for this service; 34 (81.0%), with hospital support; and 28 (67.0%), with laboratory support (Table). Table Structural and human resources characteristics and developed activities in the Reference Centers for Special Immunobiologicals (CRIE) in accordance with the services and distribution profiles (%) of the structural and human resources characteristics and developed activities. Brazil, 2011. Characteristics Profile G1 G2 G3 G4 G5 Total (n = 12) (n = 6) (n = 5) (n = 13) (n = 6) (N = 42) % % % % % % Structure             Proportion of services that are inserted into teaching institutiona 58.0 0 0 0 100 31.0 Proportion of services that have minimal physical structure, according to the recommendation (reception, doctor’s office, preparation and/or vaccination room)a 92.0 50.0 60.0 54.0 83.0 69.0 Proportion of services that are inserted into building where other services operatea 75.0 100 80.0 92.0 100 88.1 Proportion of services that share the working area with other activitiesa 17.0 17.0 100 62.0 50.0 45.2 Proportion of services that have purpose-built vaccine refrigeratora 58.0 0 20.0 54.0 50.0 42.9 Proportion of services with preventive maintenance for cold chaina 92.0 83.0 20.0 0 83.0 52.4 Proportion of services that have electrical power generatorb 91.7 50.0 80.0 46.2 100 71.4 Proportion of services that have oxygen sourcea 92.0 33.0 20.0 69.0 17.0 57.1 Proportion of services that have computera 75.0 100 0 100 100 80.1 Proportion of services that have faxa 42.0 33.0 0 69.0 83.0 50.0 Human Resources             Proportion of services that have doctor in more than 50.0% of working hoursa 66.7 0 60.0 23.1 83.3 45.2 Proportion of services that have human resources training to meet emergenciesa 100.0 17.0 20.0 69.0 67.0 64.3 Developed activities             Proportion of services that apply routine immunobiologicalsa 92.0 33.0 60.0 100 100 83.3 Proportion of services that apply more than 80.0% of the released immunobiologicals in their own CRIEa 50.0 33.3 60.0 92.3 83.3 66.7 Proportion of services that participated in vaccination campaigns in the last 5 yearsa 67.0 83.0 80.0 69.0 83.0 73.8 Proportion of services that meet non-presential solicitations (only the requests)a 100 67.0 60.0 85.0 83.0 83.3 Proportion of services that provide a counter-reference document to the patient/requester, when there is no indication of immunobiologicala 100 67.0 40.0 54.0 67.0 69.0 Proportion of services that register the cases seen to which the immunobiological was not releasedb 16.7 16.7 20.0 23.1 33.3 21.4 Proportion of services with operation to the public more than 40 hours a weeka 58.3 66.7 40.0 84.6 50.0 59.5 Proportion of services that have available telephone contact for 24 hoursa 67.0 83.0 60.0 46.0 67.0 61.9 Proportion of services that meet people who received immunobiologicals in other health units and present adverse events following immunizationa 83.0 83.0 80.0 92.0 100 88.1 Proportion of services that have support of experts for cases of adverse events following immunizationb 91.7 100 80.0 76.9 83.3 85.7 Proportion of services that have laboratory support for cases of adverse events following immunizationb 83.3 83.3 40.0 61.5 50.0 66.7 Proportion of services that have hospital support for cases of adverse events following immunizationb 91.7 100 80.0 76.9 50.0 81.0 Proportion of services that rely on technical support Group for the discussion of cases of adverse events following immunizationb 58.3 0 40.0 53.8 66.7 47.6 Proportion of services that receive traineesa 75.0 33.0 60.0 85.0 100 73.8 Proportion of services that do or did researches in immunizationsa 42.0 0 0 15.0 100 31.0 G1: best structure; G2: immunobiologicals distributor; G3: incipient implementation; G4: vaccination room; G5: teaching and research a Questions used in the statistical clustering for the creation of CRIE profiles. b Characteristics not included in the creation of the profiles. Nine services reported having no doctor, at the time of the research, and four reported lack of nurse. There were trained professionals for emergency care in 27 services (64.0%). Regarding the infrastructure, lack of sufficient cold chain equipment was reported by 29 CRIE managers (69.0%) and 35 (83.0%) reported use of domestic refrigerators. Source of oxygen was available in 24 CRIE (57.0%). In the dendrogram analysis resulting from the application of the grouping method for binary data, using the Euclidean distance squared, by the furthest neighbor method (complete linkage), to the selected data for the characterization of the services, we opted by cutting the squared Euclidean distance measure in 20 points. Five clusterings resulted from this, in which the services were considered to have relevant characteristics in common and that differed them from other clusterings (Figure 2). Figure 2 Dendrogram of the clustering for binary data with the squared Euclidean distance, by the farthest neighbor method, with formation of five groups of services with similar features. CRIE: Reference Centers for Special Immunobiologicals; G1: best structure; G2: immunobiologicals distributor; G3: incipient implementation; G4: vaccination room; G5: teaching and research Group 1 was named “best structure” profile, because it featured the largest number of services that had the minimum recommended rooms (reception, doctor’s office, and preparation or vaccination room). This group also showed greater proportion of services with presence of oxygen source, purpose-built vaccine refrigerators, and preventive maintenance of the cold chain. All services of this group had trained professionals to meet emergencies. Most received trainees, but less than half did research activities (Table). Group 2 was named “immunobiologicals distributor” profile, for being the group with services that applied fewer doses of immunobiologicals in relation to the total distributed immunogens (administered and distributed) by the service. It also featured the lowest frequency of services that applied the vaccines of the routine calendars. Only half of the services had minimal physical structure and none of them had purpose-built vaccine refrigerators. It was the only group in which no CRIE had doctor present during more than half of the working hours. This group also showed the lowest frequency of trained professionals to meet emergencies. Group 3 was named “incipient implementation” because it includes services with insufficient structure, according to the foreseen in the Ordinance that regulates the CRIE. None of them had unique physical area, few had purpose-built vaccine refrigerator and oxygen supply and the services had the lowest rate of preventive maintenance of the cold chain. None of these CRIE had computer or fax. Less than half worked more than 40 hours a week, considered as minimum necessary period for application and distribution of immunobiologicals satisfactorily in the CRIE. Group 4, named “vaccination room” profile, included services that showed features that were more similar to a conventional vaccination room. All services did routine vaccination, and most of them applied the majority of their immunobiologicals and participated in vaccination campaigns. Most services worked more than 40 hours a week. In Group 5, called “education and research” profile, all CRIE were inserted into teaching hospitals, developed research activities and received trainees. This group presented the best index of presence of doctors. Most services had appropriate minimum physical structure. All applied routine vaccines and most applied more than 80.0% of released immunobiologicals and participated in vaccination campaigns. Twelve of the 18 CRIE of the “better structure” and “teaching and research” groups were located in the Southeast and South regions (Figure 3). From the five services of the “incipient implementation” profile, four were located in the Midwest region. In the North and Northeast, there is no evidence of the predominance of any group. Figure 3 Geographic distribution and profile (G1 to G5) of the Reference Centers for Special Immunobiologicals in Brazil by state, in 2011. All CRIE of the “incipient implementation” group were created before 2000, while, in the “better structure” and “teaching and research” groups, we observed services created at different times, from 1993 to 2009 (Figure 4). Figure 4 Distribution of Reference Centers for Special Immunobiologicals (CRIE) according to the profile (G1 to G5) and opening year. The degree of CRIE implementation was evaluated considering the Ordinance that regulates the services. The CRIE with the profiles called “better structure” and “teaching and research” were considered implemented. The CRIE with the “immunobiologicals distributor” and “vaccination room” profiles were considered partially implemented. The CRIE of the “incipient implementation” profile were considered not implemented. DISCUSSION This is the first systematic study that evaluated the implementation of CRIE in national perspective based on primary data, analyzing the structure, human resources, and developed activities dimensions. The data showed the diversity of the services’ situation nearly two decades after the creation of the special immunobiologicals program by PNI, indicating that the implementation of CRIE occurred without uniformity. Regarding the CRIE distribution across the Country, we did not observe a population coverage criterion for the definition of the number of services by state. According to personal information of state managers of immunization, distribution of these immunogens to the cities via Health Regions seems to have been implemented in some states, as an alternative to increase patients access to special immunobiologicals. We expected to observe a relation between the time of opening of the services and the degree of implementation, but the results do not prove this hypothesis, because all services of “incipient implementation” were inaugurated for over a decade (Figure 4). On the other hand, there seems to be a relation between the region where the CRIE were located and the degree of implementation, since most CRIE of G1 and G5 were located in the Southeast, while most CRIE of G3 were located in the Midwest region. Some faced barriers for the implementation of CRIE may have been caused by changes in policies for the Brazilian Unified Health System (SUS) over time. The conventional vaccination rooms were decentralized with the municipalization of health, what did not occurred with the CRIE. The local context of the institutions that house the CRIE, regarding the available structure, human resources, and their priorities, appear to have been instrumental in the implementation and performance of the services. Several factors may have contributed to the increase in the number of immunobiologicals doses applied by CRIE, such as: increased target audience, disclosure of the existence of CRIE and dissemination of knowledge about the recommendations of special immunobiologicals. This increase was most evident in the Northeast, followed by the South and Southeast regions. The absence of an increase in the number of doses applied in the North and Midwest regions may be due to information bias, since more than half of the services of these regions did not inform the number of immunobiologicals doses applied. Obstacles to the growth of some services may be arising from the inadequacy of the physical area, lack of cold chain equipment (or their inadequacy regarding regulations) 14 , and the lack of essential human resources. These factors can also impair the activities of evaluation and application of immunobiologicals, as well as the care of emergencies and investigation of adverse events following immunization. This panorama may have changed in recent years, since some improvements may have been implemented, such as the transfer of funds from the Ministry of Health to the states for adequacy of the cold chain b . Despite the heterogeneity within the groups, the profiles called “better structure” (G1) and “teaching and research” (G5), which amounted to 18 services (43.0%), resembled more what was proposed in the Ordinance that regulates the CRIE, being considered implemented. The characteristics that distinguished these two groups from the others were the higher proportion of services with more complete structure (G1) and higher proportion of services that developed activities such as application of immunogens in the very CRIE, participation in campaigns, application of routine vaccinations, and teaching and research activities (G5). The initial proposal for all CRIE was that they were privileged spaces for the development of teaching and research, training and capacity building. Some CRIE have developed studies in specific populations immunizations and adverse events following immunizaiton 3 , 6 , 10 - 12 , 17 , 18 , 20 - 23 . The CRIE that comprise the two most well-structured profiles could further develop teaching and research activities, counting on greater support and encouragement of the SES and the Ministry of Health/PNI. Meanwhile, placing the same requirements for services with very different conditions may hamper the identification of which activities could be considered more relevant to the different types of CRIE. Although not presenting appropriate conditions to develop research, the “immunobiological distributor” and “vaccination room” profiles could invest in training and capacity building for health professionals in their coverage area, as well as in the strengthening of their main activities of application and dispensing of special immunobiologicals. The services with “incipient implementation” profiles require extensive adequation of the structure and human resources and reorganization of the developed activities to find their vocation. Regarding the limitations of this study, the results must be analyzed with caution, since they were based on a questionnaire applied at a distance. With this type of assessment tool and without on-the-spot inspection, the responses must be treated as mentioned information and are subjected to the subjectivities of the respondents, related to their personal opinions or interpretations of the questions. Some questions were unused because they are not well formulated or their formatting has raised questions in the online form. Besides, the choice of the questions for the statistical analysis may have reflected in the final outcome of the implementation evaluation. Additionally, the introduction of new vaccines into the PNI routine immunization calendar, such as the 10-valent pneumococcal conjugate and meningococcal C conjugate (2010), IPV (2012), varicella (2013), and hepatitis A (2014), may have altered the demand for vaccines in the CRIE. Besides, new services were created (at least three more CRIE were opened since the end of data collection until 2014, in Minas Gerais, Rio de Janeiro, and Acre). However, there is no evidence that substantial changes have occurred in the services since the data collection. Considering that less than half of the services were classified as fully implemented, the results of this study may contribute to the restructuring of the services, including the review of the role of CRIE and reformulation of the special immunobiologicals program, based on the current context. Acknowledgments To Maykon Anderson Pires de Novais, for the preparation of the online tool for data collection, and to Rogério Ruscitto do Prado, for the conduction of the statistical analysis. To the General Coordination of the National Program of Immunization, for the support during the development of the project and interviews and during data collection. To the Institute of Health Technology Assessment, for the support in the conduction of the project. a Ministério da Saúde, Secretaria de Vigilância em Saúde. Portaria nº 48, de 28 de julho de 2004. Institui diretrizes gerais para funcionamento dos Centros de Referência para Imunobiológicos Especiais – CRIE, define as competências da Secretaria de Vigilância em Saúde, dos Estados, Distrito Federal e CRIE e dá outras providências. Diario Oficial Uniao. 29 jul 2004;seção 1:63. b Ministério da Saúde, Secretaria de Vigilância em Saúde. Portaria nº 3.301, de 26 de dezembro de 2013. Autoriza o repasse financeiro de investimento do Fundo Nacional de Saúde aos Fundos de Saúde Estaduais e Municipais para aquisição de equipamentos, material permanente e/ou unidade(s) móvel(is) para fomento e aprimoramento das condições de funcionamento da Rede de Frio. Diario Oficial Uniao. 26 dez 2013;seção 1:256. ==== Refs REFERENCES 1 Assen S Agmon-Levin N Elkayam O Cervera R Doran MF Dougados M et al EULAR recommendations for vaccination in adult patients with autoimmune inflammatory rheumatic diseases Ann Rheum Dis 2011 70 3 414 422 10.1136/ard.2010.137216 21131643 21131643 2 Brenol CV Mota LMH Cruz BA Pileggi GS Pereira IA Rezende LS et al Consenso 2012 da Sociedade Brasileira de Reumatologia sobre vacinação em pacientes com artrite reumatoide Rev Bras Reumatol 2013 53 1 4 23 10.1590/S0482-50042013000100002 23588512 3 Cabral IC Costa TNA Soares MJGO Dantas RA Santos SR Conhecimento de mães de crianças com câncer sobre vacinação especial Rev Enferm UERJ 2011 19 4 552 557 4 Danziger-Isakov L Kumar D Vaccination in solid organ transplantation Am J Transplant 2013 13 s4 311 317 10.1111/ajt.12122 23465023 5 Feijó RB Cunha J Krebs LS Calendário vacinal na infância e adolescência: avaliando diferentes propostas J Pediatr Rio J 2006 82 3 supl S4 14 10.1590/S0021-75572006000400002 6 Frauches DO Matos PASB Vatanabe JH Oliveira JF Lima APNB Moreira-Silva SF Vacinação contra pneumococo em crianças com doença falciforme no Espírito Santo, entre 2004 e 2007 Epidemiol Serv Saude 2010 19 2 165 172 10.5123/S1679-49742010000200009 7 Heijstek MW Ott de Bruin LM Borrow R Klis F Koné-Paut I Fasth A et al Vaccination in paediatric patients with auto-immune rheumatic diseases: a systemic literature review for the European League against Rheumatism evidence-based recommendations Autoimmun Rev 2011 11 2 112 122 10.1016/j.autrev.2011.08.010 21896342 8 Hilgendorf I Freund M Jilg W Einsele H Gea-Banacloche J Greinix H et al Vaccination of allogeneic haematopoietic stem cell transplant recipients: report from the international consensus conference on clinical practice in chronic GVHD Vaccine 2011 29 16 2825 2833 10.1016/j.vaccine.2011.02.018 21345379 9 Johnson RA Wichern DW Applied multivariate statistical analysis 6 Upper Saddle River Pearson Education 2007 681 689 10 Lopes MH Mascheretti M Franco MM Vasconcelos R Gutierrez EB Occurrence of early adverse events after vaccination against influenza at a Brazilian reference center Clinics 2008 63 1 21 26 10.1590/S1807-59322008000100005 18305872 11 Lopes MH Sartori AM Souza TV Mascheretti M Chaves TS Hepatitis B revaccination for healthcare workers who are anti-HBs-negative after receiving a primary vaccination series Rev Soc Bras Med Trop 2012 45 5 639 642 10.1590/S0037-86822012000500018 23152350 12 Macêdo LS Freire SM Andrade JAF Acidentes ocupacionais com BCG em salas de vacina do Estado da Bahia Rev Baiana Saude Publica 2013 37 1 222 235 13 Miller MA Rathore M Immunization in special populations Adv Pediatr 2012 59 1 95 136 10.1016/j.yapd.2012.04.017 22789576 14 Ministério da Saúde Secretaria de Vigilância em Saúde Departamento de Vigilância Epidemiológica Manual de rede de frio 4 Brasília (DF) Ministério da Saúde 2013 15 Ministério da Saúde Secretaria de Vigilância em Saúde Departamento de Vigilância das Doenças Transmissíveis Manual dos centros de referência para imunobiológicos especiais 4 Brasília (DF) Ministério da Saúde 2014 16 Ministério da Saúde Secretaria de Vigilância em Saúde Programa Nacional de Imunizações: 30 anos Brasília (DF) Ministério da Saúde 2013 Série C. Projetos e Programas e Relatórios 17 Miyaji KT Luiz AM Lara AN Chaves TSS Piorelli RO Lopes MH et al Active assessment of adverse events following yellow fever vaccination of persons aged 60 years and more Hum Vaccin Immunother 2013 9 2 277 282 10.4161/hv.22714 23291944 18 Monteiro AI Bellei NC Sousa AR Santos AM Weckx LY Infecções respiratórias em crianças menores de dois anos de idade submetidas a profilaxia com palivizumabe Rev Paul Pediatr 2014 32 2 152 158 10.1590/0103-0582201432214813 25119744 19 Novaes HMD Avaliação de programas, serviços e tecnologias em saúde Rev Saude Publica 2000 34 5 547 559 10.1590/S0034-89102000000500018 11105122 20 Oliveira DS Lara AN Luiz AM Miyaji KT Sartori AM Lopes MH Spontaneous reporting of adverse events following pandemic influenza A (H1N1) immunization in a reference center in the State of São Paulo, Brazil Rev Soc Bras Med Trop 2013 46 3 348 351 10.1590/0037-8682-1265-2013 23856866 21 Oliveira VLB Moura GN Caetano JA Esquema de imunobiológicos especiais aplicados em recém-nascidos em um serviço de referência no Ceará Rev Bras Promoç Saude 2010 23 2 188 195 10.5020/18061230.2010.p188 22 Piacentini S Contrera-Moreno L Eventos adversos pós-vacinais no município de Campo Grande (MS, Brasil) Cienc Saude Colet 2011 16 2 531 536 10.1590/S1413-81232011000200016 23 Sherlock MSM Cardoso MVLML Lopes MMCO Lélis ALPA Oliveira NR Imunização em criança exposta ou infectada pelo HIV em um serviço de imunobiológicos especiais Esc Anna Nery 2011 15 3 573 580 10.1590/S1414-81452011000300019 24 Silva CA Terreri MTR Barbosa CMPL Hilário MOE Pillegi GCS Ferriani VPL et al Consenso de imunização para crianças e adolescentes com doenças reumatológicas Rev Bras Reumatol 2009 49 5 562 589 10.1590/S0482-50042009000500007 25 Strikas RA Advisory committee on immunization practices recommended immunization schedules for persons aged 0 through 18 years - United States 2015 Morb Mortal Wkly Rep 2015 64 4 93 94 26 Succi RC Farhat CK Vacinação em situações especiais J Pediatr 2006 82 3 supl S91 100 10.1590/S0021-75572006000400011

PMC005xxxxxx/PMC5001806.txt

==== Front Rev Saude PublicaRev Saude PublicarspRevista de Saúde Pública0034-89101518-8787Faculdade de Saúde Pública da Universidade de São Paulo 0013110.1590/S1518-8787.2016050007093Commemorative Articles - Fifty YearsOral health in the agenda of priorities in public health A saúde bucal na agenda de prioridades em saúde pública Antunes José Leopoldo Ferreira I Toporcov Tatiana Natasha I Bastos João Luiz II Frazão Paulo III Narvai Paulo Capel III Peres Marco Aurélio IV I Departamento de Epidemiologia. Faculdade de Saúde Pública. Universidade de São Paulo. São Paulo, SP, BrasilII Programa de Pós-Graduação em Saúde Coletiva. Centro de Ciências da Saúde. Universidade Federal de Santa Catarina. Florianópolis, SC, BrasilIII Departamento de Prática de Saúde Pública. Faculdade de Saúde Pública. Universidade de São Paulo. São Paulo, SP, BrasilIV Australian Research Centre for Population Oral Health. School of Dentistry. The University of Adelaide. Adelaide, AustraliaCorrespondence: José Leopoldo Ferreira Antunes Faculdade de Saúde Pública Av. Dr. Arnaldo, 715 Cerqueira César 01246-904 São Paulo, SP, Brasil E-mail: leopoldo@usp.br Authors’ Contribution: Conception and planning of the study: JLFA, TNT. All authors participated in the data collection and analysis, the writing of the manuscript, and the approval of the final version. All authors assume public responsibility for the content of the article. Conflict of Interest: The authors declare no conflict of interest. 26 8 2016 2016 50 5722 4 2016 10 7 2016 This is an Open Access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. ABSTRACT This study describes the scientific production on oral health diffused in Revista de Saúde Pública, in the 50 years of its publication. A narrative review study was carried out using PubMed, as it is the search database that indexes all issues of the journal. From 1967 to 2015, 162 manuscripts specifically focused on oral health themes were published. This theme was present in all volumes of the journal, with increasing participation over the years. Dental caries was the most studied theme, constantly present in the journal since its first issue. Periodontal disease, fluorosis, malocclusions, and other themes emerged even before the decline of dental caries indicators. Oral health policy is the most recurring theme in the last two decades. Revista de Saúde Pública has been an important vehicle for dissemination, communication, and reflection on oral health, contributing in a relevant way to the technical-scientific interaction between professionals in this field. RESUMO Este estudo descreve a produção científica sobre saúde bucal veiculada na Revista de Saúde Pública, nos cinquenta anos de sua publicação. Foi realizado estudo de revisão narrativa, utilizando o PubMed como mecanismo de busca que indexa todos os fascículos da revista. De 1967 a 2015, foram publicados 162 manuscritos com foco específico em temas de saúde bucal. Essa temática esteve presente em todos os volumes da revista, com participação crescente ao longo dos anos. Cárie dentária foi o tema mais estudado, marcando presença constante na revista desde seu primeiro fascículo. Doença periodontal, fluorose, oclusopatias e outros temas emergiram antes mesmo do declínio dos indicadores de cárie. Políticas de saúde bucal é o tema mais recorrente nas duas últimas décadas. A Revista de Saúde Pública tem sido importante veículo de divulgação, comunicação e reflexão sobre saúde bucal, contribuindo de modo relevante para a interação técnico-científica entre os profissionais da área. Dental CariesOral HealthPublic HealthReviewHistorical Article ==== Body INTRODUCTION Many oral health conditions are recognized as public health problems because of their prevalence, severity, individual and community impact, which entail costs to the health system and the existence of effective methods of prevention and treatment 10 , 59 . Untreated dental caries is considered the most prevalent condition throughout the world 22 ; severe periodontal disease is the sixth 21 . Recognizing the importance of oral and craniofacial diseases on the global burden of morbidity and in association with systemic diseases, treatment costs and the possibility of applying effective measures of promotion and prevention, the 60th World Assembly of the World Health Organization adopted a resolution recommending the member states to increase budgetary appropriations devoted to the control of such diseases and conditions 66 . In Brazil, oral health is one of the three most important reasons for health care demand 20 . As public health problems, confronting adverse conditions of oral health requires coordinated action on the part of the society, in particular of health services. To guide this action, it is essential to conduct epidemiological studies, of planning and management, and studies on social sciences in health specifically focusing on such conditions in its multiple dimensions. This understanding permeates the entire history of Revista de Saúde Pública (RSP). When celebrating its 40 years, we noted that oral health was one of the main thematic areas of the journal 38 . By studies published in this scientific dissemination vehicle, we can ponder the participation of oral health issues in the evolution of the agenda of priorities in public health. On the occasion of RSP’s 50th anniversary, the present study aimed at describing the scientific production on oral health issues diffused in the journal throughout its history. Published Articles We carried out a narrative review study. For the retrieval of the published articles, we used the PubMed search database, which indexes all issues of the journal. Two examiners conducted the search independently and resolved conflicts by consensus. The selected articles were classified in subthemes and synthesis for presentation purposes. From 1967 to 2015, RSP published 162 manuscripts specifically focused on oral health themes, including original articles, theoretical review, reviews, editorials, and previous notes. The Figure shows the distribution of these articles by decades (blocks) and three-year periods (line), showing that RSP approached these themes since its early years and that the concentration of the journal regarding oral health has grown over time. Figure Number of articles on oral health themes published in Revista de Saúde Pública, according to decades (blocks) and three-year periods (line). Sao Paulo, SP, Southeastern Brazil, 1967-2015. Dental Caries: Prevalence Studies and Inequality Dental caries was the oral health condition that most motivated studies on the pages of RSP. It was also a constant theme during the five decades of the journal. Sixty-nine articles assessed the prevalence of dental caries in the population in general and in specific groups, inequality in its distribution, preventive resources, and methodological aspects of the forms of measuring the severity. Already in its early years, dental caries prevalence was described for different cities, having as reference the addition of fluoride in the water supply system 62 , 63 . These studies have contributed to consolidate the conviction that water fluoridation is effective in reducing dental caries indicators, and should, therefore, be extended. Anticipating the concern with social inequalities in health, pioneering studies addressed differences in the prevalence of dental caries according to social strata. In the first issue of RSP, Souza et al. 56 compared indicators of the disease among black and white school children; this inequality would be studied again in the journals of 1970 57 and 1974 7 . Viegas 65 assessed the incidence of dental caries during pregnancy; Castellanos 7 studied the prevalence of the disease in orphanages in the city of Sao Paulo. The first study to evaluate socioeconomic differentials in the experience of the disease was published by Yankilevich et al. 67 , having as reference the city of Córdoba, in Argentina, and employing a Marxist-inspired approach, using the concept of social classes. In the latest period, a decline in the prevalence of dental caries was observed. Articles published in RSP described and analyzed empirical data justifying this observation. Narvai et al. 34 compared the results of epidemiological surveys of dental caries carried out in the city of Sao Paulo from 1970 to 1996, and concluded that the reduction of dental caries indicators was largely due to the addition of fluoride in public water supply and in dentifrices, as well as the introduction of preventive programs in the public health network. In the same period, Freysleben et al. 17 compared the presence of dental caries in school children of Florianopolis in 1971 and 1997, concluding that the reduction of the prevalence was real and could not be attributed only to changes in diagnostic criteria. However, the decline in dental caries prevalence was accompanied by increased inequality of their indicators between the social strata. This phenomenon, referred to as polarization of dental caries experience 35 , was also described and analyzed in the pages of RSP for different age groups 3 , 16 , 55 . Original Themes and Methodological Innovations Methodological aspects of measures of dental caries in the population also motivated pioneering studies on RSP. In 1973, Souza 58 evaluated a synthetic measure of dental caries experience by simplifying the traditional Decay-missing-filled (DMF) index, which accounts for the decayed teeth, lost due to dental caries and filled (restored). This theme would be explored in the pages of the journal by Guimarães and Guimarães 19 , who proposed original and simplified methodology for measuring prevalence of dental caries and which was applied in subsequent studies 12 . Analytical techniques widely used in studies on public health also had pioneer application to the area of oral health in the pages of RSP. Linear regression 61 and logistic regression 40 were applied to test the association between dental caries experience and behavioral and socioeconomic factors. RSP also innovated by publishing, in 1997, a study with georeferencing of dental caries indicators in the municipalities of the state of Sao Paulo 44 . In 2007, Celeste et al. 9 published the first article on RSP employing multilevel analysis to relate dental caries indicators in adolescents with the characteristics of individuals and their residential contexts. Still considering original themes of research in oral health, RSP has published studies that, in time, have had an impact on the prevention of dental caries. Mouthwashes and topical applications of fluoride gel in mouth trays were object of study already in the first volumes of the journal, becoming popular in Brazil 8 , 46 , 64 . Pinto 47 considered the hypothesis of the addition of fluoride to common salt and underlined the conditions under which the measure could be considered a complement to water fluoridation in the Brazilian context. Peres et al. 42 described the addition of sugar in syrups and oral drug solutions, with potential damage to children’s oral health. The participation of pediatricians 51 and community health agents 15 in promoting children’s oral health also promoted original research on RSP. The aging of the population, a theme so important to contemporary public health, motivated studies on oral health in RSP in different periods. Dental caries, periodontal disease, dental loss, and prosthetic use were discussed, in 1992, in a study identifying the need for specific odontological policies and programs for older people 50 . Still focused on this age group, several studies have addressed the data of epidemiological surveys carried out in Brazil in recent decades. Singh et al. 54 analyzed obesity and dental loss in older people and showed that these conditions relate differently between the sexes. Martins et al. 30 studied the relationship between oral health conditions and housing, individual characteristics, and behavior in a study that included more than 5,000 older people of Brazil. Figueiredo et al. 14 evaluated the masticatory capacity of adults in Florianopolis, and Ribeiro et al. 49 described the dental loss of adults and focused the preservation of functional teeth and reduced dental arch as alternatives to prosthetic treatment. Still regarding the latest period, articles on other original themes were published, which are still reverberating in professional field and have given rise to new approaches such as self-perception of oral health of teenagers 41 and adults 27 , and the manifestation of lesions of the oral mucosa in patients with HIV/AIDS 45 . The welfare of professionals working with oral health have also motivated studies. Regarding the decrease in risk of infection of these professionals, Martins and Barreto 32 described prevalence of vaccination for hepatitis B virus among dentists according to area of expertise, while Garcia and Blank 18 found the adequacy of occupational post-exposure conducts to biological material for oral health workers. Nunes and Freire 36 evaluated the quality of life of public health dentists and found low values in physical and psychological domains and high in social relations and environment domains. Emerging Themes Even before the decline of dental caries indicators, other oral health themes emerged as public health problems, requiring action of health authorities and health services. In particular, several studies of RSP focused on distribution, forms of measurement, associated factors, and consequences of malocclusion, gingival inflammation, and periodontal diseases, conditions widely prevalent. Tomita et al. 60 pointed the relationship between sucking habits and severity of occlusion such as open bite, dental crowding, and cross bite. The theme would return to the pages of the journal in 2007 39 and 2013 53 . In 1969, the need for gingival treatment was the subject of RSP 6 . Periodontal disease, gingival bleeding, and dental calculus were themes of the journal, including evaluating the possible association between their manifestation in pregnant women and the birth of children with low weight 11 . Dental fluorosis, another emerging theme on the agenda of priorities in public health, was present in the pages of RSP in the last two decades of its 50 years. Studies with different methodological schemes assessed its prevalence and its perception by the population as well as the related beliefs and attitudes. Lima and Cury 24 measured fluoride intake of children by water and dentifrice. Subsequent studies have evaluated the concentration of fluoride in several foods and bottled water for sale. Oral cancer and orofacial clefts are less prevalent conditions; but their severity justifies the inclusion on the agenda of priorities in public health. These themes are also present in RSP, in the latest period. Loffredo et al. estimated the incidence of oral clefts in Brazil 25 and carried out a case-control study 26 pointing heredity and pollution as the main risk factors. Oral cancer has been studied as to its risk factors 28 , 29 and inequalities of gender and race 2 . The planning of dental services for patients requiring special care has also been the object of studies published in RSP. Oliveira et al. 37 studied the dental care aimed at children and adolescents with Down syndrome, and underlined the importance of the orientation of health professionals who take care of these patients in order to provide full care. Elizondo et al. 13 analyzed the dental care aimed at patients with HIV/AIDS in Mexico by examining their perception regarding the persistence of stigma on the part of professionals. Oral Health Policies The study on oral health policies stood out in the last two decades. About a third of the articles published in RSP in this period addressed oral health policies and related issues of the planning and management of dental services. Such studies proposed and evaluated programs such as fluoridation of the water supply network, preventive procedures and of oral health promotion, the importance of the auxiliary professionals composing oral health teams in direct provision of services, and the dental care in public and private networks. Even before the Brazilian Unified Health System (SUS) was deployed, Vitor Gomes Pinto 48 published an article on RSP scaling the treatment needs and human resources in the dental area in order to foster the implementation of a program of basic oral health services of national range. In the following years, the dental care in SUS has been evaluated regarding its extent and effectiveness. Lacerda et al. 23 and Martins et al. 31 evaluated factors associated with self-perception of need to visit the dentist such as pain of dental origin and difficult to chew. Baldani et al. 4 analyzed the provision of public dental services in Paraná and identified an expansion in oral health actions in previous years, with a pro-equity trend in the provision and use of dental services in primary health care. Antunes and Narvai 1 also evaluated favorably the expansion of dental care in the public network after the implementation, on the part of SUS, of Family Health Strategy and of Dental Specialties Centers. Peres et al. 43 documented the reduction in inequalities between socioeconomic strata when using dental services, comparing data collected in 2003 and 2008 for National Research by Residential Samples. Moysés et al. 33 analyzed the surveillance policy of oral health in force in Brazil, highlighting, on one hand, expressive progress and, on the other, obstacles and difficulties still present. Camargo et al. 5 evaluated the use of dental services in pre-scholar children, distinguishing the reason that led to the consultation. By this strategy, they could identify factors associated with consultations aimed at routine evaluation and to solve problems. They concluded that the rate of use of dental services of pre-scholar children is still lesser than that of medical consultations (childcare) and also that, besides the socioeconomic condition, maternal behaviors have an important role for using routine dental services. On the occasion of the ten-year anniversary of the Programa Brasil Sorridente (Smiling Brazil Program), Scherer and Scherer 52 have focused the changes at work in oral health in primary health care, identifying progress achieved and challenges still persistent at work in oral health at this level of care. According to the authors, the public network professionals tend to reproduce the dominant biomedical model, requiring continued efforts of management, training, and permanent education for understanding the dynamics of the work in order to get significant changes to local realities. Final Considerations RSP has been an important vehicle of communication of scientific knowledge in the field of collective oral health. Since its first issue, and in all its volumes, the journal has been publishing search results about oral health themes of interest to public health, contributing to scientific reflection, professional training, and health planning. The Brazilian Ministry of Health recognized the importance of the journal for the scientific communication in the field of oral health, choosing it to publish, in 2013, a supplement to studies describing and analyzing the results of the National Survey on Oral Health, the epidemiological survey, also known as SBBrasil 2010. Dental caries has held a prominent position in the published articles. It was also the study on dental caries that, in part, promoted the discussion of methodological innovations, both in terms of its evaluation in epidemiological investigations, and in statistical techniques of analysis. Over time, other harms to oral health, such as adverse periodontal conditions,fluorosis, malocclusions, and oral cancer, were explored according to varied theoretical-methodological perspectives and linked to the Brazilian context. The growth of publications in the thematic area of policies, planning, management, and evaluation of oral health services sought to answer, among other priorities, the problem represented by dental caries, including demands for qualified services to handle its direct demonstration or more severe consequences. Publications diffused in RSP expose a set of major transformations in scientific thought and valuation of the oral health theme. The synthetic description of these studies provides identifying the involvement of different oral health themes in the evolution of the agenda of priorities in public health. The journal has been employed by researchers in the field as an instrument of dissemination, communication, and reflection of scientific knowledge, contributing in a relevant way to education and technical, scientific, and professional interaction. ==== Refs REFERENCES 1 Antunes JLF Narvai PC Políticas de saúde bucal no Brasil e seu impacto sobre as desigualdades em saúde Rev Saude Publica 2010 44 2 360 365 10.1590/S0034-89102010000200018 20339637 2 Antunes JLF Toporcov TN Biazevic MG Boing AF Bastos JL Gender and racial inequalities in trends of oral cancer mortality in Sao Paulo, Brazil Rev Saude Publica 2013 47 3 470 478 10.1590/S0034-8910.2013047003724 24346559 3 Ardenghi TM Piovesan C Antunes JLF Desigualdades na prevalência de cárie dentária não tratada em crianças pré-escolares no Brasil Rev Saude Publica 2013 47 3 129 137 10.1590/S0034-8910.2013047004352 24626589 4 Baldani MH Almeida ES Antunes JLF Eqüidade e provisão de serviços públicos odontológicos no estado do Paraná Rev Saude Publica 2009 43 3 446 454 10.1590/S0034-89102009000300008 19448913 5 Camargo MBJ Barros AJD Frazão P Matijasevich A Santos IS Peres MA et al Preditores da realização de consultas odontológicas de rotina e por problema em pré-escolares Rev Saude Publica 2012 46 1 87 97 10.1590/S0034-89102012005000004 22218761 6 Cardoso GM Verificação da aplicabilidade e comportamento do índice de Lowell Smith, para estimar as necessidades de tratamento gengival Rev Saude Publica 1969 3 2 117 132 10.1590/S0034-89101969000200001 5372720 7 Castellanos RA Aspectos epidemiológicos da cárie dental em escolares brancos e não brancos de ambos os sexos internos em sete orfanatos da cidade de São Paulo, Brasil, em 1972 Rev Saude Publica 1974 8 1 51 62 10.1590/S0034-89101974000100006 4152676 8 Castellanos RA Estudo comparativo do efeito de solução de NaF a 0,5%, através de bochecho, escovação e ambos, na prevenção da cárie dental Rev Saude Publica 1983 17 6 461 475 10.1590/S0034-89101983000600003 6426032 9 Celeste RK Nadanovsky P Leon AP Associação entre procedimentos preventivos no serviço público de odontologia e a prevalência de cárie dentária Rev Saude Publica 2007 41 5 830 838 10.1590/S0034-89102007000500018 17923905 10 Costa JSD Victora CG O que é “um problema de saúde pública”? Rev Bras Epidemiol 2006 9 1 144 146 10.1590/S1415-790X2006000100018 11 Cruz SS Costa MC Gomes IS Filho Vianna MI Santos CT Doença periodontal materna como fator associado ao baixo peso ao nascer Rev Saude Publica 2005 39 5 782 787 10.1590/S0034-89102005000500013 16254655 12 Cypriano S Sousa MLR Wada RS Avaliação de índices CPOD simplificados em levantamentos epidemiológicos de cárie dentária Rev Saude Publica 2005 39 2 285 292 10.1590/S0034-89102005000200021 15895150 13 Elizondo JE Treviño AC Violant D Dentistry and HIV/AIDS related stigma Rev Saude Publica 2015 49 79 10.1590/S0034-8910.2015049005877 14 Figueiredo DR Peres MA Luchi CA Peres KG Fatores associados às dificuldades de adultos na mastigação Rev Saude Publica 2013 47 6 1028 1038 10.1590/S0034-89102013000901028 24626541 15 Frazão P Marques D Efetividade de programa de agentes comunitários na promoção da saúde bucal Rev Saude Publica 2009 43 3 463 471 10.1590/S0034-89102009000300010 19330198 16 Freire MC Reis SC Figueiredo N Peres KG Moreira RS Antunes JLF Determinantes individuais e contextuais da cárie em crianças brasileiras de 12 anos em 2010 Rev Saude Publica 2013 47 3 40 49 10.1590/S0034-8910.2013047004322 24626580 17 Freysleben GR Peres MA Marcenes W Prevalência de cárie e CPO-D médio em escolares de doze a treze anos de idade nos anos de 1971 e 1997, região Sul, Brasil Rev Saude Publica 2000 34 3 304 308 10.1590/S0034-89102000000300015 10920455 18 Garcia LP Blank VL Condutas pós-exposição ocupacional a material biológico na odontologia Rev Saude Publica 2008 42 2 279 286 10.1590/S0034-89102008000200013 18372979 19 Guimarães LOC Guimarães AMR Simplificação do índice CPO dos 18 aos 25 anos Rev Saude Publica 1990 24 5 407 411 10.1590/S0034-89101990000500009 2101533 20 Instituto Brasileiro de Geografia e Estatística Um panorama da saúde no Brasil: acesso e utilização dos serviços, condições de saúde e fatores de risco e proteção à saúde 2008 Rio de Janeiro Instituto Brasileiro de Geografia e Estatística 2010 citado 2016 abr 22 http://www.ibge.gov.br/home/estatistica/populacao/panorama_saude_brasil_2003_2008/ 21 Kassebaum NJ Bernabé E Dahiya M Bhandari B Murray CJ Marcenes W Global burden of severe periodontitis in 1990-2010: a systematic review and meta-regression J Dent Res 2014 93 11 1045 1053 10.1177/0022034514552491 25261053 22 Kassebaum NJ Bernabé E Dahiya M Bhandari B Murray CJ Marcenes W Global burden of untreated caries: a systematic review and metaregression J Dent Res 2015 94 5 650 658 10.1177/0022034515573272 25740856 23 Lacerda JT Simionato EM Peres KG Peres MA Traebert J Marcenes W Dor de origem dental como motivo de consulta odontológica em uma população adulta Rev Saude Publica 2004 38 3 453 458 10.1590/S0034-89102004000300017 15243677 24 Lima YB Cury JA Ingestão de flúor por crianças pela água e dentifrício Rev Saude Publica 2001 35 6 576 581 10.1590/S0034-89102001000600012 11799472 25 Loffredo LCM Freitas JA Grigolli AA Prevalência de fissuras orais de 1975 a 1994 Rev Saude Publica 2001 35 6 571 575 10.1590/S0034-89102001000600011 11799471 26 Loffredo LCM Souza JM Yunes J Freitas JA Spiri WC Fissuras lábio-palatais: estudo caso-controle Rev Saude Publica 1994 28 3 213 217 10.1590/S0034-89101994000300009 7747081 27 Luchi CA Peres KG Bastos JL Peres MA Desigualdades na autoavaliação da saúde bucal em adultos Rev Saude Publica 2013 47 4 740 751 10.1590/S0034-8910.2013047004364 24346665 28 Marchioni DM Fisberg RM Góis JF Filho Kowalski LP Carvalho MB Abrahão M et al Dietary patterns and risk of oral cancer: a case-control study in São Paulo, Brazil Rev Saude Publica 2007 41 1 19 26 10.1590/S0034-89102007000100004 17273630 29 Marques LA Eluf J Neto Figueiredo RA Góis JF Filho Kowalski LP Carvalho MB et al Oral health, hygiene practices and oral cancer Rev Saude Publica 2008 42 3 471 479 10.1590/S0034-89102008000300012 18470367 30 Martins AM Barreto SM Silveira MF Santa-Rosa TT Pereira RD Autopercepção da saúde bucal entre idosos brasileiros Rev Saude Publica 2010 44 5 912 922 10.1590/S0034-89102010005000028 20882263 31 Martins AM Barreto SM Pordeus IA Fatores relacionados à autopercepção da necessidade de tratamento odontológico entre idosos Rev Saude Publica 2008 42 3 487 496 10.1590/S0034-89102008000300014 18470368 32 Martins AM Barreto SM Vacinação contra a hepatite B entre cirurgiões dentistas Rev Saude Publica 2003 37 3 333 338 10.1590/S0034-89102003000300011 12792684 33 Moysés SJ Pucca GA Junior Paludetto M Junior Moura L Avanços e desafios à Política de Vigilância à Saúde Bucal no Brasil Rev Saude Publica 2013 47 3 161 167 10.1590/S0034-8910.2013047004329 24626593 34 Narvai PC Castellanos RA Frazão P Prevalência de cárie em dentes permanentes de escolares do Município de São Paulo, SP, 1970-1996 Rev Saude Publica 2000 34 2 196 200 10.1590/S0034-89102000000200015 10881157 35 Narvai PC Frazão P Roncalli AG Antunes JLF Cárie dentária no Brasil: declínio, polarização, iniqüidade e exclusão social Rev Panam Salud Publica 2006 19 6 385 393 10.1590/S1020-49892006000600004 16968593 36 Nunes MF Freire MCM Qualidade de vida de cirurgiões-dentistas que atuam em um serviço público Rev Saude Publica 2006 40 6 1019 1026 10.1590/S0034-89102006000700009 17173158 37 Oliveira AC Czeresnia D Paiva SM Campos MR Ferreira EF Uso de serviços odontológicos por pacientes com síndrome de Down Rev Saude Publica 2008 42 4 693 699 10.1590/S0034-89102008000400016 18709245 38 Pereira JCR Revista de Saúde Pública: quarenta anos da produção científica no Brasil Rev Saude Publica 2006 40 Esp 148 159 10.1590/S0034-89102006000400020 16924315 39 Peres KG Barros AJ Peres MA Victora CG Effects of breastfeeding and sucking habits on malocclusion in a birth cohort study Rev Saude Publica 2007 41 3 343 350 10.1590/S0034-89102007000300004 17515986 40 Peres KG Bastos JRM Latorre MRDO Severidade de cárie em crianças e relação com aspectos sociais e comportamentais Rev Saude Publica 2000 34 4 402 408 10.1590/S0034-89102000000400014 10973161 41 Peres KG Cascaes AM Leão ATT Côrtes MIS Vettore MV Aspectos sociodemográficos e clínicos da qualidade de vida relacionada à saúde bucal em adolescentes Rev Saude Publica 2013 47 3 19 28 10.1590/S0034-8910.2013047004361 24626578 42 Peres KG Oliveira CT Peres MA Raymundo MS Fett R Sugar content in liquid oral medicines for children Rev Saude Publica 2005 39 3 486 489 10.1590/S0034-89102005000300022 15997327 43 Peres KG Peres MA Boing AF Bertoldi AD Bastos JL Barros AJ Redução das desigualdades sociais na utilização de serviços odontológicos no Brasil entre 1998 e 2008 Rev Saude Publica 2012 46 2 250 258 10.1590/S0034-89102012000200007 22437856 44 Peres MA Narvai PC Calvo MC Prevalência de cárie dentária em crianças aos doze anos de idade, em localidades do Estado de São Paulo, Brasil, período 1990-1995 Rev Saude Publica 1997 31 6 594 600 10.1590/S0034-89101997000700008 9629715 45 Petruzzi MNMR Cherubini K Salum FG Figueiredo MAZ Risk factors of HIV-related oral lesions in adults Rev Saude Publica 2013 47 1 52 59 10.1590/S0034-89102013000100008 23703130 46 Pinto IL Prevenção da cárie dental com aplicações tópicas semestrais de flúor-fosfato acidulado Rev Saude Publica 1993 27 4 277 290 10.1590/S0034-89101993000400008 8209160 47 Pinto VG Prevenção da cárie dental: a questão da fluoretação do sal Rev Saude Publica 1982 16 1 66 72 10.1590/S0034-89101982000100006 7134779 48 Pinto VG Saúde bucal no Brasil Rev Saude Publica 1983 17 4 316 327 10.1590/S0034-89101983000400006 6658352 49 Ribeiro MT Rosa MA Lima RM Vargas AM Haddad JP Ferreira EF Edentulism and shortened dental arch in Brazilian elderly from the National Survey of Oral Health 2003 Rev Saude Publica 2011 45 5 817 823 10.1590/S0034-89102011005000057 21845293 50 Rosa AGF Castellanos Fernandez RA Pinto VG Ramos LR Condições de saúde bucal em pessoas de 60 anos ou mais no Município de São Paulo (Brasil) Rev Saude Publica 1992 26 3 155 160 10.1590/S0034-89101992000300005 1342495 51 Schalka MMS Rodrigues CRMD A importância do médico pediatra na promoção da saúde bucal Rev Saude Publica 1996 30 2 179 186 10.1590/S0034-89101996000200010 9077017 52 Scherer CI Scherer MD Avanços e desafios da saúde bucal após uma década de Programa Brasil Sorridente Rev Saude Publica 2015 49 98 10.1590/S0034-8910.2015049005961 53 Silveira LM Prade LS Ruede AM Haeffner LSB Weinmann ARM Aleitamento materno e sua influência nas habilidades orais de crianças Rev Saude Publica 2013 47 1 37 43 10.1590/S0034-89102013000100006 23703128 54 Singh A Peres MA Peres KG Bernardo CO Xavier A D’Orsi E Gender differences in the association between tooth loss and obesity among older adults in Brazil Rev Saude Publica 2015 49 44 10.1590/S0034-8910.2015049005590 55 Sousa MLR Rando-Meirelles MPM Tôrres LHN Frias AC Cárie dentária e necessidades de tratamento em adolescentes paulistas Rev Saude Publica 2013 47 3 50 58 10.1590/S0034-8910.2013047004340 24626581 56 Souza JMP Arrillaga AN Ochoa FV Rocha O Prevalência da cárie dental em brancos e não brancos Rev Saude Publica 1967 1 1 38 43 10.1590/S0034-89101967000100006 5603523 57 Souza JMP Atendimento dentário de crianças de oito a doze anos de idade, nos grupos escolares estaduais da capital de São Paulo Rev Saude Publica 1970 4 1 61 70 10.1590/S0034-89101970000100010 5505839 58 Souza JMP CPO e MID: alguns resultados obtidos em meninos brancos, de 8 a 12 anos Rev Saude Publica 1973 7 2 93 101 10.1590/S0034-89101973000200003 4148019 59 Thomson WM Sheiham A Spencer AJ Sociobehavioral aspects of periodontal disease Periodontol 2000 2012 60 1 54 63 10.1111/j.1600-0757.2011.00405.x 60 Tomita NE Bijella VT Franco LJ Relação entre hábitos bucais e má oclusão em pré-escolares Rev Saude Publica 2000 34 3 299 303 10.1590/S0034-89102000000300014 10920454 61 Tomita NE Bijella VT Lopes ES Franco LJ Prevalência de cárie dentária em crianças da faixa etária de 0 a 6 anos matriculadas em creches: importância de fatores socioeconômicos Rev Saude Publica 1996 30 5 413 420 10.1590/S0034-8910199600050000310.1590/S0034-89101985000400001 9269090 62 Viegas Y Viegas AR Análise dos dados de prevalência de cárie dental na cidade de Barretos, SP, Brasil, depois de dez anos de fluoretação da água de abastecimento público Rev Saude Publica 1985 19 4 287 299 10.1590/S0034-89101985000400001 3832373 63 Viegas Y Viegas AR Análise dos dados de prevalência de cárie dental na cidade de Campinas, SP, Brasil, depois de dez anos de fluoração da água de abastecimento público Rev Saude Publica 1974 8 4 399 409 10.1590/S0034-89101974000400006 4156419 64 Viegas Y Efeito inibidor de cárie dental de uma única aplicação tópica de solução de fluofosfato acidulada em adultos jovens. Experiência de um ano Rev Saude Publica 1970 4 1 55 60 10.1590/S0034-89101970000100009 5505838 65 Viegas Y Gestação e cárie dental Rev Saude Publica 1970 4 1 71 77 10.1590/S0034-89101970000100011 5505840 66 World Health Organization World Health Assembly Oral health: action plan for promotion and integrated disease prevention: report by Secretariat Geneva World Health Organization 2007 2016 abr 22 http://apps.who.int/iris/handle/10665/22448 67 Yankilevich ERLM Cattoni STD Cornejo LS Battellino LJ Distribución de la caries dental en niños preescolares en una región urbana, Argentina, 1992 Rev Saude Publica 1993 27 6 436 444 10.1590/S0034-89101993000600006 7997814

PMC005xxxxxx/PMC5001808.txt

==== Front PLoS CurrPLoS CurrPLoS CurrplosPLoS Currents2157-3999Public Library of Science San Francisco, USA 2761717010.1371/currents.outbreaks.873427b89ab9c75eb90c8ddb8d8c7c90Research ArticleInvasion Dynamics of Teratogenic Infections in Light of Rubella Control: Implications for Zika Virus Metcalf C. Jessica E. Department of Ecology and Evolutionary Biology, Princeton University, Princeton, New Jersey, USABarrett Alan Department of Pathology and Sealy Center for Vaccine Development, University of Texas Medical Branch, Galveston, Texas, USA16 8 2016 8 ecurrents.outbreaks.873427b89ab9c75eb90c8ddb8d8c7c90© 2016 Metcalf, Barrett, et al2016Metcalf, Barrett, et alThis is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Introduction: The greatest burden for a subset of pathogens is associated with infection during pregnancy. Evidence for teratogenic effects of Zika Virus have highlighted the importance of understanding the epidemiology of such pathogens. Rubella is perhaps the most classic example, and there is much to be learned from the long history of modelling associated with this virus. Methods: We extended an existing framework for modeling age-specific dynamics of rubella to illustrate how the body of knowledge of rubella dynamics informs the dynamics of teratogenic infections more broadly, and particularly the impact of control on such infections in different transmission settings. Results: During invasion, the burden in women of childbearing age is expected to peak, but then fall to low levels before eventually levelling out. Importantly, as illustrated by rubella dynamics, there is potential for a paradoxical effect, where inadequate control efforts can increase the burden. Conclusions: Drawing on the existing body of work on rubella dynamics highlights key knowledge gaps for understanding the risks associated with Zika Virus. The magnitude and impacts of sterilizing immunity, plus antigenic maps measuring cross-protection with other flaviviruses, and the magnitude of transmission, as well as likely impact of control efforts on transmission are likely to be key variables for robust inference into the outcome of management efforts for Zika Virus. CJEM was funded by the Gates Foundation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript ==== Body Introduction The impacts of rubella infection during pregnancy have been recognized since the pioneering work of Norman Gregg following introductions of this pathogen into communities in Australia during the Second World War1. Theoretical research into the characteristics of the population burden of pathogens with this unique manifestation followed2; and the teratogenic outcomes of rubella3 and their implications across diverse demographic, epidemiological and control settings4 , 5 , 6 , 7 are currently well described. In particular, in an endemic setting, incomplete vaccination coverage can lead to an increase in the teratogenic burden associated with rubella infection (Congenital Rubella Syndrome, or CRS), an outcome also known as the ‘paradoxical effect’. This occurs because infection with rubella provides lifelong immunity, so that in the absence of vaccination, the average age of infection can be low. Vaccination tends to increase the average age of infection by making the infection rare; and if coverage has not sufficiently reduced incidence to offset this increase, the outcome can be an increase in the burden of CRS2 , 7. Perhaps the greatest impact theoretical modeling has ever had on public health8 was the resulting restriction of introduction of the rubella-containing vaccine into many parts of the world9. Other than some suggestive patterns from the veterinary literature with Japanese encephalitis and Wesselbron viruses 10, the flaviviruses were not known to be teratogenic; however, there is now compelling evidence that Zika virus (ZIKV) is teratogenic 11 , 12. Therefore, the existing body of work on rubella may provide powerful insights into core directions for global control efforts for management of ZIKV. Leveraging the existing knowledge of rubella dynamics to understand ZIKV burden and risk requires understanding similarities and differences between these two pathogens. In particular, how does the epidemiology of rubella map onto what is known of the dynamics of ZIKV? Two broad scenarios are possible for ZIKV: i) epidemic transmission between humans via the bite of a mosquito (as for dengue virus); or ii) a predominantly zoonotic cycle involving an animal reservoir and a mosquito host, where humans are incidental dead-end hosts (as for West Nile virus [WNV]). At the present time, it is unknown whether human-mosquito-human transmission can be sustained, or whether a sylvatic transmission cycle will contribute to outbreaks in the Americas. A sylvatic cycle has received relatively less attention in the Americas but has been considered13 . Low levels of viremia detected in humans, which might indicate low probability of onward transmission (further detailed below); combined with historical suggestions that ZIKV persistence across a variety of settings might require hosts beyond the primate (human and monkey) hosts characterized 14; and highly variable prevalence across sero-surveys in human populations across Africa15 suggestive of erratic extinction-reintroduction dynamics all suggest that this possibility should not be discounted before more data becomes available. Sexual, blood transfusion, or similar transmission routes have also been reported, but their relevance for similar pathogens (e.g., WNV) suggests a relatively minor contribution to dynamics, although this may change as we obtain more information. Natural wild-type ZIKV infection is also likely to induce life-long homotypic immunity, as observed for infections by other flaviviruses such as yellow fever22 , which is also the case for infection with rubella virus2; nonetheless, further research into this aspect of the biology via longitudinal serological surveys are an important direction for future research (see below). Here, we develop a simple model to illustrate how current theory on rubella may inform risk and outcomes of control for ZIKV, based on two simple transmission scenarios (epidemic and zoonotic), and interpret what this analysis suggests about core areas for further research, specifically for ZIKV, and in the context of teratogenic infections more broadly. Methods and Results Assuming sterilizing immunity following wild-type infection, for the Epidemic scenario, a simple human-to-human transmission process can approximate the dynamics of vector-transmitted infections16, allowing us to borrow directly from the theory relating to rubella (age structured model described in detail in7; here deployed with monthly age classes up to age 15, and annual thereafter). Much of this theory hinges on the value of R0, or the number of new infections per infectious individuals in a completely immunizing population; R0>1 in our Epidemic scenario, which shapes both the average age of infection, and the impact of control strategies 7. Figure 1A,B indicates basic expectations for an immunizing pathogen in endemic and invasive contexts with an R0 around 3 (based on analyses of the Yap outbreaks17 ; and in Colombia18 ; Figure 1A, 1B). Where the infection is endemic, the average age of infection is low, since much of the older population is protected by immunity; where the infection is invasive, all age ranges are affected, as observed during the invasion dynamics of ZIKV such as occurred in Yap19. Figure 1: Epidemic scenario A) Simulation of an immunising infection (R0 = 3), from invasion to endemicity (±15th year); B) Corresponding age incidence through time (arrows) indicating reduced case burden during childbearing ages (CBA, light red), as population immunity establishes; C) Effect of interventions measured by the ratio of cases occurring during potential CBA (y axis), comparing reductions in R0 (x axis), reflecting of vector control (upper plot: a positive effect in the invasion year, but the classical paradoxical effect in endemic years, as for rubella); and the effect of a single vaccination campaign occurring in the first outbreak year, but at increasing delays (x axis), targeting ages up to 30, and with low coverage (lower plot: a consistently beneficial effect, especially for interventions in the first year; since our focus is 30 years in the future, many women will have been protected by that single campaign, despite low coverage). Zoonotic scenario D) Age incidence assuming that following pathogen emergence, 20% of the susceptible population is at risk of being infected in each time-step, shown across time-steps 1, 5, and 9; and E) Expected average age of infection (y axis) in the endemic state as a function of the rate at which susceptible individuals become infected (x axis). Control efforts such as vector control are liable to reduce the force of infection, and can thus increase the average age of infection in the endemic state. Figure 1: Epidemic scenario A) Simulation of an immunising infection (R0 = 3), from invasion to endemicity (±15th year); B) Corresponding age incidence through time (arrows) indicating reduced case burden during childbearing ages (CBA, light red), as population immunity establishes; C) Effect of interventions measured by the ratio of cases occurring during potential CBA (y axis), comparing reductions in R0 (x axis), reflecting of vector control (upper plot: a positive effect in the invasion year, but the classical paradoxical effect in endemic years, as for rubella); and the effect of a single vaccination campaign occurring in the first outbreak year, but at increasing delays (x axis), targeting ages up to 30, and with low coverage (lower plot: a consistently beneficial effect, especially for interventions in the first year; since our focus is 30 years in the future, many women will have been protected by that single campaign, despite low coverage). Zoonotic scenario D) Age incidence assuming that following pathogen emergence, 20% of the susceptible population is at risk of being infected in each time-step, shown across time-steps 1, 5, and 9; and E) Expected average age of infection (y axis) in the endemic state as a function of the rate at which susceptible individuals become infected (x axis). Control efforts such as vector control are liable to reduce the force of infection, and can thus increase the average age of infection in the endemic state. For the Zoonotic scenario, where humans are likely dead end hosts, R0<1 or even zero for human-vector-human transmission. Consequently, the key driver of the burden of congenital disease is fluctuations in the force of infection on humans resulting from transmission in the zoonotic reservoir. During pathogen invasion, every human individual is susceptible, and high burdens might occur, as observed in Yap 19 . Conversely, where exposure has consistently been high, the average age of infection might be low, and the burden correspondingly low (Figure 1E), although this will depend on the magnitude of transmission into the human population – low transmission to human susceptibles can be associated with high average age. If local extinction and reintroduction are the rule (as for many vector-borne infections) populations may remain vulnerable, with high proportions of the older population being susceptible. As in the epidemic scenario, vector control may affect the average age of infection, and should be evaluated carefully (Figure 1E). The power of vaccination, again, is the opportunity of providing direct protection to those most vulnerable. Discussion Our analysis indicates similarities with core predictions that emerge from the existing body of work on rubella across the two major scenarios for ZIKV transmission. Is ZIKV more like dengue (scenario 1) or WNV (scenario 2)? The rarity of human ZIKV isolates compared to the number of infected individuals (Lanciotti et al.20 failed to make any isolates during the Yap outbreak) suggests a low and/or transient viremia, making onward transmission seem unlikely, suggestive more of a similarity with WNV. This accords with historic evidence of the importance of zoonotic reservoirs, e.g., in many parts of West Africa 15. Conversely, models fitted to more recent outbreaks17 suggest greater similarity with dengue. Overall, there is still much to be done to understand the transmission cycle of ZIKV in the Americas. In both scenarios, however, a rapidly changing burden over the course of the early phase of the invasion, followed by endemic establishment of relatively lower burden of disease is suggested by simple modeling, and analogy with rubella. Quantifying both magnitude and timing of this shifting burden will require a deeper understanding of the underlying magnitude of transmission as well as the size of the susceptible population which is at risk, likely to be a function of both the landscape of immunity, and vector dynamics11. Our model illustrates that for ZIKV, as for rubella, the potential paradoxical effects associated with control efforts could also be observed in both scenarios. Again, the details of the changes in burden following implementation of control will depend on both the core parameters described above, but also quantification of the impact of control efforts. If vaccine development is successful, deployment is likely to be targeted primarily at vulnerable groups (women of childbearing age) potentially followed by men (given evidence for sexual transmission). Such targeting is much less likely to result in the 'paradoxical effect' observed for rubella introduced into routine childhood immunization programs 5. However, over the longer term, some form of routine deployment of a ZIKV vaccine is likely to be considered should such a vaccine be developed (perhaps around age 9 years, as for HPV vaccine) and considerations of age incidence effects as for rubella re-emerge. To conclude, the existing body of research into rubella provides a powerful starting point for thinking about the implications of teratogenic infections like ZIKV, but there are also many clear directions for further work. Overall, the invasion dynamics of teratogenic infections remain under-studied. More specifically, for ZIKV, theory building around existing frameworks for rubella could be deployed to identify core biological features required to characterize both the burden and how it will respond to control efforts, in both the long and short term 21. Sensitivity to core assumptions under varied demographic, epidemiological and control settings will delineate the critical aspects of the biology that must be known to best manage the burden of this infection. The magnitude and impacts of sterilizing immunity, as well as antigenic maps measuring cross-protection with other flaviviruses, and the magnitude of transmission, or R0 7, as well as likely impact of control efforts on transmission are likely to be key variables for robust inference into the outcome of management efforts for ZIKV. Competing Interest Statement The authors have declared that no competing interests exist. Data Availability Statement There is no raw data associated with this paper. The citation 7 was used to construct the models. Acknowledgments We thank Bryan Grenfell for comments on an earlier version. ==== Refs References 1 Gregg, N.M., Congenital cataract following German measles in the mother. Transactions of the Ophthalmological Society of Australia, 1941. 3: p. 31-46. 2 Knox, E.G., Strategy for rubella vaccination. International Journal of Epidemiology, 1980. 9: p. 13–23. 3 Cooper, L.Z., The history and medical consequences of rubella. Reviews of Infectious Diseases, 1985. 7: p. S2-S10. 4 Anderson, R.M. and R.M. May, Vaccination against rubella and measles: qualitative investigations of different policies. Journal of Hygiene of Cambridge, 1983. 90: p. 259-325. 5 Anderson, R.M. and B.T. Grenfell, Quantitative investigations of different vaccination policies for the control of congenital rubella syndrome (CRS) in the United Kingdom. Journal of Hygiene of Cambridge, 1986. 96: p. 305-333. 6 Cutts, F.T. and E. Vynnycky, Modelling the incidence of congenital rubella syndrome in developing countries. International Journal of Epidemiology, 1999. 28: p. 1176-1184. 7 Metcalf, C.J.E., J. Lessler, P. Klepac, F.T. Cutts, and B.T. Grenfell, Minimum levels of coverage needed for rubella vaccination: impact of local demography, seasonality and population heterogeneity. Epidemiology and Infection, 2012. 16: p. 1-12 8 Heesterbeek, H., R.M. Anderson, V. Andreasen, S. Bansal, D. De Angelis, C. Dye, K.T. Eames, W.J. Edmunds, S.D. Frost, and S. Funk, Modeling infectious disease dynamics in the complex landscape of global health. Science, 2015. 347(6227): p. aaa4339. 9 Robertson, S.E., F.T. Cutts, R. Samuel, and J.L. Diaz-Ortega, Control of rubella and congenital rubella syndrome (CRS) in developing countries, part 2: vaccination against rubella. Bulletin of the World Health Organization, 1997. 75: p. 69-80. 10 Hubálek, Z., I. Rudolf, and N. Nowotny, Arboviruses pathogenic for domestic and wild animals. Adv Virus Res, 2014. 89: p. 201-275. 11 Lessler, J., L.H. Chaisson, L.M. Kucirka, Q. Bi, K. Grantz, H. Salje, A.C. Carcelen, C.T. Ott, J.S. Sheffield, N.M. Ferguson, D.A.T. Cummings, C.J.E. Metcalf, and I. Rodriguez-Barraquer, Assessing the global threat from Zika virus. Science, 2016: p. aaf8160. 12 Cauchemez, S., M. Besnard, P. Bompard, T. Dub, P. Guillemette-Artur, D. Eyrolle-Guignot, H. Salje, M.D. Van Kerkhove, V. Abadie, C. Garel, A. Fontanet, and H.-P. Mallet, Association between Zika virus and microcephaly in French Polynesia, 2013–15: a retrospective study. Lancet, 2016. 13 Bueno, M.G., N. Martinez, L. Abdala, C.N.D. dos Santos, and M. Chame, Animals in the Zika virus life cycle: what to expect from megadiverse Latin American countries. bioRxiv, 2016: p. 062034. 14 McCrae, A. and B. Kirya, Yellow fever and Zika virus epizootics and enzootics in Uganda. Transactions of the Royal Society of Tropical Medicine and Hygiene, 1982. 76(4): p. 552-562. 15 Brès, P., Données récentes apportées par les enquêtes sérologiques sur la prévalence des arbovirus en Afrique, avec référence spéciale à la fièvre jaune. Bulletin of the World Health Organization, 1970. 43(2): p. 223. 16 Reich, N.G., S. Shrestha, A.A. King, P. Rohani, J. Lessler, S. Kalayanarooj, I.-K. Yoon, R.V. Gibbons, D.S. Burke, and D.A. Cummings, Interactions between serotypes of dengue highlight epidemiological impact of cross-immunity. Journal of The Royal Society Interface, 2013. 10(86): p. 20130414. 17 Kucharski, A.J., S. Funk, R.M. Eggo, H.-P. Mallet, W.J. Edmunds, and E.J. Nilles, Transmission dynamics of Zika virus in island populations: a modelling analysis of the 2013–14 French Polynesia outbreak. PLoS Negl Trop Dis, 2016. 10(5): p. e0004726. 18 Nishiura, H., K. Mizumoto, W.E. Villamil-Gómez, and A.J. Rodríguez-Morales, Preliminary estimation of the basic reproduction number of Zika virus infection during Colombia epidemic, 2015–2016. Travel Medicine and Infectious Disease, 2016. 19 Duffy, M.R., T.-H. Chen, W.T. Hancock, A.M. Powers, J.L. Kool, R.S. Lanciotti, M. Pretrick, M. Marfel, S. Holzbauer, and C. Dubray, Zika virus outbreak on Yap Island, federated states of Micronesia. New England Journal of Medicine, 2009. 360(24): p. 2536-2543. 20 Lanciotti, R.S., O.L. Kosoy, J.J. Laven, J.O. Velez, A.J. Lambert, A.J. Johnson, S.M. Stanfield, and M.R. Duffy, Genetic and serologic properties of Zika virus associated with an epidemic, Yap State, Micronesia, 2007. Emerg Infect Dis, 2008. 14(8): p. 1232-9. 21 Ferguson, N.M., Z.M. Cucunubá, I. Dorigatti, G.L. Nedjati-Gilani, C.A. Donnelly, M.-G. Basáñez, P. Nouvellet, and J. Lessler, Countering the Zika epidemic in Latin America. Science, 2016. 353(6297): p. 353-354. 22 Sawyer, W., The Persistence of Yellow Fever Immunity. Journal of Preventive Medicine, 1931. 5(6): p. 413-428.

PMC005xxxxxx/PMC5001836.txt

==== Front eLifeElifeeLifeeLifeeLife2050-084XeLife Sciences Publications, Ltd 275281931835710.7554/eLife.18357BiochemistryShort ReportThe aspartyl protease DDI2 activates Nrf1 to compensate for proteasome dysfunction Koizumi Shun 1Irie Taro 1Hirayama Shoshiro 1Sakurai Yasuyuki 1Yashiroda Hideki 1Naguro Isao 2Ichijo Hidenori 2Hamazaki Jun 1Murata Shigeo http://orcid.org/0000-0002-3177-35031*1 Laboratory of Protein Metabolism, Graduate School of Pharmaceutical Sciences, The University of Tokyo, Tokyo, Japan2 Laboratory of Cell Signaling, Graduate School of Pharmaceutical Sciences, The University of Tokyo, Tokyo, JapanDikic Ivan Reviewing editorGoethe University Medical School, Germanysmurata@mol.f.u-tokyo.ac.jp16 8 2016 2016 5 e1835731 5 2016 12 8 2016 © 2016, Koizumi et al2016Koizumi et alThis article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.In response to proteasome dysfunction, mammalian cells upregulate proteasome gene expression by activating Nrf1. Nrf1 is an endoplasmic reticulum-resident transcription factor that is continually retrotranslocated and degraded by the proteasome. Upon proteasome inhibition, Nrf1 escapes degradation and is cleaved to become active. However, the processing enzyme for Nrf1 remains obscure. Here we show that the aspartyl protease DNA-damage inducible 1 homolog 2 (DDI2) is required to cleave and activate Nrf1. Deletion of DDI2 reduced the cleaved form of Nrf1 and increased the full-length cytosolic form of Nrf1, resulting in poor upregulation of proteasomes in response to proteasome inhibition. These defects were restored by adding back wild-type DDI2 but not protease-defective DDI2. Our results provide a clue for blocking compensatory proteasome synthesis to improve cancer therapies targeting proteasomes. DOI: http://dx.doi.org/10.7554/eLife.18357.001 eLife digest The proteasome is a machine that destroys unnecessary or damaged proteins inside cells. This role of the proteasome is essential for cell survival, and so when the proteasome is inhibited, cells produce new proteasomes to compensate. Upon proteasome inhibition, a protein called Nrf1 is activated and executes this “bounce-back” response. Some cancer treatments aim to kill cancer cells by inhibiting proteasomes, but these treatments may be unsuccessful if the bounce-back response is not also prevented. Therefore, understanding how Nrf1 is activated is an important issue. Nrf1 is produced at a structure called the endoplasmic reticulum in cells and is continually destroyed by the proteasome. On the other hand, when proteasomes are inhibited, Nrf1 accumulates and is cleaved into an active form, which moves to the cell nucleus to start producing proteasomes. However, it was not known which molecule cleaves Nrf1. Koizumi et al. set out to discover this molecule by screening the genetic material of human cells, and identified a gene that encodes a protease (an enzyme that cleaves other proteins) called DDI2. The loss of DDI2 from cells prevented Nrf1 from being cleaved and entering the nucleus, resulting in low levels of proteasome production. Further experiments showed that a mutant form of DDI2 that lacked protease activity was unable to cleave Nrf1, confirming DDI2’s role in activating Nrf1. Deleting DDI2 from cells does not completely prevent the cleavage of Nrf1, and so some other cleaving enzyme might exist; the identity of this enzyme remains to be discovered. Future work is also needed to establish exactly how DDI2 cleaves Nrf1. This could help to develop a DDI2 inhibitor for cancer treatment that could be used in combination with existing proteasome inhibitors. DOI: http://dx.doi.org/10.7554/eLife.18357.002 Author Keywords proteasomeNrf1proteasetranscriptionResearch Organism Humanhttp://dx.doi.org/10.13039/501100001691Japan Society for the Promotion of Science25221102Murata Shigeo http://dx.doi.org/10.13039/501100001691Japan Society for the Promotion of Science26000014Murata Shigeo http://dx.doi.org/10.13039/100007449Takeda Science FoundationMurata Shigeo The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.elife-xml-version2.5Author impact statementPeptidase activity of DDI2 is required to activate Nrf1 in order to enable proteasome recovery in response to proteasome inhibition. ==== Body Introduction Proteasome inhibition elicits a response to restore proteasome activity, or a 'bounce-back response,' where Nrf1 is the responsible transcription factor that upregulates expression of all proteasome subunit genes in a concerted manner in human cells (Radhakrishnan et al., 2010; Steffen et al., 2010). Proteasome inhibitors such as bortezomib and carfilzomib have been in clinical use for treatment of cancers, especially multiple myeloma, but this bounce-back response attenuates the ability of proteasome inhibitors to kill cancer cells (Radhakrishnan et al., 2010). Therefore, genes regulating Nrf1 activation could be useful drug targets for increasing efficacy of proteasome inhibition in cancer treatment. When Nrf1 is produced, the bulk of the polypeptide is inserted into the ER lumen and glycosylated, with a short cytosolic N-terminus followed by a single transmembrane domain (Radhakrishnan et al., 2014; Zhang et al., 2007). The luminal region of Nrf1 is continually retrotranslocated to the cytosol by the p97/VCP ATPase complex, accompanied by deglycosylation and ubiquitination. Under normal circumstances, Nrf1 is promptly degraded by the proteasome. In contrast, when proteasome activity is compromised, Nrf1 escapes degradation and is proteolytically cleaved to the active form which enters the nucleus and enhances expression of target genes including proteasome subunits (Radhakrishnan et al., 2014, 2010; Sha and Goldberg, 2014; Steffen et al., 2010). However, the processing enzyme for Nrf1 remains obscure. Results and discussion To identify genes important for Nrf1 activation, we performed a genome-wide small interfering RNA (siRNA) screen (Figure 1—figure supplement 1A). Our approach used the well-characterized subcellular localization of Nrf1 accumulation in the nucleus in response to proteasome inhibition. HEK293A cells were transfected with pooled siRNA (a pool of 4 unique siRNAs per gene) and then treated with the proteasome inhibitor bortezomib to induce accumulation and nuclear translocation of Nrf1 (Steffen et al., 2010). Cells were then fixed and stained with anti-Nrf1 antibody. The ratio of the nuclear to cytoplasmic fluorescent intensities was assessed by high-content microscopy and automated image analysis (Figure 1—figure supplement 1B). p97 siRNA treatment served as a positive control, which abolished Nrf1 translocation following bortezomib treatment while increasing cytoplasmic Nrf1 (Figure 1A) (Radhakrishnan et al., 2014). We observed a high degree of assay robustness (Z’-factor > 0.5, Figure 1—figure supplement 1C) in the primary screen. The initial candidate genes with B score < –3.2 (Figure 1B) or which were picked up by visual inspection of the raw image data were further tested using four individual siRNAs in two different cell lines (HEK293A and HT1080 cells) (Figure 1C). The subsequent candidates that had more than two hits in either cell line were finally examined whether the siRNAs mitigated upregulation of PSMA3, a proteasome subunit gene. Consequently, we obtained 14 candidate genes that may impair activation of Nrf1 in response to bortezomib treatment (Figure 1D). These hits included SEL1L, a co-factor of the ubiquitin ligase HRD1, which catalyzes ER-associated degradation (ERAD) of Nrf1 (Iida et al., 2011; Steffen et al., 2010; Tsuchiya et al., 2011) and FAF2/UBXD8, a p97-recruiting molecule in ERAD (Meyer et al., 2012), validating our screening approach (Figure 1A,B, and D).10.7554/eLife.18357.003Figure 1. A genome-wide siRNA screen for regulators of Nrf1 translocation to the nucleus in response to proteasome inhibition. (A) Representative images of Nrf1 localization in control cells (no siRNA) and cells transfected with siRNA targeting p97 or SEL1L in the primary screen. Yellow-boxed regions are magnified and displayed in the right panels. (B) B score of all samples in the primary screen. Data are ordered from lowest to highest. Dashed blue line represents a cutoff value for positive hits. Some of the representative final hits are shown as red dots. The list of B scores for all samples in the primary screen are available in the Figure 1—source data 1. (C) Workflow and summary of the genome-wide siRNA screen. (D) List of the 14 final hit genes and the score in each assay throughout the screening process. DOI: http://dx.doi.org/10.7554/eLife.18357.003 10.7554/eLife.18357.004Figure 1—source data 1. List of B-score in the primary screen. DOI: http://dx.doi.org/10.7554/eLife.18357.004 10.7554/eLife.18357.005Figure 1—figure supplement 1. Methods for the genome-wide screen. (A) A schematic view of the workflow in the primary screen using HEK293A cells. (B) Definition of nucleus and cytoplasm in the screen. Region of interest (ROI) of the nucleus was defined as a circle drawn two pixels inside of the outermost DAPI signal. ROI of the cytoplasm was defined as a three-pixel wide ring around DAPI signal. No gap between the rings and DAPI signals was set in the analysis using HEK293A cells, whereas a one-pixel gap was set in the analysis using HT1080 cells. (C) The well score of negative (no siRNA) and positive (si p97) controls. The well score is the fluorescence intensity ratio of ROI of nucleus to cytoplasm as defined in (B). The Z’ factor was calculated from 12 individual wells on each plate. DOI: http://dx.doi.org/10.7554/eLife.18357.005 Among the final hit genes, we focused on DDI2, because it has a typical retroviral aspartyl protease domain, and therefore is a candidate Nrf1 processing enzyme (Krylov and Koonin, 2001). In negative and positive (p97 siRNA) control cells treated with bortezomib, the majority of Nrf1 is localized in the nucleus and the cytoplasm, respectively (Figure 2A). DDI2 knockdown partially inhibited nuclear translocation of Nrf1, accompanied by an increase in cytoplasmic Nrf1. To determine if there is a defect in Nrf1 processing by DDI2 knockdown, we examined which Nrf1 species were observed in each knockdown. In negative control cells, Nrf1 was hardly detected in the absence of bortezomib, but bortezomib treatment increased the processed, active form of Nrf1 as well as the full-length, cytosolic form that is retrotranslocated into the cytosol by p97 (Figure 2B) (Radhakrishnan et al., 2014). In p97 knockdown cells, the luminal ER form of Nrf1 that is N-glycosylated accumulated while the processed form almost disappeared both in the presence and absence of bortezomib. In DDI2 knockdown cells, Nrf1 was not detected in the absence of bortezomib, similar to control cells. However, in the presence of bortezomib, the full-length, cytosolic form of Nrf1 markedly accumulated (Figure 2B). These results indicate that DDI2 is involved in conversion of full-length, cytosolic Nrf1 to the processed, active form.10.7554/eLife.18357.006Figure 2. DDI2 is involved in Nrf1 processing and translocation to the nucleus. (A) Representative images of Nrf1 localization. HEK293A cells were transfected with a non-targeting control (negative control), DDI2, or p97 siRNA and then treated with 50 nM bortezomib for 14 hr before fixation. (B) Immunoblotting of whole-cell lysates of cells in (A) treated with or without bortezomib. Nrf1 is detected as three different forms; a glycosylated form (G), full-length form (FL), and processed form (P). (C) Immunoblotting of Nrf1 after deglycosylation treatment. HEK293A cells were transfected with DDI2 or p97 siRNA, followed by transfection with Nrf1-3×Flag, and then treated with or without 50 nM bortezomib. The cell lysates were treated with or without Endo H. DeG denotes deglycosylated Nrf1. DOI: http://dx.doi.org/10.7554/eLife.18357.006 10.7554/eLife.18357.007Figure 2—figure supplement 1. The expression and localization of DDI2 were not affected by bortezomib treatment. (A) Relative mRNA expression of DDI2 in HCT116 cells treated with or without bortezomib (50 nM, 14 hr). The data represent mean + standard error of the mean (SEM) (n = 3, biological replicates). (B) Subcellular localization of DDI2. HEK293A cells stably expressing Venus-Sec61β (ER marker) were transfected with Flag-DDI2 and treated with or without bortezomib. DOI: http://dx.doi.org/10.7554/eLife.18357.007 We further examined the N-glycosylation status of Nrf1 using cells transfected with Nrf1 tagged with 3×Flag at the C-terminus. N-glycosylated Nrf1 accumulated in p97 knockdown cells was sensitive to endoglycosidase H (Endo H) treatment and the deglycosylated form migrated faster in SDS-PAGE, consistent with a previous report (Figure 2C) (Radhakrishnan et al., 2014). In contrast, the full-length form of Nrf1 that is significantly accumulated in DDI2 knockdown cells was not sensitive to Endo H treatment (Figure 2C). Note that bortezomib treatment alone causes some accumulation of Endo H-sensitive N-glycosylated Nrf1 and that the deglycosylated species was detected at almost the same molecular weight as the processed, active form of Nrf1 observed in cells treated with bortezomib alone, similar to previously reported observations (Figure 2C) (Radhakrishnan et al., 2014). These results demonstrate that the form of Nrf1 accumulated in DDI2-depleted cells is not N-glycosylated, further supporting the role of DDI2 in the processing of Nrf1 rather than in deglycosylation or retrotranslocation. The X-ray crystal structure analysis of the retroviral aspartyl protease (RVP) domain of budding yeast Ddi1p has revealed that it is a dimer with a similar fold to that of the human immunodeficiency virus type 1 (HIV-1) protease, with identical geometry of the double D[S/T]GA motif of the active site (Sirkis et al., 2006). The HIV-1 protease typically cleaves substrates between two hydrophobic residues (Konvalinka et al., 2015). Nrf1 has been shown to be cleaved between Trp103 and Leu104 to become active (Radhakrishnan et al., 2014), which conforms with the cleavage motif by retroviral aspartyl proteases. Accordingly, we asked whether the protease activity of DDI2 is required for Nrf1 processing. DDI2 has a ubiquitin-like domain (UBL) at the N-terminus and a RVP domain near the C-terminus (Figure 3A). Bortezomib treatment increased the processed form of Nrf1 (Figure 3B). Knockdown of DDI2 reduced the processed form and increased full-length Nrf1 (Figure 3B). This effect was rescued by introducing siRNA-resistant wild-type DDI2 but not a protease-dead DDI2 in which the active site aspartic acid 252 was replaced with asparagine (D252N). We also found that a DDI2 mutant lacking the UBL domain only partially restored the effect of DDI2 knockdown (Figure 3B). These results suggest that the protease activity of DDI2 is required for cleavage of Nrf1 and that the UBL domain plays some role in the cleavage.10.7554/eLife.18357.008Figure 3. The protease activity of DDI2 is necessary for Nrf1 processing and its transcriptional activity. (A) Schematic diagram of wild-type (WT) and each mutant of DDI2. Ubiquitin-like (UBL) domain and retroviral protease-like (RVP) domain are represented as filled rectangles. The putative aspartyl protease active site amino acid sequence is shown. (B) HEK293A cells were transfected with DDI2 siRNA and after 24 hr were transfected with a plasmid encoding WT or mutant DDI2 shown in (A), followed by 50 nM bortezomib treatment for 14 hr before harvest. The signal intensity ratio of Nrf1 full-length form (FL) to the processed form (P) was calculated, where the ratio for bortezomib treatment alone was set as 1. (C) Immunoblotting of whole cell lysates of DDI2 WT knock-in (KI), DDI2 knockout (KO), and DDI2 D252N KI HCT116 cells. The cells transfected with Nrf1-3×Flag were treated with or without 50 nM borteaomib. (D) Relative mRNA expression of the proteasome genes PSMA3 and PSMB5 in WT, DDI2 KO, DDI2 WT KI, and DDI2 D252N KI HCT116 cells. mRNA levels of target genes were normalized by GUSB mRNA levels. The data represent mean + standard error of the mean (SEM) (n = 3, biological replicates). Statistical comparison was made by Tukey’s test (*p<0.05). (E) Proteasome peptidase activity of cell lysates of the indicated cell lines. The data represent mean + SEM (n = 3, biological replicates). Statistical comparison was made by Tukey’s test (**p<0.01). DOI: http://dx.doi.org/10.7554/eLife.18357.008 10.7554/eLife.18357.009Figure 3—figure supplement 1. Genome editing of DDI2 locus by CRISPR-Cas9 system. (A) Schematic diagram of DDI2 wild-type (WT), knockout (KO), and knock-in (KI) alleles. Red arrowhead indicates sgRNA targeting site, and blue arrows indicate PCR primers for confirmation of homologous recombination. puroR indicates a puromycin resistant cassette. (B) Confirmation of successful genome editing at the DDI2 locus. The genome DNA of each cell line was extracted and amplified by PCR using primer pairs shown in (A). DOI: http://dx.doi.org/10.7554/eLife.18357.009 To confirm the necessity of the protease activity of DDI2, we generated DDI2 knockout (KO) and protease-dead DDI2 (D252N) knock-in (KI) cells as well as wild-type DDI2 knock-in cells (Figure 3, Figure 3—figure supplement 1A,B). In DDI2 knockout cells and DDI2 D252N knock-in cells, the full-length form of Nrf1 was accumulated upon bortezomib treatment, whereas the processed form was accumulated in wild-type DDI2 knock-in cells (Figure 3C). These results further support the requirement of DDI2 protease activity for Nrf1 activation. We then examined whether a lack of the catalytic activity of DDI2 abolishes the 'bounce-back' response after proteasome inhibition. In parental HCT116 cells, bortezomib treatment caused an increase in mRNA levels of the proteasome subunit genes PSMA3 and PSMB5 (Figure 3D). Knockout of DDI2 strongly suppressed this response, further supporting the importance of DDI2 in Nrf1 activation (Figure 3D). Interestingly, the basal expression of proteasome subunits was also decreased in DDI2-deficient cells. In wild-type DDI2 knock-in cells, mRNA levels of the proteasome subunits were upregulated in response to bortezomib, similar to the parental cells. In contrast, DDI2 D252N knock-in cells did not undergo such a response, similar to DDI2 knockout cells (Figure 3D). These results suggest that the processing of Nrf1 by the aspartyl protease activity of DDI2 is required for upregulation of proteasome gene expression mediated by Nrf1 in response to proteasome inhibition. Nrf1 has also been found to regulate basal expression of proteasome subunits, the extent of which varies between cell types (Lee et al., 2013, 2011). We observed that knockout and D252N DDI2 knock-in cells had significantly lower proteasome activity compared to wild-type DDI2 knock-in cells, suggesting that DDI2 is also involved in basal expression of proteasomes through its catalytic activity (Figure 3E). In conclusion, we identified DDI2 as a protease that is required for Nrf1 processing and the bounce-back response induced by proteasome inhibition. However, there remain several questions to be answered. How can the involvement of DDI2 be reconciled with a previous report that demonstrated a defect in Nrf1 processing by strong inhibition of the proteasome, leading to the conclusion that the proteasome is the processing enzyme for Nrf1 (Sha and Goldberg, 2014)? In terms of substrate specificity, the cleavage site of Nrf1 (P1: W, P1’: L) does not seem to be a sequence preferred by the proteasome (Toes et al., 2001); rather it conforms to a cleavage motif of RVP (Konvalinka et al., 2015). It could be that the proteasome activity is required for function of DDI2 or some other factors that is involved in Nrf1 processing. Related to this, the mechanism by which DDI2 acts as a Nrf1 processing protease remains unclear. DDI2 is not induced by bortezomib at either protein or mRNA level (Figure 2B and Figure 2—figure supplement 1A). Furthermore, the subcellular localization of DDI2 seems to be unaffected by bortezomib treatment (Figure 2—figure supplement 1B). Since DDI2 is suggested to be active even when the proteasome activity is not compromised (Figure 3D and E), a specific activation mechanism under proteasome impairment may not exist. An intriguing finding is that the UBL domain of DDI2 plays some role in Nrf1 processing (Figure 3B). It has been shown that the UBL domain of Ddi1p is an atypical UBL that binds ubiquitin (Nowicka et al., 2015). Binding of DDI2 with ubiquitinated proteins, possibly Nrf1 itself, would be promoted by proteasome inhibition and may facilitate Nrf1 processing by DDI2. Lastly, whether DDI2 directly cleaves Nrf1 remains unknown. We have tested a recombinant fragment of Nrf1 encompassing the processing site as a substrate for recombinant DDI2, but failed to detect its cleavage. Other factors might be required for in vitro reconstitution of Nrf1 processing by DDI2, such as substrate unfolding, co-activators of DDI2, and a set of specific experimental conditions. Understanding the mechanism by which DDI2 cleaves Nrf1 and establishing an in vitro assay for the enzymatic activity of DDI2 should provide useful information for developing a DDI2 inhibitor that would block compensatory proteasome synthesis to improve cancer therapies targeting proteasomes. Materials and methods Genome-wide siRNA screening In the primary screen, Dharmacon siGENOME SMARTpool siRNA library (GE Dharmacon, Lafayette, CO) was used. To prepare screening plates, the siRNAs in each well were suspended in 1 × siRNA buffer (Thermo Fisher Scientific, Waltham, MA) and 2.5 pmol siRNA (2.5 μL/well) was dispensed into black, clear bottom, 384-well plates (Greiner, Kremsmünster, Austria). For each well, a mixture of 10 μL DMEM and 0.1 μL Lipofectamine RNAiMAX (Invitrogen, Carlsbad, CA) was added. After 40 min incubation, 2000 cells/well of HEK293A cells were seeded. After 48 hr culture, bortezomib was added into each well to a final concentration of 10 nM. Cells were fixed with 4% PFA after 12 hr bortezomib treatment. Cells were then stained with Nrf1 antibody (sc-13031; Santa Cruz Biotechnology, Dallas, TX) and DAPI, and the fluorescent images were acquired and analyzed by CellInsight High Content Screening Platform (Thermo Fisher Scientific). The fluorescence signal ratio of the nucleus to the cytoplasm was used as a raw measured value. The value was fitted in a two-way median polish method to exclude positional effects in the 384-well plate, and then the B score was calculated on a per-plate basis using the following formula. B score = (Xi − Median) / MAD (Xi: measured value, MAD: median of absolute deviation) In the secondary screen, four individual siRNAs contained in the library were purchased from Dharmacon and used. HEK293A and HT1080 cells transfected with each siRNA were analyzed by the same method as in the primary screening. In the third screen, HEK293A cells were treated with each hit siRNA and the expression level of the proteasome gene PSMA3 was measured by quantitative RT-PCR. DNA constructs Human DDI2 cDNA was synthesized from total RNA extracted from HT1080 cells using the indicated primers. Forward: 5’-ATGCTGCTCACCGTGTACTGTGTGC-3’, Reverse: 5’-TCATGGCTTCTGACGCTCTGCATCC-3’. DDI2 UBL deletion mutant was synthesized using 5’-AACTTACCCCGAATAGATTTCAG-3’ for a forward primer. siRNA resistant mutations were introduced without changing amino acid sequence using the following primers. Forward: 5’- TAATGTTGTATATTAACTGCAAAGTGAATGGACATCCTG-3’, Reverse: 5’- CGACCTGTCCAAAACTTTCCGGAGCCTCTTCCATAGC-3’. Human Nrf1 cDNA was synthesized from total RNA extracted from HEK293A cells using the indicated primers. Forward: 5’-ATGCTTTCTCTGAAGAAATACTTAACG-3’, Reverse: 5’-TCACTTTCTCCGGTCCTTTGG-3’. PCR was performed using PrimeSTAR Max DNA polymerase (Takara Bio, Shiga, Japan). Amplified fragments were subcloned into pIRES vector (Clontech Laboratories, Mountain View, CA) and all constructs were confirmed by sequencing. Cell culture and transfection HEK293A cells were purchased from Thermo Fisher Scientific. HCT116 cells were obtained from RIKEN BRC. The cell lines were tested negative for mycoplasma contamination by DAPI staining. The authors performed no further authentication of the cell lines. HEK293A cells and HCT116 cells were cultured under standard conditions. cDNAs were transfected into cells using PEI-MAX (Mw: 40,000). siRNAs targeting DDI2 or p97 and siGENOME Non-Targeting siRNA #2 were purchased from GE Dharmacon. siRNA was transfected into cells with Lipofectamine RNAi MAX (Invitrogen). The sequences of siRNAs targeting DDI2 and p97 were as follows: DDI2, 5’-GCCAAGUAGUGAUGCUUUA-3’; p97, 5’-GUAAUCUCUUCGAGGUAUA-3’. Establishment of DDI2 knockout and knock-in cell lines The cell lines were established using the CRISPR/Cas9 system. Single guide RNAs (sgRNA) were designed using CRISPR direct (http://crispr.dbcls.jp/) and cloned into a pX330 vector. The sgRNA sequence for DDI2 was 5’-ACTCGAGCTCGCACAGCGCG-3’. Targeting constructs for gene knockout were designed to insert a puromycin resistance cassette at the locus of the start codon. Targeting constructs for DDI2 knock-in were designed to insert DDI2 wild type or D252N cDNA in-frame downstream of DDI2 exon 1. A puromycin resistance cassette was also inserted into this region. The sgRNA vector and targeting vector were transfected in HCT116 cells. After 48 hr transfection, cells were cultured in medium supplemented with 4 μg/mL puromycin. After two weeks drug selection, colonies were picked up and successful homologous recombination was confirmed by PCR method. PCR was performed using EmeraldAmp PCR Master Mix (Takara Bio). The following primers were used for confirmation of genome editing. DDI2 Forward: 5’-ATGCTGCTCACCGTGTACTGTGTGC-3’, DDI2 intron 1 Reverse: 5’-GCAAGCTGAGTAGGGAAATGAAACCACCAA-3’, Puro forward: 5’-GTCACCGAGCTGCAAGAACTCTTCC-3’. Quantitative RT-PCR Cells were harvested 12 hr after 20 nM bortezomib treatment. Total RNA of cells was isolated using High Pure RNA isolation kit (Roche, Basel, Switzerland) and were reverse-transcribed using ReverTra Ace qPCR RT kit (Toyobo, Osaka, Japan). Quantitative RT-PCR was performed using THUNDERBIRD Probe qPCR Mix (Toyobo), Universal ProbeLibrary Probe (Roche), and LightCycler 480 (Roche). The sequences of primers used were as follows: PSMA3, 5’-GAAGAAGCAGAGAAATATGCTAAGG-3’ and 5’-GGCTAAATAGTTACATTGGACTGGAG-3’; PSMB5, 5’-CATGGGCACCATGATCTGT-3’ and 5’-GAAATCCGGTTCCCTTCACT-3’; GUSB, 5’-CGCCCTGCCTATCTGTATTC-3’ and 5’-TCCCCACAGGGAGTGTGTAG-3’. Immunoblotting 24 hr after transfection of siRNA, cells were transfected with cDNA and cultured for a further 48 hr. 50 nM bortezomib was added 14 hr prior to cell lysis. Cells were lysed in buffer containing 42 mM Tris-HCl (pH 6.8), 1.72% SDS, 5.6% glycerol, 5% 2-mercaptoethanol, and 0.01% bromophenol blue (SDS sample buffer) for whole-cell lysate. The samples were subjected to SDS-PAGE, transferred to polyvinylidene fluoride membrane, and analyzed by immunoblotting. All images were taken using Fusion SL4 (M&S Instruments). Rabbit polyclonal antibody against DDI2 was raised by immunizing keyhole limpet hemocyanin (KLH) conjugated synthetic DDI2 C-terminal (residues 385–399) peptides. The following antibodies were purchased: Nrf1 (sc-13031; Santa Cruz), p97 (MA3-004; Invitrogen), GAPDH (sc-32233; Santa Cruz), Flag (F1804; Sigma Aldrich, St. Louis, MO). Immunostaining Cells were fixed in 4% paraformaldehyde 72 hr after transfection of siRNA and 16 hr after 10 nM bortezomib treatments. The cells were incubated with primary antibodies, and then incubated with DAPI (Nacalai Tesque, Kyoto, Japan) and secondary antibodies, either Goat anti-rabbit or anti-mouse IgG secondary antibody Alexa Fluor 488 or Alexa Fluor 647 conjugate (Invitrogen). All images were acquired by TCS SP5 or TCS SP8 (Leica Microsystems, Wetzlar, Germany). Proteasome activity measurement Cells were lysed in ice-cold buffer containing 25 mM Tris-HCl (pH 7.5), 0.2% Nonidet P-40, 1 mM dithiothreitol, 2 mM ATP, and 5 mM MgCl2. The hydrolysis of the fluorogenic peptide, succinyl-Leu-Leu-Val-Tyr-7-amino-4-methylcoumarin (Suc-LLVY-MCA) (Peptide Institute, Osaka, Japan) was measured in 50 mM Tris-HCl (pH 8.0) at 37°C by ARVO MX 1420 (PerkinElmer, Waltham, MA). Deglycosylation assay Cells were lysed in ice-cold phosphate buffered saline (PBS) containing 0.5% Triton X-100. After centrifugation (20,000 g, 10 min), the cell lysates were subjected to deglycosylation reactions with Endo Hf (New England BioLabs, Ipswich, MA) following the manufacturer’s protocol. Statistical analysis A biological replicate was considered as each independent experiment. Each different clone of the same genotypes was also considered as a biological replicate in the experiments using mutant cell lines obtained by the CRISPR-Cas9 system. Technical replicates were multiple analyses of the same sample in an experiment. The results are expressed as mean + standard error of the mean (SEM) of three biological replicates (n = 3). Significant differences were considered as probabilities less than 5% (p<0.05). Funding Information This paper was supported by the following grants: http://dx.doi.org/10.13039/501100001691Japan Society for the Promotion of Science 25221102 to Shigeo Murata. http://dx.doi.org/10.13039/501100001691Japan Society for the Promotion of Science 26000014 to Shigeo Murata. http://dx.doi.org/10.13039/100007449Takeda Science Foundation to Shigeo Murata. Acknowledgements We are grateful to the members of the Murata laboratory for helpful advice and discussion and the members of the Ichijo laboratory for experimental support. This work was funded by JSPS grants 25221102 and 26000014 to SM. Additional information Competing interests The authors declare that no competing interests exist. Author contributions SK, Conception and design, Acquisition of data, Analysis and interpretation of data. TI, Conception and design, Acquisition of data, Analysis and interpretation of data. SH, Acquisition of data, Analysis and interpretation of data. YS, Acquisition of data, Analysis and interpretation of data. HY, Acquisition of data, Analysis and interpretation of data. IN, Acquisition of data, Analysis and interpretation of data. HI, Acquisition of data, Analysis and interpretation of data. JH, Conception and design, Acquisition of data, Analysis and interpretation of data. SM, Conception and design, Analysis and interpretation of data, Drafting or revising the article. 10.7554/eLife.18357.010Decision letter Dikic Ivan Reviewing editorGoethe University Medical School, GermanyIn the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included. Thank you for submitting your article "The aspartyl protease DDI2 activates Nrf1 to compensate for proteasome dysfunction" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by Ivan Dikic as the Senior and Reviewing Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Raymond J Deshaies (Reviewer #1) and Anne Bertolotti (Reviewer #2). The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission. Summary: The manuscript by Koizumi et al. identifies DDI2 as a candidate enzyme that mediates the cleavage and activation of the transcription factor Nrf1 (NFE2L1). The manuscript starts with an unbiased genome-wide siRNA screen aimed at identifying genes important for Nrf1 activation. This yielded 14 candidate genes, including DDI2, which is known to have an aspartyl protease domain homologous to that of the HIV protease. Subsequent experiments demonstrated diminished nuclear localization and processing of Nrf1, and diminished induction of proteasome genes, in DDI2-deficient cells treated with proteasome inhibitor. It was also shown that the active site of DDI2 is required to promote processing of Nrf1 and bortezomib-induced expression of proteasome genes. The authors also showed that DDI2 can be targeted by inhibitors of HIV protease and that these inhibitors improve the efficacy of Bortezomib. The findings presented in this manuscript are technically convincing and targeting the bounce-back response upon proteasome inhibition is indeed of importance at several levels: to improve our knowledge on the regulation of proteasome upregulation and to improve cancer therapies based on proteasome inhibition. Essential revisions: 1) Did the authors attempt a protease protection experiment to examine the disposition of Nrf1 in DDI2-depleted cells? Based on the data Figure 2B, the authors conclude that Nrf1 has been retrotranslocated and deglycosylated in DDI2-depleted cells, but this is not formally proven. It is possible that DDI2 regulates Nrf1 in some other manner that is upstream of the processing step (e.g. retrotranslocation), and the MW shift occurs for some other reason other than deglycosylation. While this may be unlikely, it is certainly possible and should be commented. 2) Figure 4 is weak, and all reviewers felt that needs to be significantly improved or removed. This is not a point that is strictly necessary for the central finding of the paper that DDI2 is required for processing. The real question is whether it can be improved by additional data (see below) or should be removed from the current manuscript. For example, the effects in panel A are weak and would benefit from a titration to higher concentrations. Panel B is even weaker. Other groups have reported that there is some synthetic effect of combining proteasome inhibitors with HIV protease inhibitors (e.g. the Kraus and Driessen papers cited in the Results and Discussion). So, what is new about this figure? The big question is, what is the underlying mechanism? Is it really due to inhibition of DDI2? A couple of questions related to this experiment immediately come to mind. Are DDI2∆ cells, like NRF1∆ cells, more sensitive to bortezomib? Is the synthetic effect of combining bortezomib and NFV or LPV still observed in DDI2∆ cells? Is it observed in NRF1∆ cells? So we recommend to the authors to decide one of these two options related to Figure 4. 3) The authors have established DDI2 knock-out cells (DDI2 KO) and DDI2 KO cells in which the WT or protease-dead mutant DDI2 cDNA was knock-in to the DDI2 locus. It would be useful to analyse the processing of Nrf1 in these cell lines upon proteasome inhibition to complement the siRNA knocked down experiments. 4) The authors should discuss why both forms (FL and P) of Nrf1 are decreasing after co-treatment with bortezomib and lopinavir. Could it be due to cell death or off-target effect of this drug? Is lopinavir treatment alone also decreasing the levels of Nrf1? 10.7554/eLife.18357.011Author response Essential revisions: 1) Did the authors attempt a protease protection experiment to examine the disposition of Nrf1 in DDI2-depleted cells? Based on the data Figure 2B, the authors conclude that Nrf1 has been retrotranslocated and deglycosylated in DDI2-depleted cells, but this is not formally proven. It is possible that DDI2 regulates Nrf1 in some other manner that is upstream of the processing step (e.g. retrotranslocation), and the MW shift occurs for some other reason other than deglycosylation. While this may be unlikely, it is certainly possible and should be commented. We examined whether the Nrf1 species accumulated in DDI2-depleted cells is deglycosylated using endoglycosidase H (Endo H), which removes N-linked glycosylation. In p97-depleted cells, the accumulated Nrf1 band was sensitive to Endo H treatment, consistent with the previous studies and indicating that the two bands correspond to glycosylated “G” and deglycosylated “DeG” forms of Nrf1. In contrast, the Nrf1 form “FL” accumulated in DDI2-depleted cells is not sensitive to Endo H treatment, indicating that this form is not glycosylated and that the “FL” form is a product after retrotranslocation. Curiously, the migration of “DeG” and “FL” does not correspond to each other. This phenomenon was also observed in a previous report, though the precise reason is yet unknown (Radhakrishnan et al. 2014). These results are newly added as Figure 2C and described in the third paragraph of the Results and Discussion in the revised manuscript. We also performed protease protection experiments to examine whether Nrf1 retrotranslocation was inhibited, but we were unable to obtain clear results. However, DDI2-depleted cells show accumulation of deglycosylated Nrf1, in contrast to p97-depleted cells (Figure 2B and C). As deglycosylation of ER-associated degradation substrates are generally completed in the cytoplasm after retrotranslocation from the ER, this result suggests that Nrf1 is retrotranslocated in DDI2-depleted cells. 2) Figure 4 is weak, and all reviewers felt that needs to be significantly improved or removed. This is not a point that is strictly necessary for the central finding of the paper that DDI2 is required for processing. The real question is whether it can be improved by additional data (see below) or should be removed from the current manuscript. For example, the effects in panel A are weak and would benefit from a titration to higher concentrations. Panel B is even weaker. Other groups have reported that there is some synthetic effect of combining proteasome inhibitors with HIV protease inhibitors (e.g. the Kraus and Driessen papers cited in the Results and Discussion). So, what is new about this figure? The big question is, what is the underlying mechanism? Is it really due to inhibition of DDI2? A couple of questions related to this experiment immediately come to mind. Are DDI2∆ cells, like NRF1∆ cells, more sensitive to bortezomib? Is the synthetic effect of combining bortezomib and NFV or LPV still observed in DDI2∆ cells? Is it observed in NRF1∆ cells? So we recommend to the authors to decide one of these two options related to Figure 4. We decided to remove Figure 4 from this manuscript after performing some experiments related to this figure. We examined bortezomib sensitivity of DDI2 knock-in and knock-out cells as well as Nrf1 knock-out cells. Although DDI2 knock-out cells tended to be more sensitive to bortezomib than DDI2 wild-type knock-in cells, the effect was not significant. Also, at least in our hands, Nrf1 knock-out, which has been shown to increase bortezomib sensitivity, did not confer significant sensitivity to bortezomib in HCT116 cells. As it is known that contribution of Nrf1 on proteasome expression depends on cell types, the results might not be generalized. We also examined whether a synergistic effect of bortezomib and LPV or NFV was observed in DDI2 knock-out or Nrf1 knock-out cells. The synergistic effect was slightly decreased but still observed in both cell lines. Thus, the HIV protease inhibitor could inhibit Nrf1 processing via DDI2 as suggested in our previous Figure 4B. However, since the decrease in the synergistic effect was so modest that we cannot conclude that synergistic cytotoxicity in combination treatment with bortezomib is mainly caused by Nrf1 or DDI2. Accordingly, we deleted sentences regarding Figure 4 in the revised manuscript. 3) The authors have established DDI2 knock-out cells (DDI2 KO) and DDI2 KO cells in which the WT or protease-dead mutant DDI2 cDNA was knock-in to the DDI2 locus. It would be useful to analyse the processing of Nrf1 in these cell lines upon proteasome inhibition to complement the siRNA knocked down experiments. We examined Nrf1 processing in DDI2 knockout, DDI2 WT knock-in, and DDI2 D252N knock-in cells, and the result is newly inserted in Figure 3C. Defect of Nrf1 cleavage is observed in DDI2 knockout cells and DDI2 D252N knock-in cells as in Figure 3B, in which the cells were treated with DDI2 siRNA and transfected with D252N cDNA. This results support the idea that DDI2 is a processing enzyme of Nrf1. 4) The authors should discuss why both forms (FL and P) of Nrf1 are decreasing after co-treatment with bortezomib and lopinavir. Could it be due to cell death or off-target effect of this drug? Is lopinavir treatment alone also decreasing the levels of Nrf1? The decrease of Nrf1 after co-treatment with bortezomib and lopinavir is reproducible, but the reason is not clear. Cell death contributed to this result to some extent because co-treatment with bortezomib and 10 μM lopinavir caused cell death. However, Nrf1 was also decreased in combination treatment with bortezomib and 1 μM lopinavir, though cell death was not induced in this condition. This effect was not observed in co-treatment with bortezomib and nelfinavir, suggesting that this was an off-target effect of lopinavir. It is possible that lopiavir itself could decrease the amount of Nrf1. However, it was not confirmed because Nrf1 was hardly detected without proteasome inhibitor. As we decided to delete Figure 4, we did not include discussion regarding this point. ==== Refs References Iida Y Fujimori T Okawa K Nagata K Wada I Hosokawa N 2011 SEL1L protein critically determines the stability of the HRD1-SEL1L endoplasmic reticulum-associated degradation (ERAD) complex to optimize the degradation kinetics of ERAD substrates Journal of Biological Chemistry 286 16929 16939 10.1074/jbc.M110.215871 21454652 Konvalinka J Kräusslich HG Müller B 2015 Retroviral proteases and their roles in virion maturation Virology 479–480 403 417 10.1016/j.virol.2015.03.021 Krylov DM Koonin EV 2001 Correspondence Current Biology 11 R584 R587 10.1016/S0960-9822(01)00357-8 11516960 Lee CS Ho DV Chan JY 2013 Nuclear factor-erythroid 2-related factor 1 regulates expression of proteasome genes in hepatocytes and protects against endoplasmic reticulum stress and steatosis in mice FEBS Journal 280 3609 3620 10.1111/febs.12350 23702335 Lee CS Lee C Hu T Nguyen JM Zhang J Martin MV Vawter MP Huang EJ Chan JY 2011 Loss of nuclear factor E2-related factor 1 in the brain leads to dysregulation of proteasome gene expression and neurodegeneration PNAS 108 8408 8413 10.1073/pnas.1019209108 21536885 Meyer H Bug M Bremer S 2012 Emerging functions of the VCP/p97 AAA-ATPase in the ubiquitin system Nature Cell Biology 14 117 123 10.1038/ncb2407 22298039 Nowicka U Zhang D Walker O Krutauz D Castañeda CA Chaturvedi A Chen TY Reis N Glickman MH Fushman D 2015 DNA-damage-inducible 1 protein (Ddi1) contains an uncharacteristic ubiquitin-like domain that binds ubiquitin Structure 23 542 557 10.1016/j.str.2015.01.010 25703377 Radhakrishnan SK den Besten W Deshaies RJ 2014 p97-dependent retrotranslocation and proteolytic processing govern formation of active Nrf1 upon proteasome inhibition eLife 3 e18357 10.7554/eLife.01856 Radhakrishnan SK Lee CS Young P Beskow A Chan JY Deshaies RJ 2010 Transcription factor Nrf1 mediates the proteasome recovery pathway after proteasome inhibition in mammalian cells Molecular Cell 38 17 28 10.1016/j.molcel.2010.02.029 20385086 Sha Z Goldberg AL 2014 Proteasome-mediated processing of Nrf1 is essential for coordinate induction of all proteasome subunits and p97 Current Biology 24 1573 1583 10.1016/j.cub.2014.06.004 24998528 Sirkis R Gerst JE Fass D 2006 Ddi1, a eukaryotic protein with the retroviral protease fold Journal of Molecular Biology 364 376 387 10.1016/j.jmb.2006.08.086 17010377 Steffen J Seeger M Koch A Krüger E 2010 Proteasomal degradation is transcriptionally controlled by TCF11 via an ERAD-dependent feedback loop Molecular Cell 40 147 158 10.1016/j.molcel.2010.09.012 20932482 Toes RE Nussbaum AK Degermann S Schirle M Emmerich NP Kraft M Laplace C Zwinderman A Dick TP Müller J Schönfisch B Schmid C Fehling HJ Stevanovic S Rammensee HG Schild H 2001 Discrete cleavage motifs of constitutive and immunoproteasomes revealed by quantitative analysis of cleavage products The Journal of Experimental Medicine 194 1 12 10.1084/jem.194.1.1 11435468 Tsuchiya Y Morita T Kim M Iemura S Natsume T Yamamoto M Kobayashi A 2011 Dual regulation of the transcriptional activity of Nrf1 by β-TrCP- and Hrd1-dependent degradation mechanisms Molecular and Cellular Biology 31 4500 4512 10.1128/MCB.05663-11 21911472 Zhang Y Lucocq JM Yamamoto M Hayes JD 2007 The NHB1 (N-terminal homology box 1) sequence in transcription factor Nrf1 is required to anchor it to the endoplasmic reticulum and also to enable its asparagine-glycosylation Biochemical Journal 408 161 172 10.1042/BJ20070761 17705787

PMC005xxxxxx/PMC5001837.txt

==== Front eLifeElifeeLifeeLifeeLife2050-084XeLife Sciences Publications, Ltd 275254871419310.7554/eLife.14193Tools and ResourcesNeuroscienceCellular resolution circuit mapping with temporal-focused excitation of soma-targeted channelrhodopsin Baker Christopher A http://orcid.org/0000-0002-0604-84491*Elyada Yishai M 2†Parra Andres 2Bolton M McLean 1*1 Disorders of Neural Circuit Function, Max Planck Florida Institute for Neuroscience, Jupiter, United States2 Functional Architecture of the Cerebral Cortex, Max Planck Florida Institute for Neuroscience, Jupiter, United StatesHäusser Michael Reviewing editorUniversity College London, United Kingdomchristopher.baker@mpfi.org (CAB);mclean.bolton@mpfi.org (MMcLB)† Department of Neurobiology, Institute of Life Sciences, Edmond and Lily Safra Center for Brain Sciences, Hebrew University of Jerusalem, Jerusalem, Israel. 15 8 2016 2016 5 e1419305 1 2016 14 8 2016 © 2016, Baker et al2016Baker et alThis article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.We describe refinements in optogenetic methods for circuit mapping that enable measurements of functional synaptic connectivity with single-neuron resolution. By expanding a two-photon beam in the imaging plane using the temporal focusing method and restricting channelrhodopsin to the soma and proximal dendrites, we are able to reliably evoke action potentials in individual neurons, verify spike generation with GCaMP6s, and determine the presence or absence of synaptic connections with patch-clamp electrophysiological recording. DOI: http://dx.doi.org/10.7554/eLife.14193.001 eLife digest Nerve cells called neurons carry information around the body in the form of electrical impulses and pass signals to each another to form circuits that link different organs and tissues. Mapping out the neurons in the brain can reveal how different circuits contribute to an animal’s behavior. Yet, because the brains of mammals contain millions of neurons, these circuits are often difficult to untangle. One way to tease apart circuits of neurons uses a technique called optogenetics, which involves manipulating the genes inside neurons such that the cells produce a light-sensitive protein and respond to blasts of light. The aim is to activate a specific neuron and then see which other neurons are activated shortly afterwards, revealing a connected circuit. However, exposure to light can be imprecise. Also, the neurons in the brain are so densely packed that the nerve endings from neighboring neurons often overlap without actually being connected. This makes it unclear if activated neurons are truly part of the same circuit or simply bystanders reacting to the same nearby blast of light. To overcome this limitation, Baker et al. developed a new optogenetic approach with two important features. First, the approach makes use of a light-sensitive protein called channelrhodopsin that had been modified to confine it to the cell body of each neuron and exclude it from the nerve endings. Second, pulses of laser light were specifically shaped to target only the cell body of an individual neuron. Baker et al. show that this new method can activate neurons inside slices of mouse brain without affecting the neighboring neurons. This allowed circuits of neurons to be mapped in fine detail. This new optogenetic method is expected to shed light on the patterns of nerve signals that contribute to animal behavior. The approach may also be modified to use other light-sensitive proteins or investigate how neural circuits are altered in animal models of human disorders like autism and schizophrenia. DOI: http://dx.doi.org/10.7554/eLife.14193.002 Author Keywords optogeneticscircuitconnectivityResearch Organism MouseMax Planck Florida InstituteBaker Christopher A Elyada Yishai M Parra-Martin Andres Bolton M McLean The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.elife-xml-version2.5Author impact statementCombining spatial restriction of channelrhodopsin to the neuronal cell body with two photon excitation and calcium imaging will enable production of high resolution maps of neural circuitry. ==== Body Introduction The synaptic organization of individual neurons into circuits is the physiological basis for the interpretation of sensory input and production of behavioral responses. Understanding the precise patterns of connectivity among the distinct types of neurons that comprise neural circuits is critical for elucidating circuit function and ultimately requires methods that can map functional connectivity with single-cell resolution. Optical activation of single neurons using two-photon excitation of caged neurotransmitters or optogenetic probes such as channelrhodopsin (ChR2) provides a powerful approach for assessing the synaptic connections of single neurons. In particular, optogenetic mapping utilizing ChR2 and the rapidly expanding family of opsin variants have increased the flexibility and precision of mapping paradigms. Variations in the single-channel properties of the opsins can be exploited to generate rapid action potential trains or sustained depolarizations (Mattis et al., 2012), and new red-shifted variants have facilitated excitation deeper in tissue and have enabled simultaneous optical control of two distinct neuronal populations (Klapoetke et al., 2014; Lin et al., 2013; Yizhar et al., 2011). In addition, genetic restriction of opsin expression using transgenic mouse lines enhances the ability to activate and assess the connectivity of specific cell types. Despite the great potential of optogenetics for mapping the synaptic connections of single neurons, there are multiple issues that have limited its effectiveness. First, two-photon activation of single neurons with ChR2 is complicated by its kinetics and low single-channel conductance. A diffraction-limited spot does not activate sufficient channels simultaneously to reliably bring neurons expressing ChR2 to action potential threshold. Several solutions have been implemented to address this. Rapid scanning of a diffraction-limited two-photon excitation spot across an opsin-expressing cell allows sufficient temporal integration to generate action potentials (Packer et al., 2012; Prakash et al., 2012; Rickgauer and Tank, 2009). Alternatively, scanless two-photon excitation by temporal focusing (Oron et al., 2005; Zhu et al., 2005) increases the number of simultaneously excited opsin molecules by expanding the beam in the imaging plane without sacrificing the optical sectioning of multiphoton microscopy (Andrasfalvy et al., 2010; Losonczy et al., 2010; Papagiakoumou et al., 2010; Rickgauer et al., 2014). The implementation of diffractive optical elements (Fino and Yuste, 2011; Nikolenko et al., 2007; Packer and Yuste, 2011) or spatial light modulators (Dal Maschio et al., 2010; Nikolenko et al., 2008; Packer et al., 2012, 2015; Papagiakoumou et al., 2010, 2009, 2008) has also allowed for more complicated excitation profiles that encompass multiple spots around a cell, ensembles of neurons, or particular branches of dendritic trees. While these approaches have made it possible to provide sufficient two-photon illumination to reliably drive action potentials, the ability to use optogenetic stimulation to selectively target single neurons remains challenging because the opsins are expressed throughout the dendritic and axon terminal fields, generating a potential confounding source of light-induced electrical responses. For example, it can be difficult to know whether a recorded electrophysiological event is due to stimulation of a presynaptic cell or 'direct' stimulation of a portion of the recorded cell’s dendritic arbor. Although these possibilities can be distinguished by their kinetics, large amplitude direct responses may obscure simultaneous smaller synaptic events. This essentially leaves an indeterminate region of any circuit 'map' coinciding with the dendritic arbors of the recorded neuron, which can extend for 200 µm or more around the soma. In addition, the optogenetic approach could be compromised by unintended activation of fibers of passage or local axonal boutons, which are known to respond to temporal focusing of two-photon excitation (Andrasfalvy et al., 2010; Losonczy et al., 2010). Exclusion of opsins from axonal compartments has previously been achieved by fusing the opsin with targeting domains that bind to myosin Va motors necessary for transporting proteins into dendrites; one such motif from melanophilin is sufficient to exclude ChR2 from axons and enhance the resolution of neural circuit maps (Lewis et al., 2009). The resulting ChR2 distribution, however, remains throughout the dendritic tree and thus does not solve the problem of undesired direct activation of a neuron’s processes while trying to stimulate other neurons in close proximity. To overcome these limitations, we combined temporal focusing with spatial confinement of ChR2 expression to the neuronal cell body and proximal dendrites. Our alternative targeting approach took advantage of the Kv2.1 potassium channel, which has a particularly unique localization to clusters at the neuronal soma and proximal dendrites (Trimmer, 1991) achieved through a 66-amino acid domain in its C-terminus (Lim et al., 2000) that drives its association with myosin IIb and specific post-Golgi transport vesicles destined for the somatic compartment (Jensen et al., 2014). Furthermore, this targeting signal is sufficient to alter ChR2 trafficking in retinal ganglion cells (Wu et al., 2013). We have employed this approach to target ChR2 to the soma and proximal dendrites of neurons in somatosensory cortex and added a nuclear fluorescent tag to identify ChR2-expressing neurons for targeting a two-photon temporal focused mapping beam. We combined this soma-targeted ChR2 with verification of successful action potential generation with a genetically encoded calcium indicator in our mapping protocol. Our approach allows robust and precise activation of neurons in brain slices for the construction of functional synaptic connectivity maps with single-cell resolution without loss of information about local connections in the region of the dendritic arbor of the recorded neuron or inadvertent activation of axons. Results We stimulated neurons expressing soma-targeted ChR2 in acute slices of mouse cortex using scanless temporal focusing (TF), which has been used successfully for optogenetic stimulation at high axial resolution in scattering tissue samples (Andrasfalvy et al., 2010; Losonczy et al., 2010; Papagiakoumou et al., 2010, 2013; Rickgauer et al., 2014). TF uses a diffraction grating (Figure 1—figure supplement 1) to separate the spectral components of a pulsed laser beam, resulting in a temporally broadened pulse that is inefficient at excitation except at the focal plane, where the components are recombined. This yields a volume of excitation in which the diameter in the x-y plane and the thickness in the axial plane are controlled independently (Oron et al., 2012, 2005). We designed our excitation volume to approach the dimensions of a typical neuronal soma (Figure 1—figure supplement 2). To restrict expression of ChR2 to the soma and proximal dendrites, we generated ChR2 fusion proteins by attaching a 65 amino acid motif from the Kv2.1 voltage-gated potassium channel to the carboxy terminus of ChR2-EYFP (Lim et al., 2000). Nontargeted ChR2-EYFP fluorescence was distributed throughout the processes of dissociated cortical neurons that had been filled with a fluorescent dye (Figure 1A). In contrast, targeted ChR2-EYFP-Kv2.1 was located primarily in the soma and excluded from the axon and distal dendrites (Figure 1A).10.7554/eLife.14193.003Figure 1. Characterization of soma-targeted channelrhodopsin (ChR2). (A) Fluorescence of ChR2-EYFP (green) in live dissociated cortical neurons that had been patched and filled with Alexa 594 dye (magenta) as visualized by two-photon microscopy. Scale bar = 100 µm. (B) Representative two-photon maximum intensity projections of Alexa 594 fluorescence and current responses to a single 150 ms TF stimulation pulse (red bar) for patched and dye-filled pyramidal cells in acute slices expressing nontargeted (N) or targeted (T) ChR2. Scale bar = 100 µm. (C) Mean (solid lines) ± s.e.m. (shaded regions) for current responses to TF stimulation measured at specific points along a single dendrite (N=10 nontargeted cells and 13 targeted cells). Fitting each curve with an exponential function demonstrated that the nontargeted and targeted datasets were significantly different (*p<0.05 by F-test). (D) Maps of direct currents in representative cells expressing ChR2 constructs when stimulating a 15 × 15 grid of locations. Each pixel in a map represents the direct current observed in the patched cell when that location in the slice was stimulated. Each position in a map was stimulated with the minimum power that evoked action potentials in 10/10 trials when stimulation was applied to the soma. Currents evoked by such powers ranged from 20 to over 250 pA, as indicated by the scale at bottom. Scale bar at lower left = 100 µm. DOI: http://dx.doi.org/10.7554/eLife.14193.003 10.7554/eLife.14193.004Figure 1—figure supplement 1. Custom two-photon microscope setup for temporal focusing stimulation with simultaneous imaging. DOI: http://dx.doi.org/10.7554/eLife.14193.004 10.7554/eLife.14193.005Figure 1—figure supplement 2. Orthogonal projections from a z-stack of a spin-coated layer of 200 nm fluorescent beads imaged with the temporal focusing beam and collected with a CCD camera. Scale bar = 20 µm. The full width at half maximum was 9.0 µm in the xy plane and 9.5 µm in the axial plane. DOI: http://dx.doi.org/10.7554/eLife.14193.005 10.7554/eLife.14193.006Figure 1—figure supplement 3. Characterization of targeted ChR2 in axonal processes. (A) Maximum intensity projections of confocal image stacks of brain sections from animals expressing nontargeted (N) or targeted (T) ChR2. Dashed boxes in the upper panels indicate the positions of the images shown in the lower panels. Scale bars = 100 µm. (B) Representative two-photon maximum intensity projections of Alexa 594 fluorescence and current responses to a single 150 ms TF stimulation pulse (red bar) for patched and dye-filled pyramidal cells in acute slices. Scale bar = 50 µm. (C) Mean (solid lines) ± s.e.m. (shaded regions) for current responses to TF stimulation measured at specific points along an axonal process (N=16 nontargeted cells and 9 targeted cells). Fitting each curve with an exponential function demonstrated that the nontargeted and targeted datasets were significantly different (*p<0.05 by F-test). DOI: http://dx.doi.org/10.7554/eLife.14193.006 As a functional assay to compare the distribution of ChR2 in the soma and dendrites of targeted and non-targeted constructs, we used TF stimulation at intervals along the apical dendrite of patched layer II/III pyramidal neurons in acute coronal brain slices of somatosensory cortex from mice expressing opsins from viral constructs and recorded light activated currents in voltage clamp (Figure 1B). The stimulation power for these experiments was determined independently for each neuron to be the minimum that elicited a single action potential in 10 out of 10 trials when the TF spot was placed directly over the soma; this value was 0.92 ± 0.24 mW/µm2 for targeted ChR2 and 1.65 ± 0.26 mW/µm2 for the nontargeted version, owing to the increased sensitivity of the targeted construct (see below). Dendrites were followed throughout the depth of the tissue and areas were selected for stimulation such that the dendrite was planar throughout the extent of the 10 µm TF disc. The TF-stimulated current declined ~10-fold at 50 µm from the soma with the targeted construct, versus two fold for the non-targeted construct (Figure 1C; N=10 untargeted cells from four animals and N=13 targeted cells from six animals). The reduced dendritic ChR2 in the targeted construct opens up the possibility of using light stimulation with patch recording to identify synaptic currents that originate from nearby neurons. To illustrate this point, we compared the spatial distribution of direct currents recorded in Layer II/III pyramidal neurons in acute slices of somatosensory cortex expressing non-targeted and targeted constructs while sequentially stimulating points at 20 µm intervals over a 300 × 300 um grid surrounding the soma. Stimulation of cells expressing non-targeted ChR2-evoked currents up to 200 µm away from the neuronal soma, often delineating a pattern indicative of the dendritic arbor of the cell (Figure 1D). In contrast, cells expressing the targeted construct exhibited significant currents only when stimulated within the 25–50 µm immediately adjacent to the cell body. The lack of direct light-activated currents throughout much of the neuron’s dendritic field makes it possible to visualize synaptic currents that would be evoked by ChR2 activation of nearby presynaptic neurons, even those that lie within the neuron’s dendritic field. In brain sections from mice injected with viruses encoding ChR2 constructs, examination of regions near the edge of the extent of virus-driven expression revealed markedly denser labeling of neuronal processes around a single-neuron expressing nontargeted ChR2 versus the area around several neurons expressing the targeted version (Figure 1—figure supplement 3A). We also saw processes reminiscent of axons at distances of several hundred microns away from the cell bodies of conventional ChR2-positive neurons. To functionally characterize the extent of axonal opsin expression, we again patched layer II/III pyramidal neurons in acute coronal brain slices of somatosensory cortex and recorded light-activated currents in voltage clamp while stimulating multiple positions along putative axonal processes, which were distinguished by their thin profile, absence of spines, and presence of occasional varicosities. Although the amount of detected current decreased rapidly for both constructs as TF stimulation was moved along the axon, noticeable currents could be detected at distances of 50 µm or greater from the soma of cells expressing nontargeted ChR2 (Figure 1—figure supplement 3B). In contrast, cells expressing targeted ChR2 demonstrated an 80% reduction of current when stimulating just 20 µm down the axon and significantly less current than cells expressing nontargeted ChR2 throughout its examined length (Figure 1—figure supplement 3C). Interestingly, the targeted ChR2 also exhibited increased peak current amplitude in response to TF activation, presumably by concentrating channel density at the soma (Figure 2A; from 308.3 ± 43.5 pA in 12 nontargeted cells to 760.5 ± 146.0 pA in 12 targeted cells at 3.63 mW/µm2; p=0.0101 for main effect of targeting by two-way repeated measures ANOVA), leading to an over three-fold reduction of the power required to reach action potential threshold (Figure 2B; from 2.75 ± 0.31 mW/µm2 in 17 nontargeted cells to 0.88 ± 0.23 mW/µm2 in 19 targeted cells; p=8.028 × 10–5 by Mann-Whitney U Test). At the threshold stimulation power for each cell, the amount of evoked current was similar between targeted and untargeted constructs (261.06 ± 31.62 pA for untargeted ChR2 and 222.70 ± 27.21 pA for targeted ChR2 p=0.3677 by two sample t-test), consistent with the observation that the targeting modification had no effect on rheobase, other intrinsic physiological parameters, or action potential properties in response to current injections (Figure 2—figure supplement 1 and Table 1). The ability to trigger action potentials at lower incident power, coupled with the somatic restriction of the targeted construct, should provide enhanced spatial and temporal resolution for mapping neuronal circuits. We measured the spatial resolution of action potential generation by moving the TF spot to different lateral and axial locations relative to a patched neuron expressing targeted ChR2 and examining the proportion of 10 trials that resulted in an action potential when using the threshold stimulation intensity. The full-width at half maximum of these measurements was 11.1 µm laterally and 23.3 µm axially (N=10 cells from 7 animals; Figure 2C and D). Similar measurements using nontargeted ChR2 demonstrated resolution of 19.6 µm laterally and 36.2 µm axially (N=13 cells from 5 animals; Figure 2C and D); both curve fits were significantly different from the targeted ChR2 versions (p<0.05 by F-test). In terms of temporal precision, the mean latency from light onset to generation of a single action potential in the above experiments was 38.98 ± 17.33 ms; the average jitter (defined as the standard deviation of the latency across the ten trials for a given cell) was 6.8 ± 2.1 ms. Increasing stimulation power beyond threshold provoked additional action potentials and eventually shortened the latency to first spike to 9.3 ± 1.8 ms (Figure 2—figure supplement 2), consistent with reported values for stimulating cortical neurons with sculpted light (Papagiakoumou et al., 2010). Targeted ChR2-expressing cells exhibited shorter action potential latencies than nontargeted ChR2 cells at equivalent stimulation intensity, consistent with latency being power-dependent and the earlier observation that targeted ChR2 is more sensitive. Although we concentrated efforts on longer stimulation pulses to increase the number of spikes and facilitate calcium imaging (see below), we verified the performance of targeted ChR2 under a shorter stimulation regimen. Temporal focusing pulses of 32 ms were effective at generating action potentials in opsin-expressing cells, and targeted ChR2 exhibited greatly enhanced photocurrents and lower power thresholds under these conditions (Figure 2—figure supplement 3). Moreover, the ability to reduce latency to less than 10 ms with increased power (Figure 2—figure supplement 2) suggests that even shorter stimulation pulses may still be effective. Therefore, the combination of temporal focusing and soma-targeted ChR2 expression yields a highly reliable and spatially precise means to stimulate action potentials with somatic illumination.10.7554/eLife.14193.007Figure 2. Enhanced sensitivity and resolution with soma-targeted ChR2. (A) Evoked currents in acute slices in response to 150 ms TF stimulation of increasing intensity; targeted ChR2 exhibited significantly greater current (two-way repeated measures ANOVA, p=0.0101 for main effect of targeting, p=0.0056 for targeting × power interaction, *p<0.05 Holm-Bonferroni post hoc test). (B) Minimum TF power required to elicit a single action potential. Each point indicates a single cell; cells expressing targeted ChR2 had a much lower threshold (*p=8.028 × 10–5, Mann-Whitney U Test). (C–D) Probability of evoking spikes per 10 trials in cells expressing targeted (red) or nontargeted (blue) ChR2 when stimulating the soma and at laterally and axially displaced locations. DOI: http://dx.doi.org/10.7554/eLife.14193.007 10.7554/eLife.14193.008Figure 2—figure supplement 1. Action potentials triggered in neurons expressing nontargeted (N) or targeted (T) ChR2 in response to stepwise current injections ranging from −100 to 110 pA in 30 pA steps. DOI: http://dx.doi.org/10.7554/eLife.14193.008 10.7554/eLife.14193.009Figure 2—figure supplement 2. Action potential latency as a function of stimulation power in cells expressing targeted (red) or nontargeted (blue) ChR2. Mean (solid lines) ± s.e.m. (shaded regions) latency to first action potential in response to 150 ms TF stimulation of increasing intensity in cells from acute cortical slices (N=9 cells per group). Fitting each curve with an exponential function demonstrated that the nontargeted and targeted datasets were significantly different (*p<0.05 by F-test). DOI: http://dx.doi.org/10.7554/eLife.14193.009 10.7554/eLife.14193.010Figure 2—figure supplement 3. Latency to first action potential with targeted and nontargeted ChR2. (A) Evoked currents in acute slices in response to shorter 32 ms TF stimulation of increasing intensity; targeted ChR2 exhibited significantly greater current (two-way repeated measures ANOVA, p=1.951 × 10–4 for main effect of targeting, p=0.01 for targeting × power interaction, *p<0.05 Holm-Bonferroni post hoc test). (B) Minimum TF power required to elicit a single action potential. Each point indicates a single cell; cells expressing targeted ChR2 had a much lower threshold (*p=9.124 × 10–6, Mann-Whitney U Test). (C) Mean (solid lines) ± s.e.m. (shaded regions) for current responses to TF stimulation measured at specific points along a single dendrite (N=6 nontargeted cells and 6 targeted cells). Fitting each curve with an exponential function demonstrated that the nontargeted and targeted datasets were significantly different (*p<0.05 by F-test). DOI: http://dx.doi.org/10.7554/eLife.14193.010 10.7554/eLife.14193.011Table 1. Intrinsic electrophysiological properties of neurons expressing normal or targeted channel rhodopsin constructs. Rheobase measurements were made in response to current injections. All measurements are mean ± s.e.m. No significant differences as an effect of targeting were found (Mann-Whitney U Tests, all p>0.05). DOI: http://dx.doi.org/10.7554/eLife.14193.011 Nontargeted Targeted p Input resistance (MΩ) 129.16 ± 8.91 (N = 11) 148.36 ± 17.76 (N = 18) 0.9462 Capacitance (pF) 125.25 ± 16.99 (N = 11) 89.85 ± 7.66 (N = 18) 0.0887 Resting potential (mV) −72.71 ± 3.86 (N = 7) −76.87 ± 1.87 (N = 8) 0.3532 Rheobase (pA) 160.00 ± 7.69 (N = 12) 147.39 ± 12.92 (N = 23) 0.5357 Spikes at 1.5× rheobase 7.18 ± 0.52 (N = 12) 7.56 ± 0.83 (N = 23) 0.8377 To demonstrate the general utility of these techniques for mapping synaptic connections, we performed connectivity experiments in acute somatosensory cortical slices. We generated a bicistronic adeno-associated virus (AAV) construct encoding targeted ChR2-Kv2.1 followed by a P2A ribosomal skipping sequence and a histone 2B-mRuby2 fusion protein to fluorescently label neuronal nuclei to identify cells for TF stimulation. We patched a cell in layer II/III, and then stimulated surrounding cells that had been identified by nuclear mRuby2 fluorescence. Action potential firing in a light-stimulated neuron was verified by an increase in the fluorescence signal measured with the genetically encoded calcium sensor GCaMP6s (Chen et al., 2013) expressed from a separate AAV construct. For these experiments, we used an average stimulation power (2.29 ± 0.55 mW/µm2) that, when combined with a 150 ms stimulation pulse, frequently led to trains of 2–4 action potentials in the patched cells and facilitated detection of larger calcium transients. We used an imaging power and dwell time that did not lead to action potential generation in any cells recorded, instead causing an average inward current of 29.5 ± 9.5 pA (N = 7 cells from 5 animals)—well below the average rheobase. The ability to identify a ChR2-Kv2.1-expressing neuron for TF stimulation by the presence of a fluorescent nuclear label and verify that the neuron has in fact been stimulated successfully by detecting calcium transients with GCaMP is a key advantage for the execution and interpretation of mapping experiments. Excitatory synaptic connections were identified by the presence of GCaMP fluorescence increase in only the TF stimulated neuron and a reproducible inward current with appropriate synaptic delay, kinetics and reversal potential in the patched cell (see Materials and methods). To separate spontaneous currents occurring during the stimulation epoch from bona fide synaptic events, we also required occurrence of synaptic responses on multiple stimulus repetitions and with a post-stimulus onset jitter of less than 14 ms. A lack of connection was defined as failure to reach these criteria following the presence of a GCaMP6s response to light stimulation in a potential presynaptic neuron. In a representative experiment, we identified 43 nuclear-labeled cells in a single axial plane (Figure 3A), of which 35 cells yielded calcium transients in response to TF stimulation (Figure 3B). Three of these 35 cells elicited reproducible postsynaptic responses in the recorded neuron when photostimulated (Figure 3C), exhibiting multiple currents consistent with the production of trains of presynaptic action potentials. Every connection detected was associated with an unequivocal calcium transient in only the stimulated neuron (Figure 3B). In repeated experiments (seven neurons from five animals), another example of which is shown in Figure 3—figure supplement 1, the average rate of detecting a calcium response to TF stimulation of a cell expressing the ChR2-Kv-P2A-H2B-mRuby2 bicistronic construct was 80.00 ± 2.60%. Given that the GCaMP6s is expressed by a separate AAV construct and coinfection is not necessarily 100%, the 'nonresponsive' cells may express insufficient GCaMP for detection of single action potentials. Indeed, we identified three reproducible postsynaptic events (out of 316 presynaptic stimulations) that could not be correlated with a calcium transient, perhaps again due to lower GCaMP expression in those neurons. Overall, the average connectivity rate was 10.27 ± 2.60% (27 connections out of 252 cells showing a calcium transient to photostimulation), which was not significantly different from our own results using paired recording (7 connections out of 115 cells; Fisher’s exact test, p=0.18).10.7554/eLife.14193.012Figure 3. Typical mapping experiment with action potential validation by GCaMP6s. (A) In acute slices expressing GCaMP6s (green) and targeted ChR2-P2A-H2B-mRuby2 (magenta), cells were identified for photostimulation based on nuclear mRuby2 fluorescence (yellow circles). A cell was patched and dye-filled (triangle) and assessed for postsynaptic currents as surrounding cells were stimulated (1.40 mW/µm2 incident power) and GCaMP fluorescence was simultaneously recorded. Scale bar = 100 µm. (B) Changes in GCaMP fluorescence over the experimental timecourse for each cell identified in (A); each cell was stimulated sequentially with ~2 s between cells. Signals are the average of four trials. In this experiment, 35 out of 43 cells yielded calcium transients in response to optical stimulation; overall the average probability of detecting an induced calcium response was 80%. (C) Calcium (blue) responses for a subset of cells reacting to TF stimulation; four trials are overlaid in light blue, with the average in dark blue. Red lines in the calcium traces indicate the onset of stimulation for each cell. The recorded currents for the four trials (black) in the patched and putatively postsynaptic cell are shown expanded below each calcium trace; the shaded red region indicates the 150 ms stimulation epoch. Three cells showed a calcium response to TF stimulation and triggered postsynaptic currents (fast onset kinetics but delayed with respect to the TF stimulation) in the patched cell. Five representative cells showing calcium transients but no reproducible postsynaptic currents are also shown. For the displayed three connected cells and the five unconnected cells, the cell numbers, spatial locations and full calcium traces are marked with plus signs and minus signs, respectively, in panels (A) and (B). Across multiple experiments, the average connection probability was 10% (27 connections out of 252 responsive cells). DOI: http://dx.doi.org/10.7554/eLife.14193.012 10.7554/eLife.14193.013Figure 3—figure supplement 1. Second typical mapping experiment with action potential validation by GCaMP6s. (A) In acute slices expressing GCaMP6s (green) and targeted ChR2-P2A-H2B-mRuby2 (magenta), cells were identified for photostimulation based on nuclear mRuby2 fluorescence (yellow circles). A cell was patched and assessed for postsynaptic currents as surrounding cells were stimulated (1.40 mW/µm2 incident power) and GCaMP fluorescence was simultaneously recorded. Scale bar = 100 µm. (B) Changes in GCaMP fluorescence over the experimental timecourse for each cell identified in (A); each cell was stimulated sequentially as described in Figure 3. Signals are the average of four trials. (C) Calcium (blue) responses for a subset of cells reacting to TF stimulation; four trials are overlaid in light blue, with the average in dark blue. Red lines in the calcium traces indicate the onset of stimulation for each cell. The recorded currents for the four trials (black) in the patched and putatively postsynaptic cell are shown expanded below each calcium trace; the shaded red region indicates the 150 ms stimulation epoch. Four cells with a calcium response to TF stimulation and coincident postsynaptic currents (fast onset kinetics but delayed with respect to the TF stimulation) in the patched cell are shown, as are four representative cells showing calcium transients but no reproducible postsynaptic currents. For the displayed connected and unconnected cells, the cell numbers, spatial locations and full calcium traces are marked with plus signs and minus signs, respectively, in panels (A) and (B). DOI: http://dx.doi.org/10.7554/eLife.14193.013 For comparison, we executed similar GCaMP6s-monitored mapping experiments (three neurons from two animals) using nontargeted ChR2 and the same average stimulation power (Figure 4). Although we were able to detect putative synaptic connections, they were frequently coincident with direct current responses indicative of stimulation of the patched cell’s dendritic arbor (Figure 4C). Across the experiments, such direct responses occurred during stimulation of 28.75 ± 4.46% of the target cells within a single field of view. These direct responses were often large in amplitude and could easily obfuscate much smaller synaptic events associated with bona fide connections. Moreover, we also observed reproducible calcium transients in off-target cells when other target cells were being stimulated (note events in Figure 4B that lie off of the diagonal). These off-target responses occurred with a probability (the number of events divided by the number of target cells) of 15.17 ± 1.79% and likely result from unintentional stimulation of a ChR2-containing sensitive dendrite of one cell while intending to activate the soma of a separate cell. Consistent with this interpretation, these off-target calcium events were observed at a significantly lower frequency in experiments with the targeted opsin (5.42 ± 1.39%, p<0.01 by two sample t-test). Moreover, the average distance between a cell exhibiting an off-target calcium transient and the intended target cell was greater with nontargeted ChR2 (40.6 ± 6.4 µm versus 18.9 ± 2.4 µm; p<0.01 by two sample t-test), further suggesting that cells expressing nontargeted ChR2 were firing action potentials in response to unintended stimulation of distal dendrites. Our combination of TF and restricted ChR2 thus facilitated the mapping of local circuitry within the 300 µm surrounding a neuron without confounding signals from its dendritic arbor and with a higher throughput than that achieved with electrophysiological techniques alone.10.7554/eLife.14193.014Figure 4. Typical mapping experiment using nontargeted ChR2. (A) In acute slices expressing GCaMP6s (green) and conventional ChR2-mRuby2, cells were identified for photostimulation based on responsiveness to widefield ChR excitation (yellow circles). A cell was patched and dye-filled (triangle) and assessed for postsynaptic currents as surrounding cells were stimulated (2.13 mW/µm2 incident power) and GCaMP fluorescence was simultaneously recorded. Scale bar = 100 µm. (B) Changes in GCaMP fluorescence over the experimental timecourse for each cell identified in (A); each cell was stimulated sequentially as described in Figure 3. Signals are the average of four trials. (C) Calcium (blue) responses for a subset of cells reacting to TF stimulation; four trials are overlaid in light blue, with the average in dark blue. Red lines in the calcium traces indicate the onset of stimulation for each cell. The recorded currents for the four trials (black) in the patched and putatively postsynaptic cell are shown expanded below each calcium trace; the shaded red region indicates the 150 ms stimulation epoch. Several types of events were observed: one cell showed a synaptic event coincident with a low amplitude current from direct stimulation of the dendritic arbor of the patched cell, and one cell showed what might be a synaptic event buried within another direct stimulation current of large amplitude (red arrows). Three other cells with a calcium response to TF stimulation and coincident direct stimulation of the patched cell’s dendritic arbor are also shown, along with three cells with calcium transients but no detectable currents in the patched cell. The cell numbers, spatial locations and full calcium traces corresponding to these events are marked in panels (A) and (B). DOI: http://dx.doi.org/10.7554/eLife.14193.014 Discussion This study demonstrates that combining temporal focusing for two-photon activation of ChR2 with restriction of ChR2 expression to the soma and proximal dendrites of neurons yields a reliable method for evaluating synaptic connectivity with single-neuron resolution. The spatially restricted ChR2 expression we describe allows unmasking of synaptic connections from neurons whose somata lie close to the dendrites of the postsynaptic cell and would have been occluded by direct activation of ChR2 on the dendrite. In addition, depletion of ChR2 from axons prevents inadvertent depolarization of boutons or fibers of passage that could compromise attempts to identify the source of a synaptic event. ChR2 harboring the Kv2.1 localization motif also showed enhanced sensitivity, which is of particular use in situations where excitation power is at a premium. Together these features of the soma-restricted construct significantly enhance the ability to map synaptic connections with single-cell resolution. The development of spatially-restricted optogenetic constructs is probably one of the most effective means for achieving single-cell resolution in local circuit mapping experiments. This restriction requires the identification of a relatively small compartment near the neuronal soma characterized by selective expression of particular proteins with known motifs responsible for their localization. The distribution of particular voltage-gated potassium channels defines such a subcellular domain along the soma and proximal dendrites (Trimmer, 2015), and the sufficiency of a short Kv2.1 C-terminal sequence for driving heterologous proteins to this region (Lim et al., 2000) made this strategy ideal for restricting ChR2 expression. The only obvious alternative is restriction to the axon initial segment (AIS) mediated by an ankyrin G binding motif present in voltage-gated sodium channels (Garrido et al., 2003). Although incorporating this motif into ChR2 drives its localization to the AIS (Grubb and Burrone, 2010; Wu et al., 2013), the resulting construct does not support light-generated action potentials under physiological conditions (Grubb and Burrone, 2010). Moreover, AIS-targeted ChR2 alters the intrinsic firing properties of retinal neurons, presumably by displacing endogenous voltage-gated sodium channels whose subcellular localization is also dependent on ankyrin G binding (Zhang et al., 2015). Therefore, the Kv2.1 targeting strategy currently remains the optimal means of concentrating ChR2 such that action potentials will not be generated by TF stimulation of dendrites >20 µm away from the soma. Indeed, this approach has been successfully used with opsins to artificially generate center-surround receptive fields in retinal ganglion cells (Wu et al., 2013). The Wu et al. (2013) study showed functional restriction of opsins to the soma and proximal dendrites by examining receptive field responses to wide (200 µm) bars of one-photon light. We have now measured the somatic targeting at a finer scale, demonstrated two-photon excitation of targeted ChR2 that is enhanced relative to normal ChR2 under the same conditions, and established the utility of targeted opsins for enhancing the resolution of local connectivity maps of neural circuits. The soma-targeted ChR2 could be combined with many of the techniques for two-photon optical control of neuronal activity previously developed for caged neurotransmitters or optogenetic probes. The relative merit of the optical activation method depends on the axial resolution, temporal precision required and the number of neurons to be stimulated simultaneously in a given experimental paradigm. For example, rapid scanning of a diffraction-limited two-photon excitation spot in a pattern on the soma can generate action potentials (Packer et al., 2012; Prakash et al., 2012; Rickgauer and Tank, 2009) and combination with spatial light modulators allows for simultaneous excitation of neuronal ensembles (Packer et al., 2012, 2015). The temporal resolution of this rapid scanning approach, however, is limited by the time required to scan along the cell body. Scanless activation of untargeted ChR in brain slices using TF, alone or in combination with spatial light modulators, is capable of generating action potentials in hippocampal pyramidal neurons within 1–3 ms of light onset (Andrasfalvy et al., 2010) and in cortical neurons in less than 10 ms (Papagiakoumou et al., 2010). We did not take full advantage of the temporal precision capability of TF to fire action potentials in our current study, instead focusing on a screening method that would identify connections without optimizing the amount of power that would fire each potential presynaptic neuron with minimal latency. We therefore chose longer pulses at a power sufficient to fire most neurons and generate trains of action potentials, which would elicit stronger signals with calcium indicators. For experiments requiring temporal precision, the minimization of action potential latency requires optimization of excitation area and laser power (Papagiakoumou et al., 2008). Because power is also the limiting factor in extending the area of two-photon activation to large numbers of neurons with spatial light modulators, the reduced power required to bring soma-targeted opsins to threshold would be an asset to such experiments. The optimal performance of our method relies on sufficient co-expression of an opsin, a marker for opsin expression, and a genetically encoded calcium sensor. For GCaMP6s and similar sensors, there appears to be a balance between sufficient expression to detect single action potentials and excessive levels that lead to a lack of responsiveness (Chen et al., 2013; Packer et al., 2015; Tian et al., 2009). Targeting this expression window can be complicated, particularly when simultaneously trying to achieve high opsin levels using a separate AAV construct. Indeed, heterogeneity in GCaMP expression levels between neurons with viral infection may be partially responsible for the 20% of neurons in which we did not detect a change in GCaMP fluorescence with TF stimulation. We have also explored the co-expression of GCaMP6s as the second member of a single P2A-mediated bicistronic construct, but the GCaMP6s expression level was lower than desired, resulting in inadequate detection of action potentials (data not shown). Future development of these techniques will therefore benefit from bypassing viral systems and instead expressing the calcium sensor or the opsin in transgenic animals. The enhanced sensitivity of the targeted ChR2 should alleviate the concern that there would be sufficient opsin expression from a single genetic locus for activation of neurons with two-photon excitation. There are several opportunities for future extension of this method. The somatic targeting approach could also be exploited in the context of other opsins such as C1V1, which has recently been leveraged for two-photon stimulation of neural circuits in acute slices and in vivo (Packer et al., 2012, 2015; Prakash et al., 2012; Rickgauer et al., 2014). Other molecules which may be good candidates for somatic targeting include the red-shifted ReaChR for its sensitivity (Lin et al., 2013) and Chronos for its rapid temporal characteristics (Klapoetke et al., 2014). Moreover, the somatic restriction of membrane-bound genetically encoded voltage sensors could dramatically reduce background associated with the neuropil and facilitate an all-optical version of our current method. Because the Kv2.1 targeting sequence is unable to restrict the distribution of the single-pass membrane protein CD8, however, there may be certain structural constraints on the effectiveness of the motif (Lim et al., 2000). We suspect that at least for the opsins, which are all seven transmembrane domain proteins likely to have a similar structure, the Kv2.1 motif will be sufficient to achieve somatic localization. In summary, our combination of TF and soma-restricted ChR2 enables functional connectivity mapping and is straightforward and easy to implement with standard two-photon microscopes. Furthermore, these techniques could also be used in vivo, where the enhanced resolution of the targeted ChR2 makes it especially attractive for selective cell stimulation in behavioral paradigms. These enhancements to probing brain microcircuitry through optical stimulation promise to reveal much about nervous system function and how it might be modified by experience and perturbed in animal models of neurologic or psychiatric disease. Materials and methods Construct generation The vector pAAV-hSyn-hChR2(H134R)-EYFP (Addgene plasmid 26973) served as the backbone for generating certain modified constructs. The 'proximal restriction and clustering signal' (Lim et al., 2000) of the Kv2.1 voltage-gated potassium channel (QSQPILNTKEMAPQSKPPEELEMSSMPSPVAPLPARTEGVIDMRSMSSIDSFISCATDFPEATRF), codon optimized for mouse, was generated by automated gene synthesis (Integrated DNA Technologies, Coralville, IA) and amplified by PCR using primers CGGCATGGACGAGCTGTACAAGCAGTCCCAGCCTATTCTGAAC and TGATATCGAATTCTTACTTAAACCGCGTAGCCTCTGG. The resulting product was inserted into the BsrGI site at the C-terminus of the ChR2-EYFP fusion protein sequence using the Gibson Assembly kit (New England Biolabs, Ipswich, MA). For mapping experiments using nontargeted hChR2 with coincident visualization of GCaMP6s, the EYFP-coding sequence between the PshAI and BsrGI sites was replaced with the sequence of mRuby2. To better visualize cells for stimulation during mapping experiments, we generated a bicistronic AAV construct consisting of hChR2 followed immediately by the Kv2.1 targeting sequence, a P2A ribosomal skipping sequence, and a histone 2B-mRuby2 fusion protein. The Kv2.1 sequence was amplified with primers ATCGAGGTCGAGACTCTCGTCGAAGACGAAGCCGAGGCCGGAGCCGTGCCAGCGGCCGCCACCCAGTCCCAGCCTATTCT and ACGTCTCCTGCTTGCTTTAACAGAGAGAAGTTCGTGGCTCCGGATCCAAACCGCGTAGCCTCTGG, histone 2B was amplified from Addgene plasmid 11680 with primers CTGTTAAAGCAAGCAGGAGACGTGGAAGAAAACCCCGGTCCTGGTTCTATGCCAGAGCCAGCGAAG and TCGGCTTCGTCTTCGACGAGCATGGTGGCGACCGGTG, and mRuby2 was amplified from Addgene plasmid 40260 with primers CTCGTCGAAGACGAAGCCGAGGCCGGAGCCGTGCCAGCGGCCGCCACCATGGTGTCTAAGGGCGAA and GATAAGCTTGATATCGTTACTTGTACAGCTCG. The amplified products were incorporated into an EcoRI-PshAI fragment from Addgene plasmid 26973 by Gibson Assembly; the P2A sequence was reconstituted from the primers during the assembly process. All plasmids were propagated in Stbl3 cells (Life Technologies, Grand Island, NY) and made into recombinant adeno-associated viral (AAV) particles of serotype 1 by the Penn Vector Core (Philadelphia, PA). Animals and viral injections All animal work was conducted according to the Guide for the Care and Use of Laboratory Animals from National Institute of Health. C57BL6/J mice were maintained on a 12 hr light-dark cycle with ad libitum access to food and water. Under isofluorane anesthesia, P21-P25 mice were injected with AAV particles (800 nl containing ~1 × 1013 genome copy units/ml) at 9.2 nl per 10 s through a pipette positioned 500 µm beneath the surface of the somatosensory cortex and attached to a Nanoject II microinjector. For mapping experiments with nuclear labeling and verification of action potential generation by calcium influx, full strength AAV1 encoding ChR2-Kv2.1-P2A-H2B-mRuby2 (5 × 1012 units/ml) was coinjected with a diluted concentration of AAV1 GCaMP6s (final concentration = 1.4 × 1012 units/ml); diluted GCaMP6s reportedly results in nuclear exclusion of the calcium indicator and lower toxicity (Packer et al., 2015). For all data reported here, animals receiving targeted or untargeted ChR2 viruses were all examined between 3 and 4 weeks after injection. Somatic restriction was also observed in animals 5 weeks after injection. Dissociated cortical neurons Neurons were prepared by dissecting the corticies from postnatal day 1 animals and digesting for 30 min at 37°C in Earle’s Balanced Salt Solution supplemented with 1.5 mM MgSO4, 1 mM CaCl2, and 8.3 units/ml papain (Worthington Biochemical, Lakewood, NJ) under 95%O2/5%CO2. Just prior to plating, cells were transfected with 1 µg DNA per 2.5 × 106 cells using the Amaxa nucleofector system. Neurons were then seeded onto laminin-coated coverslips containing a feeder layer of astrocytes prepared as previously described (McCarthy and de Vellis, 1980). Cells were maintained in neuronal growth media (Neurobasal (Life Technologies) supplemented with 5 µg/ml insulin, 110 µg/ml sodium pyruvate, 100 units/ml penicillin, 100 µg/ml streptomycin, 40 ng/ml thyroxine, 292 µg/ml L-glutamine, 5 µg/ml N-acetyl-L-cysteine, 100 µg/ml BSA, 100 µg/ml transferrin, 16 µg/ml putrescine, 60 ng/ml progesterone, 40 ng/ml sodium selenite, 50 ng/ml BDNF, and 5 ng/ml forskolin) for ~10 days before use in experiments. Slice preparation and electrophysiology Three weeks after viral injection, 300–400 µm slices were prepared in ice-cold cutting solution containing (in mM): 124 choline chloride, 26 NaHCO3, 2.5 KCl, 3.3 MgCl2, 1.2 NaH2PO4, 1 glucose and 0.5 CaCl2. After cutting, slices were allowed to recover for 30 min at 32°C in artificial CSF containing (in mM): 124 NaCl, 26 NaHCO3, 3 KCl, 1.25 NaH2PO4, 20 glucose, 1 MgCl2, 2 CaCl2, 5 sodium ascorbate, 3 sodium pyruvate and 2 thiourea. All solutions were maintained under constant 95%O2/5%CO2. Whole-cell recordings were made through 4-7 MΩ pipettes, filled with intracellular solution containing (in mM): 145 potassium gluconate, 5 NaCl, 10 HEPES, 0.5 EGTA, 4 MgATP, 0.3 NaGTP, 0.02 Alexa Fluor 488 or 594 hydrazide. In some experiments, the intracellular solution was supplemented with 0.2% biocytin. All recordings were collected using Axon Multiclamp 700B amplifiers and Digidata 1440A digitizers (Molecular Devices, Sunnyvale, CA) at 10 kHz, controlled by Clampex software. Optical setup Our customized optical setup was based on a AxioImager Z1 platform (Zeiss, Thornwood, NY) fitted with an Ultima dual path scan head containing two pairs of galvanometric mirrors (Bruker, Middleton, WI) for separate control of imaging and stimulating optical pathways (Figure 1—figure supplement 1). The imaging pathway used a Chameleon Ti:sapphire laser tuned to 920 nm (Coherent, Santa Clara, CA). The stimulation pathway employed a MaiTai DeepSee Ti:sapphire laser tuned to 880 nm (SpectraPhysics, Santa Clara, CA). The spectral broadening necessary for temporal focusing was achieved on this path by placing a 300 line/mm diffraction grating (Thorlabs, Newton, NJ) approximately 1 m away from the galvanometric mirrors; the laser spot on the grating was imaged onto the plane between the mirrors using a 500 mm focal length lens (Thorlabs) placed 500 mm away from the grating along the path of the first diffraction order from the grating. The stimulation laser was recombined with the imaging laser in the scan head using a 900 nm long-pass dichroic mirror (Chroma, Bellows Falls, VT). The power applied by either laser was controlled with a separate Pockels cell (Conoptics, Danbury, CT). Confocal microscopy Three weeks after viral injection, animals were transcardially perfused with saline followed by 4% paraformaldehyde. Brains were removed, cryoprotected in 30% sucrose, and sectioned at 50 µm intervals on a sliding microtome. Sections were stained for ChR2-EYFP fusion proteins with a rabbit antibody raised against GFP followed by an Alexa 488-labeled secondary antibody (Thermo Fisher Scientific, Waltham, MA). Sections were mounted in Prolong Gold (Thermo Fisher Scientific) and images were collected on a Zeiss 780 confocal microscope using 10× air- and 40× oil-immersion objectives. Acquisition settings for laser power and PMT voltages were kept constant between nontargeted and targeted ChR2 samples. Stimulation and resolution studies Temporal focusing stimulation was controlled by Prairie View 5.0 software (Bruker). Standard stimulation protocols used 150 ms pulses, except as noted in the text. Stimulation powers were measured after the objective, and ranged from 15 to 285 mW; assuming our spot size to be at least 10 µm in diameter, we are using powers of no greater than 0.2 to 3.6 mW/µm2. This calculation assumes a completely flat excitation profile; in reality, there is some spread in the axial dimension and thus, the power density is probably even lower. Cells filled with fluorescent Alexa dyes were visualized by two-photon microscopy and recorded in whole-cell voltage or current clamp as the temporal focusing spot was moved to different axial planes, different cell bodies or particular spots along dendrites. For each dendrite examined, the temporal focusing-induced current was plotted as a function of distance from the cell soma; the resulting curves were averaged for each construct to generate the traces shown in Figure 1C. All responses in Figure 1 were verified to exhibit kinetics associated with direct stimulation of the patched cells: that is, exhibiting onset and offset precisely locked to the start and end of the optical stimulation. Stimulation powers were chosen separately for each cell by identifying the minimum power that evoked action potentials continually across ten trials. Resolution of spike generation was measured by moving the stimulation spot at 9 µm intervals laterally or 10 µm axially above or below the cell soma and recording the proportion of trials out of 10 leading to an action potential. Data were collected from cells until a minimum of ten cells was obtained for following ChR2 expression along dendrites; given a standard deviation of ~11% for these measurements, our sample size should be able to detect a 15% difference between groups with a power of 0.8. Data for other measures were collected in parallel, and no data were ever excluded even if a recording failed prior to collecting the dendrite expression results. Calcium imaging and connection analysis Images of 512 × 512 pixels were simultaneously collected at 1.5 Hz using a raster scanning galvonometric system. Each stimulation epoch was timed to coincide with the onset of every other imaging frame, that is, the first spot was stimulated at the onset of the second frame, the second spot was stimulated 1.26 s later at the onset of frame 4, and so on. In this manner, a calcium peak in any frame could be associated with stimulation of a particular point in the field. During each stimulation epoch, microscope PMTs were protected from saturating fluorescence signals by Uniblitz shutters (Vincent Associates, Rochester, NY). Because this resulted in a lack of signal in the first 150 ms of every other frame as the top portion of the microscope field was scanned by the imaging laser, all stimulation points and calcium data collection were confined to a ~200 × 300 µm region of the microscope field. First-pass regions of interest (ROIs) were based on the stimulation points defined by labeled nuclei; frames containing significant calcium transients were then used to manually refine each ROI to define the soma of a responsive cell. For each ROI, the change in fluorescence relative to baseline (△F/F) was computed based on a baseline period of 10 frames prior to the onset of any stimulation. Significant calcium transients were defined as events of greater than three standard deviations above the mean for a duration of at least 2 frames. Cells with baseline GCaMP6s signals of more than 2 standard deviations above the mean in a given experiment were excluded from analysis, as GCaMP overexpression can lead to aberrant responsiveness (Chen et al., 2013; Packer et al., 2015; Tian et al., 2009). A synaptic connection was scored if the following criteria were met: (1) a latency of at least 2 ms from stimulus onset, (2) occurring in at least three out of four trials, (3) a jitter of less than 14 ms, (4) a rise time from 10% of peak to 90% of peak of less than 10 ms, and (5) the presence of a calcium transient in the photostimulated cell. Detected connections were confirmed to be excitatory in nature by altering the holding potential of the patched cell to demonstrate a reversal potential of ~0 mV. Connectivity rate was defined as then number of pairs with calcium transients in the presynaptic neuron and postsynaptic currents divided by the number of presynaptic neurons with calcium transients stimulated. Funding Information This paper was supported by the following grant: Max Planck Florida Institute to Christopher A Baker, Yishai M Elyada, Andres Parra-Martin, M McLean Bolton. Acknowledgements We would like to thank Alipasha Vaziri for advice on temporal focusing, Jason Christie for consultations on implementing temporal focusing on our specific two-photon setup, David Whitney and Dan Wilson for valuable conversations regarding calcium imaging, David Fitzpatrick for comments on the manuscript, and Asnel Joseph, Nowrin Ahmed, and Laura Conatser for performing stereotaxic AAV injections. GCaMP6s is courtesy of V Jayaraman, R Kerr, D Kim, L Looger, and K Svoboda from the GENIE Project at the Janelia Farm campus of the Howard Hughes Medical Institute. This work was supported by the Max Planck Florida Institute. Additional information Competing interests The authors declare that no competing interests exist. Author contributions CAB, Conceived the study, Designed and built the optical setup, Engineered DNA constructs and performed temporal focusing experiments, Analyzed data, Wrote the manuscript. YME, Designed and built the optical setup, Drafting or revising the article, Contributed unpublished essential data or reagents. AP, Determined intrinsic electrophysiological properties and conducted paired recording experiments, Analyzed data. MMcLB, Conceived the study, Analyzed data, Wrote the manuscript. Ethics Animal experimentation: This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health, and all animals were handled according to protocols approved by the Institutional Animal Care and Use Committee of the Max Planck Florida Institute for Neuroscience. 10.7554/eLife.14193.015Decision letter Häusser Michael Reviewing editorUniversity College London, United KingdomIn the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included. [Editors’ note: a previous version of this study was rejected after peer review, but the authors submitted for reconsideration. The first decision letter after peer review is shown below.] Thank you for choosing to send your work entitled "Single-cell resolution circuit mapping with temporal-focused excitation of soma-targeted channelrhodopsin" for consideration at eLife. Your full submission has been evaluated by Eve Marder as the Senior editor and three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the decision was reached after discussions between the reviewers. Based on our discussions and the individual reviews below, we regret to inform you that the work does not meet the standard for publication in eLife. While the reviewers agreed that the strategy you propose is elegant and potentially very powerful, and could have a major impact on the field of neuroscience, unfortunately they also expressed serious concerns regarding the details of the experiments and the presentation of the results, which dampened enthusiasm for the manuscript. In particular, the section of the manuscript on circuit mapping was felt to be especially weak and would require a major series of additional experiments. As you probably know, eLife has a policy of not asking for significant new experiments as part of a revision, thus resulting in the present decision. Reviewer #1: In this paper, the authors describe an optogenetics method to reach optogenetics activation with single cell resolution. The approach combines the use of temporal focusing with targeted opsins, which restrict the chanelrhodopsin expression to the soma and the apical dendrites. The development of somatic opsins is one of the most promising ways of reaching a true cellular resolution with optogenetics and will surely strongly impact the neuroscience community. However, in my opinion, the manuscript cannot be published in the present form and requires significant major modifications. Introduction: References reporting optical methods using two-photon excitation, diffractive optics and temporal focusing for optogenetics are not extensively cited. Also a state of the art (including eventual references) on alternative approaches to achieve somatic ChR2 targeting will help in appreciating the novelty of the paper. The decision of using Kv2.1 voltage gated potassium channel to confine ChR2 targeting should be discussed more extensively. Results and Discussion: Figure 1 (left): The meaning of this figure is not clear. It seems that the authors want here to characterize the optical resolution of the system. If this is the case, lateral resolution will be better characterized by performing a lateral displacement of the excitation spot along the x and y direction and plotting the corresponding curves (similarly as was done in figure 1, right panel, for axial resolution). The 3D image is very confusing and not necessary. Figure 2—figure supplement 2: Cross sections through the images are needed in order to appreciate the values for lateral and axial FWHM. Figure 2A: The image showing the Alexa 594 distribution (central bottom panel) has a very reduced fluorescence spreading with respect to the central top one: this difference is not justified as the spreading should be comparable in the two cells: authors should probably choose a better example. Figure 2D: The experiment on acute slice has been done only once: this is not enough to support their conclusion more statistics is needed. They should be able to derive for acute slice a figure similar to Figure 2B and C. They need to discuss the effect of the planarity of the dendrites in the experiments: the excitation spot has been moved laterally along the objective focal plane, if the dendritic process is axially tilted this could also induce a decrease in the current (more statistic will enables removing this ambiguity). In order to compare data from non targeted and targeted cells, authors should comment on the time they wait after injections in the two cases. Is this comparable? How long the somatic targeting stays somatic? Is there a critical time window after which the somatic ChR2 starts spreading along dendritic processes? Scale bar should be indicated in the bottom image. Why for the targeted ChR2, data have been taken with larger step? Figure 2F: The authors should better explain how they obtained this figure; how do they define the threshold? Figure 2G: In the caption they write "each position in a map was stimulated with the minimum power that reliably evoked action potential when stimulation was applied to the soma": they should better quantify the meaning of "reliably evoked action potential". Stimulation protocol (pulse duration, pulse frequency) should be indicated in the caption for all the experiments. Figure 2—figure supplement 1: Not needed. Figure 3: The data and procedure reported in this figures needed to be better presented and explained. A picture showing the GCamp6 fluorescence before photostimulation is needed to visualize the distribution of the cells in rest condition. It is not clear if the cell dye-filled and imaged in A is a ChR2 positive cell. If this is the case, authors need to show the current when the photostimulation spot is placed on the cell. The experiment should be repeated more than a single time to be convincing. The construct used in Figure 3 uses ChR2 directly linked to GCamp6: this is a very powerful idea and should be better highlighted. Results section: "[…] owing to the lower efficiency of spike generation by ChR2 in the absence of TF […]" this sentence is wrong. TF does not increase the efficiency of ChR2 excitation but only reduce the out of focus contribution, thus improving axial resolution. "[…] and a reproducible current with appropriate synaptic delay and kinetics […]" this sentence is very vague, authors should define and quantify what is an "appropriate synaptic delay and kinetics". The discussion on the biological results of in Figure 3 should be toned down. The paper is a methodological paper with interesting results and does not need in my opinion a biological conclusion that is not supported by enough data. Reviewer #2: The authors created a new construct that localizes ChR2 to the soma and proximal dendrites of neurons. When combined with two-photon beam shaping methods (e.g. temporal focusing), this should improve the ability to target and stimulate individual densely packed neurons without concurrently activating their neighbors. While the new construct may alleviate some of the concerns typically associated with optical mapping of connectivity (i.e. the inability to precisely stimulate only neurons of choice), the data presented in this manuscript are far too preliminary to make an impact in the field of circuit mapping. Detailed comments: Figure 1: Much more quantification is needed. The important variable for circuit mapping (Figure 3) is whether or not a spike is elicited, rather than the inward current. The authors should determine on what fraction of trials a spike is elicited for each power, for each location. Currently only single-trial raw current-clamp data is shown in Figure 1, but some quantification of this is required, for example: For the final power chosen, for each neuron, what fraction of trials led to a spike when the spot was directly on the soma, and what fraction of trials led to a spike when the spot was directly, vertically, above the neuron (i.e. position iv), which seems to be the most vulnerable position for eliciting unwanted action potentials? What was the final power used for the example shown? 61mW is on the threshold of activating the neuron soma directly (position iii), and 89mW (the next power tested) is on the threshold of activating the neuron when the beam is not directed to the soma (position iv). As far as I can make out, the authors go on to change the protocol later in the paper (Figure 3, "circuit mapping"), using 150ms long pulses in order to generate trains of action potentials. However, all of the analysis in Figure 1 needs to be redone with these experimental parameters, since longer stimulation pulses will increase the probability of unwanted spikes away from the location of light stimulation. What is the latency to action potential for each of the laser powers? Figure 2: In panel 2D, the authors should show an example of a "targeted" neuron (i.e. ChR2 localized to the soma), whilst stimulating at points along the apical dendrite at the same density as that shown for the "non-targeted" neuron. Also, the current elicited in the targeted neuron is here lower than the current elicited for the non-targeted neuron, which contradicts panel E, and is not "representative" – what was the stimulation power used in the two cases? In panel 2G, the interesting variable is the average number of spikes elicited in current clamp and these data would have been more valuable. Figure 3: The image quality needs to be refined, and some of the somata are poorly defined. This applies particularly to the cells that are assumed to be connected. The voltage-clamp traces in panel 3C are single trial data. The authors should show multiple traces for each connection to convince the reader that a true connection is present, rather than an EPSC which happens to coincide with light stimulation. The authors should quantify the calcium signals in all the neurons in the imaged population when a single neuron has been targeted for stimulation (beyond what's shown in Video 1 & 2, which are not informative). Crucially, the authors must show unambiguously that there was only one neuron active on each stimulation trial. Reviewer #3: The authors present a novel combination of two known methods, light shaping and opsin targeting, for the purpose of mapping synaptic connections in vitro. This is in principle a very elegant approach for improving the spatial precision of optogenetic activation, currently a key limitation in the field. However, the manuscript has a rather preliminary flavor (several of the key observations appear to be n = 1). The authors are in the position to provide a major advance here by performing a detailed quantification of how accurate and reliable their method is, using ground truth calibrations. For example, the authors have not quantified how accurate their method is with any paired recordings to prove the connections they find are real. They only state that the average connectivity is similar to that in other experiments in which pairs were directly recorded. Most importantly, the lack of detailed quantification (with mean, SD, and N) needs to be addressed prior to publication. Major comments: 1) There are major details missing in Figure 1. What is the mean action potential reliability and resolution, i.e. the grand average result of Figure 1A across all neurons? What powers were typically used for AP generation at the soma in these experimental conditions? What are the max currents observed? Please provide mean, SD, and N. Note that the figure was not created with the construct that was ultimately used, which is a weakness. 2) How many cells were used to generate the data in Figure 2—figure supplement 2? It appears that some of the differences are statistically borderline and without complete data including the sample size it is difficult to determine the reliability of this result. Also, how did the authors determine the number of significant digits to include? 3) In describing Figure 3, the authors mention that 3 photostimulation trials are performed at each location. Could the authors please show raw trials, perhaps in a lighter shade behind the average, to indicate the reliability of observed connections? 4) Figure 3 uses a different stimulation duration that rest of paper – the photostimulation time has been increased to 150 ms for Figure 3. As this value doesn't match the previous calibrations, it is very difficult to use the data in Figure 2 to calibrate Figure 3. How does the longer duration affect spatial resolution, action potential threshold, etc.? 5) Many of the calcium imaging transients in Figure 3 are quite large, and sometimes double-peaked when there is only one EPSC observed (Figure 3C, bottom row, red square). How do the authors explain the discrepancy between the fact that these long photostimulations (150 ms!) may very well induce more than one action potential, but only one post-synaptic response is observed? Many cortical synapses may depress, but not sufficiently to explain these observations. 6) What is the cutoff for a connection and how reliable is this? For example, in the bottom row of Figure 3C, fourth from the right, there is a large calcium transient and some tiny EPSCs – could these be a weak connection? 7) How often do the authors observe failures to confirm pre-synaptic action potential generation with imaging? They only say "occasionally". Excluding these from analysis could heavily bias estimation of connectivity rates! 8) In Figure 3, the authors photostimulate 192 different locations in a grid-like fashion. They don't aim to zap neurons directly, but rather by shooting at many locations, they hope to hit some neurons by chance. A quick segmentation of the image to find neurons and shoot them directly would improve accuracy, reliability, and potentially even be more efficient! Why do the authors not target neurons initially? 9) Have the authors repeated the experiment shown in Figure 3 more than once? If so, please present some grand average data. 10) Losonczy et al. 2010, cited by the authors, shows effective activation of axons. How can the authors be sure that is not occurring here? Can they provide some presynaptic patch confirmations of any of the connections they see? [Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.] Thank you for submitting your article "Single-cell resolution circuit mapping with temporal-focused excitation of soma-targeted channelrhodopsin" for consideration by eLife. Your article has been reviewed by two peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Eve Marder as the Senior Editor. The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission. We hope you will be able to submit the revised version within two months, so please let us know if you have any questions first. Summary: The development of somatically targeted opsins is one of the most promising ways of achieving true cellular resolution with optogenetics and will surely have a strong impact the neuroscience community. The authors have considerably improved the quality of the manuscript by adding more data and statistical analysis. There is a strong case now for arguing that this new construct is better suited for mapping connectivity in the circuit using targeted optogenetic stimulation. However a few points of the paper still need additional work before the manuscript is ready to proceed towards publication. We encourage the authors to proceed with these final experiments as a matter of urgency, as this is a highly competitive field. Essential revisions: 1) The power levels used in the different experiments are often missing, and this information is crucial to appreciate the spatial resolution achieved in the experiments (e.g. how far are the powers used from the saturation value?). The powers used to evoke a single AP are rather high and latency and jittering are rather long compared to what has been reported in the literature. This point is particularly weak considering that in several parts of the manuscript the authors insist on the "enhanced sensitivity "of the targeted opsin. Overall this implies that the opsins (somatic or not) used in this experiments are not very efficient and may not be suitable for experiments requiring e.g. multi-spot stimulation. Many datapoints e.g. the ones showing cellular resolution, or the connectivity experiments, are only performed using the targeted opsins and it is difficult to appreciate their importance if one can't compare the same experiments performed with the non-targeted version. 2) In order to showcase the advantage of the 'new targeted' construct, it is crucial to include the axial and lateral profile of spike probability also for the 'non-targeted' construct in Figure 2C, D. Please add this quantification to existing panels in this figure. State the power at which these curves were obtained. 3) Figure 1B: line scans to demonstrate somatic targeting are all done along dendritic processes, while no information or data are provided to show the expression confinement along axons. This would be helpful. 4) Figure 1D: "each pixel in the map show the direct current" are the authors plotting the peak current here? Moreover from this map it is difficult to understand the depolarization achieved. The same experiment performed in current clamp would allow us to learn about the spike probability for spot placed out of the target, which ultimately is the key elements to support the necessity of the somatic opsin for the connectivity experiments in Figure 3 or to appreciate the enhanced spatial resolution (see next comment). 5) Figure 3 is nice. The quantification of these connectivity mapping experiments could be included in this figure rather than only in the Results section of the text. For completion, please add an additional example of another such slice experiment in an extra supplementary figure. Also, the information on the stimulation protocol used here is very vague: "each cell was stimulated in series with 2 seconds between stimuli": how many stimuli? What power did they use? How confined is the response if experiments as the ones showed in Figure 1 C-D are done using this protocol? How do these results compare if similar experiments are done with a non-somatic opsin? 6) In the discussion the authors justify the use of high power and long photostimulation power: "We did not take full advantage of the temporal precision capability of TF to fire action potentials in our current study, instead focusing on a screening method that would identify connections without optimizing the amount of power that would fire each potential presynaptic neuron with minimal latency. We therefore chose longer pulses at a power sufficient to fire most neurons and generate trains of action potentials, which would elicit stronger signals with calcium indicators. For experiments requiring temporal precision, the minimization of action potential latency requires optimization of excitation area and laser power" This paper should convince us about the use of a new optogenetic construct, and (as discussed above) a more detailed characterization of the opsin showing the photostimulation area and laser power that enables AP generation with a temporal resolution and precision comparable to what has been achieved in the literature is important and should be carried out. 7) The sentence "Furthermore, these techniques could also be used in vivo, where the enhanced sensitivity of the targeted ChR2 makes it especially attractive" is misleading: in the paper the authors do show that the targeted version is more sensitive than the non-targeted one. But in both cases they use excitation powers much higher than what has been achieved in the literature and demonstrate performances (temporal resolution, latency and jittering) inferior to what has been achieved with ChR2 or C1V1 by other labs. This should be reworded. [Editors' note: further revisions were requested prior to acceptance, as described below.] Thank you for resubmitting your work entitled "Single-cell resolution circuit mapping in mouse brain with temporal-focused excitation of soma-targeted channelrhodopsin" for further consideration at eLife. Your revised article has been favorably evaluated by Eve Marder as the Senior editor, a Reviewing editor, and two reviewers. The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below: 1) There remain concerns about the intensity and duration of the illumination pulses used (150 ms). This suggests that the construct is not very efficient. Why were such long pulses used? Were shorter pulses used in some experiments? The authors should either demonstrate that their construct is also effective in triggering spikes when using shorter pulses, or provide a convincing justification for the use of longer pulses. 2) Please add to the Methods section some of the text that is currently a response to point 1, related to stimulation power needed to excite the cells. ("Our manuscript reports excitation powers of between 15 and 285 mW (for both types of opsins); assuming our spot size to be at least 10 μm wide and 10 μm thick, we are using powers of no greater than 0.2 to 3.8 mW/μm2). Expressing power as mW/μm2 (rather than incident power in hundreds of mW) will be useful for readers. 3) 'Single-cell resolution' is advertised in the title, but is not well supported. We suggest changing the beginning of the title to 'Cellular resolution…'. 4) Please extend the comparison between your results and those of Wu et al. 2013 Plos ONE (since they originated the somatic restriction strategy). 10.7554/eLife.14193.016Author response [Editors’ note: the author responses to the first round of peer review follow.] Reviewer #1: Introduction: References reporting optical methods using two photon excitation, diffractive optics and temporal focusing for optogenetics are not extensively cited. Also a state of the art (including eventual references) on alternative approaches to achieve somatic ChR2 targeting will help in appreciating the novelty of the paper. The decision of using Kv2.1 voltage gated potassium channel to confine ChR2 targeting should be discussed more extensively. We have lengthened the Introduction to provide a larger context for the mapping experiments including additional citations for temporal focusing and two-photon microscopy. A separate paragraph in the Discussion has been included that provides a review of other approaches that have been taken to achieve somatic targeting, and the advantages of using Kv2.1 to achieve opsin subcellular restriction. Results and Discussion: Figure 1 (left): The meaning of this figure is not clear. It seems that the authors want here to characterize the optical resolution of the system. If this is the case, lateral resolution will be better characterized by performing a lateral displacement of the excitation spot along the x and y direction and plotting the corresponding curves (similarly as was done in figure 1, right panel, for axial resolution). The 3D image is very confusing and not necessary. Indeed, the reviewer understood our original aim and provided a better alternative to achieving this end. The original figure referenced has been omitted and replaced with lateral and axial resolution measurements in Figure 2C and 2D. Figure 2—figure supplement 2: Cross sections through the images are needed in order to appreciate the values for lateral and axial FWHM. Fluorescence measurements have been added, now as Figure 1—figure supplement 2. Fitting Gaussian functions to the measurements revealed the full width at half maximum to be even smaller than previously estimated. Figure 2A: The image showing the Alexa 594 distribution (central bottom panel) has a very reduced fluorescence spreading with respect to the central top one: this difference is not justified as the spreading should be comparable in the two cells: authors should probably choose a better example. Two different neurons have been chosen for the representation. Figure 2D: The experiment on acute slice has been done only once: this is not enough to support their conclusion more statistics is needed. They should be able to derive for acute slice a figure similar to Figure 2B and C. Experiments on acute slices were repeated for 10 cells expressing untargeted ChR2 and 13 cells expressing targeted ChR2. Representative data are presented in Figure 1B, with averages in Figure 1C. The data from cultured neurons have been dropped to make room for the new slice experiments. They need to discuss the effect of the planarity of the dendrites in the experiments: the excitation spot has been moved laterally along the objective focal plane, if the dendritic process is axially tilted this could also induce a decrease in the current (more statistic will enables removing this ambiguity). We have tried to clarify that each dendrite under study was followed carefully as its depth changed within the slice, and that stimulation spots were chosen such that the temporal focusing spot contained a complete stretch (>10μm) of dendrite within a single focal plane. In addition, our inclusion of more cells should average out any remaining effects of dendrite planarity, as the reviewer suggests. In order to compare data from non targeted and targeted cells, authors should comment on the time they wait after injections in the two cases. Is this comparable? How long the somatic targeting stays somatic? Is there a critical time window after which the somatic ChR2 starts spreading along dendritic processes? This information was not as explicit or extensive in the original version of the manuscript. It now appears in the Materials and methods section. All preparations were examined between 3 and 4 weeks post-injection. Somatic restriction has been observed at 5 weeks after injection, but we have not examined later time points. Scale bar should be indicated in the bottom image. Why for the targeted ChR2, data have been taken with larger step? As described above, the exact positions along dendrites were chosen in part based on planarity.For the representative example (Figure 1B), we tried to choose a pair of neurons where the stimulation locations were similarly spaced. Figure 2F: The authors should better explain how they obtained this figure; how they define the threshold? The threshold is defined as the minimum intensity that produced action potentials in ten out of ten consecutive trials. This is now explained at multiple points within the text. Figure 2G: In the caption they write "each position in a map was stimulated with the minimum power that reliably evoked action potential when stimulation was applied to the soma": they should better quantify the meaning of "reliably evoked action potential". This is the aforementioned threshold intensity for a given neuron. Stimulation protocol (pulse duration, pulse frequency) should be indicated in the caption for all the experiments. The captions for Figures 1–3 now contain descriptions of the stimulation protocol. Figure 2—figure supplement 1: Not needed. Figure 3: The data and procedure reported in this figures needed to be better presented and explained. A picture showing the GCamp6 fluorescence before photostimulation is needed to visualize the distribution of the cells in rest condition. This figure has been completely redone in the new manuscript. Baseline GCaMP fluorescence is shown in Figure 3A. It is not clear if the cell dye-filled and imaged in A is a ChR2 positive cell. If this is the case, authors need to show the current when the photostimulation spot is placed on the cell. The experiment should be repeated more than a single time to be convincing. In the original manuscript, the cell shown was indeed ChR2-positive. In all our mapping experiments, the patched cells express ChR2 and we use their responsiveness to calibrate the photostimulation power. We have included examples of current traces found with direct stimulation of neuronal somata in Figure 1B. As for repetition of the experiment, we have indeed done this and report average statistics in the text of the Results section (seventh paragraph). The construct used in figure 3 uses ChR2 directly linked to GCamp6, this is a very powerful idea and should be better highlighted. We agree with the reviewer that this is a potentially powerful approach, however, in practice we found that this method actually led to decreased GCaMP responsiveness that contributed to poor signal to noise ratios and suboptimal image quality. We have not completely characterized the reasons for this but suspect that linking high opsin expression to high GCaMP expression on the same construct leads to many cells with GCaMP in their nuclei that no longer exhibit calcium transients. As a result, we turned to the approaches that other groups employ—using a dilution of the GCaMP virus—and combined this with nuclear labeling of neurons making identification easier for stimulation by temporal focusing. Results section: "[…] owing to the lower efficiency of spike generation by ChR2 in the absence of TF […]" this sentence is wrong. TF does not increase the efficiency of ChR2 excitation but only reduce the out of focus contribution, thus improving axial resolution. We thank the reviewer for this correction. The efficiency of ChR2 excitation is quite high as indicated by its two-photon cross section. We intended to refer to the low single-channel conductance of ChR2 and suggest that when the excitation is axially confined, a rapid scanning approach or a light sculpting technique is needed to quickly activate many ChR2 molecules and drive spike generation. We have clarified this in the text and actually expanded the portion of the Results that discusses ChR2 activation by the imaging beam (fourth paragraph). "[…] and a reproducible current with appropriate synaptic delay and kinetics […]" this sentence is very vague, authors should define and quantify what is an "appropriate synaptic delay and kinetics" We define a bona fide postsynaptic response as exhibiting a rise time of less than 10 ms, and having a greater than 2 ms latency from the photostimulation and with less than 14 ms jitter. These criteria are now described in the Materials and methods section (subsection “Calcium imaging and resolution studies”). The discussion on the biological results of in Figure 3 should be toned down. The paper is a methodological paper with interesting results and does not need in my opinion a biological conclusion that is not supported by enough data. We agree, and do not desire to make a biological statement about layer 3-local excitatory connectivity. Rather, we think it is important to provide the reader with a measure of the sensitivity and reliability of this new approach for mapping synaptic connections with the current gold standard: paired patch recording. We have revised the text to clarify the rationale for inclusion of this comparison. Reviewer #2: […] Figure 1: Much more quantification is needed. The important variable for circuit mapping (Figure 3) is whether or not a spike is elicited, rather than the inward current. The authors should determine on what fraction of trials a spike is elicited for each power, for each location. Currently only single-trial raw current-clamp data is shown in Figure 1, but some quantification of this is required, for example: For the final power chosen, for each neuron, what fraction of trials led to a spike when the spot was directly on the soma, and what fraction of trials led to a spike when the spot was directly, vertically, above the neuron (i.e. position iv), which seems to be the most vulnerable position for eliciting unwanted action potentials? What was the final power used for the example shown? 61mW is on the threshold of activating the neuron soma directly (position iii), and 89mW (the next power tested) is on the threshold of activating the neuron when the beam is not directed to the soma (position iv). In response to the reviewer’s concerns we have abandoned the original figure in favor of an analysis examining spike probability for ten neurons over ten trials at axially and laterally displaced locations. These data are now reported in the current manuscript as Figure 2C and 2D. As far as I can make out, the authors go on to change the protocol later in the paper (Figure 3, "circuit mapping"), using 150ms long pulses in order to generate trains of action potentials. However, all of the analysis in Figure 1 needs to be redone with these experimental parameters, since longer stimulation pulses will increase the probability of unwanted spikes away from the location of light stimulation. All experiments in the paper have now been performed with the longer 150 ms stimulation protocol. What is the latency to action potential for each of the laser powers? The average latency for the mapping experiments was 39 ms, which is now mentioned in the text. Figure 2: In panel 2D, the authors should show an example of a "targeted" neuron (i.e. ChR2 localized to the soma), whilst stimulating at points along the apical dendrite at the same density as that shown for the "non-targeted" neuron. Also, the current elicited in the targeted neuron is here lower than the current elicited for the non-targeted neuron, which contradicts panel E, and is not "representative" – what was the stimulation power used in the two cases? The new version of this experiment is presented in Figure 1B. The stimulation density is similar for both the targeted and nontargeted examples, and examining over 10 cells per group allowed us to choose truly representative cells that showcase the heightened current achieved with the targeted construct even with equivalent power (167mW in these cases). In panel 2G, the interesting variable is the average number of spikes elicited in current clamp and these data would have been more valuable. As mentioned previously, spike probability as a function of lateral or axial displacement is now included in Figure 2C and 2D. For our mapping experiments evaluating small synaptic currents in response to stimulation of putative presynaptic cells, however, showing a reduction in the currents evoked by direct stimulation of the dendritic arbor of the patched cell is still relevant. Figure 3: The image quality needs to be refined, and some of the somata are poorly defined. This applies particularly to the cells that are assumed to be connected. New images of cells are provided in Figure 3A; a fluorescent nuclear label and better GCaMP expression allowed for more defined images of all cells. The voltage-clamp traces in panel 3C are single trial data. The authors should show multiple traces for each connection to convince the reader that a true connection is present, rather than an EPSC which happens to coincide with light stimulation. All four voltage clamp traces used to determine the presence or absence of a synaptic connection are now shown in Figure 3C. The authors should quantify the calcium signals in all the neurons in the imaged population when a single neuron has been targeted for stimulation (beyond what's shown in Video 1 & 2, which are not informative). Crucially, the authors must show unambiguously that there was only one neuron active on each stimulation trial. These data are now included in Figure 3B, where the calcium signal is shown for each targeted neuron in a given experiment. Every connection is unequivocally associated with only one active neuron. Reviewer #3: […] 1) There are major details missing in Figure 1. What is the mean action potential reliability and resolution, i.e. the grand average result of Figure 1A across all neurons? What powers were typically used for AP generation at the soma in these experimental conditions? What are the max currents observed? Please provide mean, SD, and N. Note that the figure was not created with the construct that was ultimately used, which is a weakness. The resolution of spike generation in 10 cells expressing the targeted construct is now presented in Figures 2C and 2D. These measurements were determined at the threshold power for each particular cell, the distribution of which is shown in Figure 2B. The average currents at each power are shown in Figure 2A. There was no difference between targeted and untargeted cells in the average current necessary to evoke action potentials (i.e., rheobase), only a difference in the amount of stimulation power required to evoke such current. 2) How many cells were used to generate the data in Figure 2—figure supplement 2? It appears that some of the differences are statistically borderline and without complete data including the sample size it is difficult to determine the reliability of this result. Also, how did the authors determine the number of significant digits to include? These data are now in Figure 2—figure supplement 2, and include N for all measurements. Significant digits are determined by the measurement limits of the Clampex patch clamp software. 3) In describing Figure 3, the authors mention that 3 photostimulation trials are performed at each location. Could the authors please show raw trials, perhaps in a lighter shade behind the average, to indicate the reliability of observed connections? All raw trials are now shown in the new Figure 3C, with calcium traces overlaid in a lighter shade and electrophysiological traces displayed adjacent to each other. 4) Figure 3 uses a different stimulation duration that rest of paper – the photostimulation time has been increased to 150 ms for Figure 3. As this value doesn't match the previous calibrations, it is very difficult to use the data in Figure 2 to calibrate Figure 3. How does the longer duration affect spatial resolution, action potential threshold, etc.? All data in the paper are now derived from 150ms stimulation. 5) Many of the calcium imaging transients in figure 3 are quite large, and sometimes double-peaked when there is only one EPSC observed (Figure 3C, bottom row, red square). How do the authors explain the discrepancy between the fact that these long photostimulations (150 ms!) may very well induce more than one action potential, but only one post-synaptic response is observed? Many cortical synapses may depress, but not sufficiently to explain these observations. Indeed, the longer stimulation protocol was used in hopes of generating trains of presynaptic action potentials. While single postsynaptic currents are observed (some neurons only generate one action potential regardless of stimulation intensity), we frequently saw large calcium transients and trains of postsynaptic currents, which are shown in the new Figure 3C. 6) What is the cutoff for a connection and how reliable is this? For example, in the bottom row of Figure 3C, fourth from the right, there is a large calcium transient and some tiny EPSCs – could these be a weak connection? In the original version of the manuscript, the currents described by the reviewer did not occur across multiple trials. In the new experiments reported here, we require that events occur in at least three out of four trials and with less than 14 ms jitter and a greater than 2 ms latency from the photostimulation. These criteria are now described in the Materials and methods section (subsection “Calcium imaging and connection analysis”). 7) How often do the authors observe failures to confirm pre-synaptic action potential generation with imaging? They only say "occasionally". Excluding these from analysis could heavily bias estimation of connectivity rates! The exact numbers of excluded connections from the current study are now described in the Results. For the majority of cells (5 out of 7 in the current study), every connection found could be associated with a somatic calcium transient. Out of the 36 connections identified across all cells in the present study, 3 were excluded due to lack of a somatic calcium signal. 8) In Figure 3, the authors photostimulate 192 different locations in a grid-like fashion. They don't aim to zap neurons directly, but rather by shooting at many locations, they hope to hit some neurons by chance. A quick segmentation of the image to find neurons and shoot them directly would improve accuracy, reliability, and potentially even be more efficient! Why do the authors not target neurons initially? We originally hoped that baseline GCaMP fluorescence would be sufficient to target neurons as the reviewer suggests, however, we found that many neurons with significant calcium responses to TF stimulation were impossible to identify under baseline imaging conditions. We agree with the reviewer that segmentation would be preferable, and developed a new construct in which neuronal nuclei were fluorescently labeled for easy targeting for TF stimulation. All mapping studies in the current version of the paper use this construct. 9) Have the authors repeated the experiment shown in Figure 3 more than once? If so, please present some grand average data. The data for seven different mapping experiments are now described in the Results section (seventh paragraph). 10) Losonczy et al. 2010, cited by the authors, shows effective activation of axons. How can the authors be sure that is not occurring here? Can they provide some presynaptic patch confirmations of any of the connections they see? Targeted ChR2 expression is not seen in the axons of dye-filled cultured neurons, which is consistent with the known exclusion of Kv2.1 from axons beyond the initial segment. [Editors’ note: the author responses to the re-review follow.] Essential revisions: 1) The power levels used in the different experiments are often missing, and this information is crucial to appreciate the spatial resolution achieved in the experiments (e.g. how far are the powers used from the saturation value?). The powers used to evoke a single AP are rather high and latency and jittering are rather long compared to what has been reported in the literature. This point is particularly weak considering that in several parts of the manuscript the authors insist on the "enhanced sensitivity "of the targeted opsin. Overall this implies that the opsins (somatic or not) used in this experiments are not very efficient and may not be suitable for experiments requiring e.g. multi-spot stimulation. Many datapoints e.g. the ones showing cellular resolution, or the connectivity experiments, are only performed using the targeted opsins and it is difficult to appreciate their importance if one can't compare the same experiments performed with the non-targeted version. The power levels used in most experiments were calibrated for each cell to generate a single action potential; we chose this approach to compare resolution and sensitivity across multiple cells that may differ in absolute levels of opsin expression. We have made an extra effort to explicitly mention this rationale as well as the average powers chosen for each experiment; for sample mapping experiments in Figures 3 and 4, we have also included specific power levels in the figure legends. With temporal focusing excitation distributed across a volume, the power levels for opsin excitation are expected to be higher than those used when scanning a much smaller diffraction-limited spot. In earlier papers using temporal focusing stimulation of ChR2, Andrasfalvy et al. (PNAS, 2010) reported action potential generation anywhere with 300-400 mW power at the sample plane, and Losonczy et al (Nature Neurosci., 2010) report power levels between 50 and 500 mW. Our manuscript reports excitation powers of between 15 and 285 mW (for both types of opsins); assuming our spot size to be at least 10μm wide and 10μm thick, we are using powers of no greater than 0.2 to 3.8 mW/μm2. This approaches but does not surpass excitation powers of 0.45 mW/μm2 reported by the Emiliani group, but differences in opsin delivery and expression may account for some of the discrepancy. Finally, Rickgauer et al. (Nature Neurosci., 2014) were able to use less than 100 mW over the entire temporal focusing disc when employing C1V1, which could be due to the enhanced optical response of that opsin relative to ChR2. We agree with the reviewers that additional comparisons between nontargeted and targeted opsins would reveal the utility of a somatic targeting approach. We have taken this approach to heart and included new comparisons in Figure 2, and also added a new Figure 2—figure supplement 3 and Figure 4 (see below). 2) In order to showcase the advantage of the 'new targeted' construct, it is crucial to include the axial and lateral profile of spike probability also for the 'non-targeted' construct in Figure 2C, D. Please add this quantification to existing panels in this figure. State the power at which these curves were obtained. The manuscript now has a new version of Figure 2 with the nontargeted resolutions incorporated (see also sixth paragraph of Results). As alluded to above, the power levels for all the resolution experiments (including both Figure 1 and Figure 2) vary from cell to cell and are set to be the minimum power that generates a single action potential in ten out of ten trials. The average powers for these experiments are reported in third paragraph of Results and a mistaken power level was corrected. 3) Figure 1B: line scans to demonstrate somatic targeting are all done along dendritic processes, while no information or data are provided to show the expression confinement along axons. This would be helpful. We conducted similar optical stimulation experiments on axonal processes, added more detailed imaging, and included the results as Figure 1—figure supplement 3 and in the Results (fourth paragraph). 4) Figure 1D: "each pixel in the map show the direct current" are the authors plotting the peak current here? Moreover from this map it is difficult to understand the depolarization achieved. The same experiment performed in current clamp would allow us to learn about the spike probability for spot placed out of the target, which ultimately is the key elements to support the necessity of the somatic opsin for the connectivity experiments in Figure 3 or to appreciate the enhanced spatial resolution (see next comment). Indeed, we are plotting the peak current amplitudes in Figure 1D. This voltage clamp experiment remains an important demonstration of the extent of direct currents that overwhelm smaller synaptic events in mapping experiments, which can be seen now in the new Figure 4. We agree that the spike probability at off target locations is worthwhile information and were able to investigate this in the context of actual connectivity mapping experiments by extracting information from our previous data as well as new experiments with nontargeted opsin. When using a nontargeted construct, we found an increased probability of calcium transient generation in off-target cells when stimulating separate cells in the slice. Moreover, the average distance between an off-target activated cell and the stimulation point was greater when using nontargeted ChR2. These observations are now reported in the Results. 5) Figure 3 is nice. The quantification of these connectivity mapping experiments could be included in this figure rather than only in the Results section of the text. For completion, please add an additional example of another such slice experiment in an extra supplementary figure. Also, the information on the stimulation protocol used here is very vague: "each cell was stimulated in series with 2 seconds between stimuli": how many stimuli? What power did they use? How confined is the response if experiments as the ones showed in Figure 1 C-D are done using this protocol? How do these results compare if similar experiments are done with a non-somatic opsin? We appreciate the reviewer’s comments about the new figure, and have incorporated more quantification from multiple connectivity experiments into the figure legend, as requested. We have also included another example slice experiment as Figure 3—figure supplement 1. We have also tried to explain in the figure legend that each cell is stimulated sequentially at the onset of every other imaging frame (i.e., 2.4 seconds after the previous cell in the field). After going through all the cells in the field (40-80 cells), the entire protocol is repeated for three additional iterations to produce the results in the figures. We have reported the average power value across the mapping experiments seventh paragraph of Results and the specific power levels for each representative experiment within the figure legend. Finally, we appreciate the suggestion to perform mapping experiments with nontargeted ChR2, and have included an example as Figure 4. This allowed us to demonstrate unintended activation of off-target cells as well as the extent to which direct activation of the dendritic arbor occurs and can make interrogation of local connectivity problematic. 6) In the discussion the authors justify the use of high power and long photostimulation power: "We did not take full advantage of the temporal precision capability of TF to fire action potentials in our current study, instead focusing on a screening method that would identify connections without optimizing the amount of power that would fire each potential presynaptic neuron with minimal latency. […] For experiments requiring temporal precision, the minimization of action potential latency requires optimization of excitation area and laser power" This paper should convince us about the use of a new optogenetic construct, and (as discussed above) a more detailed characterization of the opsin showing the photostimulation area and laser power that enables AP generation with a temporal resolution and precision comparable to what has been achieved in the literature is important and should be carried out. Our goal with this manuscript was to provide a baseline system from which to base further refinements of optogenetic mapping experiments and to establish the utility of somatic opsin targeting for such an approach. We therefore focused on greater characterization of spatial resolution of the targeted ChR2 and less on temporal precision. Nevertheless, we now include data that show action potential latency consistent with that achieved in the literature can be obtained by increasing stimulation power (Figure 2—figure supplement 3). Our mapping experiments in Figures 3 and 4 also demonstrate that the advantages in spatial resolution with the targeted construct are maintained at higher stimulation powers that produce shorter latencies. 7) The sentence "Furthermore, these techniques could also be used in vivo, where the enhanced sensitivity of the targeted ChR2 makes it especially attractive" is misleading: in the paper the authors do show that the targeted version is more sensitive than the non-targeted one. But in both cases they use excitation powers much higher than what has been achieved in the literature and demonstrate performances (temporal resolution, latency and jittering) inferior to what has been achieved with ChR2 or C1V1 by other labs. This should be reworded. We altered the statement to focus more on spatial resolution (final paragraph of Discussion). [Editors' note: further revisions were requested prior to acceptance, as described below.] The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below: 1) There remain concerns about the intensity and duration of the illumination pulses used (150 ms). This suggests that the construct is not very efficient. Why were such long pulses used? Were shorter pulses used in some experiments? The authors should either demonstrate that their construct is also effective in triggering spikes when using shorter pulses, or provide a convincing justification for the use of longer pulses. We implemented a longer stimulation protocol as this provided the greatest probability of eliciting trains of action potentials, leading to increased calcium signals in mapping experiments. In an earlier version of our manuscript, we characterized responses for both types of opsins using a shorter stimulation time (32 ms). We have now included some of these data as an additional figure supplement (Figure 2—figure supplement 3) and added appropriate commentary in the text (fourth paragraph of Results). Moreover, the ability to reduce latency to less than 10 ms with increased power (Figure 2—figure supplement 2) suggests that even shorter stimulation pulses may still be effective. 2) Please add to the Methods section some of the text that is currently a response to point 1, related to stimulation power needed to excite the cells. ("Our manuscript reports excitation powers of between 15 and 285 mW (for both types of opsins); assuming our spot size to be at least 10 μm wide and 10 μm thick, we are using powers of no greater than 0.2 to 3.8 mW/μm2). Expressing power as mW/μm2 (rather than incident power in hundreds of mW) will be useful for readers. We have now expressed all powers in the text and figures as mW/μm2, and added an explanatory note to the Materials and methods section (subsection “Stimulation and resolution studies”). 3) 'Single-cell resolution' is advertised in the title, but is not well supported. We suggest changing the beginning of the title to 'Cellular resolution…'. We have made the requested change. 4) Please extend the comparison between your results and those of Wu et al. 2013 Plos ONE (since they originated the somatic restriction strategy). We extended the discussion of the Wu paper, mentioning their demonstration of somatic restriction by physiologic criteria (second paragraph of Discussion). As they measured the physiologic responses of cells under very different conditions than in our study, it is difficult to directly compare their estimates of opsin restriction with our own. ==== Refs References Andrasfalvy BK Zemelman BV Tang J Vaziri A 2010 Two-photon single-cell optogenetic control of neuronal activity by sculpted light PNAS 107 11981 11986 10.1073/pnas.1006620107 20543137 Chen TW Wardill TJ Sun Y Pulver SR Renninger SL Baohan A Schreiter ER Kerr RA Orger MB Jayaraman V Looger LL Svoboda K Kim DS 2013 Ultrasensitive fluorescent proteins for imaging neuronal activity Nature 499 295 300 10.1038/nature12354 23868258 Dal Maschio M Difato F Beltramo R Blau A Benfenati F Fellin T 2010 Simultaneous two-photon imaging and photo-stimulation with structured light illumination Optics Express 18 18720 18731 10.1364/OE.18.018720 20940765 Fino E Yuste R 2011 Dense inhibitory connectivity in neocortex Neuron 69 1188 1203 10.1016/j.neuron.2011.02.025 21435562 Garrido JJ Giraud P Carlier E Fernandes F Moussif A Fache MP Debanne D Dargent B 2003 A targeting motif involved in sodium channel clustering at the axonal initial segment Science 300 2091 2094 10.1126/science.1085167 12829783 Grubb MS Burrone J 2010 Channelrhodopsin-2 localised to the axon initial segment PLoS One 5 e14193 10.1371/journal.pone.0013761 21152024 Jensen CS Watanabe S Rasmussen HB Schmitt N Olesen SP Frost NA Blanpied TA Misonou H 2014 Specific sorting and post-Golgi trafficking of dendritic potassium channels in living neurons Journal of Biological Chemistry 289 10566 10581 10.1074/jbc.M113.534495 24569993 Klapoetke NC Murata Y Kim SS Pulver SR Birdsey-Benson A Cho YK Morimoto TK Chuong AS Carpenter EJ Tian Z Wang J Xie Y Yan Z Zhang Y Chow BY Surek B Melkonian M Jayaraman V Constantine-Paton M Wong GK Boyden ES 2014 Independent optical excitation of distinct neural populations Nature Methods 11 338 346 10.1038/nmeth.2836 24509633 Lewis TL Mao T Svoboda K Arnold DB 2009 Myosin-dependent targeting of transmembrane proteins to neuronal dendrites Nature Neuroscience 12 568 576 10.1038/nn.2318 19377470 Lim ST Antonucci DE Scannevin RH Trimmer JS 2000 A novel targeting signal for proximal clustering of the Kv2.1 K+ channel in hippocampal neurons Neuron 25 385 397 10.1016/S0896-6273(00)80902-2 10719893 Lin JY Knutsen PM Muller A Kleinfeld D Tsien RY 2013 ReaChR: a red-shifted variant of channelrhodopsin enables deep transcranial optogenetic excitation Nature Neuroscience 16 1499 1508 10.1038/nn.3502 23995068 Losonczy A Zemelman BV Vaziri A Magee JC 2010 Network mechanisms of theta related neuronal activity in hippocampal CA1 pyramidal neurons Nature Neuroscience 13 967 972 10.1038/nn.2597 20639875 Mattis J Tye KM Ferenczi EA Ramakrishnan C O'Shea DJ Prakash R Gunaydin LA Hyun M Fenno LE Gradinaru V Yizhar O Deisseroth K 2012 Principles for applying optogenetic tools derived from direct comparative analysis of microbial opsins Nature Methods 9 159 172 10.1038/nmeth.1808 22179551 McCarthy KD de Vellis J 1980 Preparation of separate astroglial and oligodendroglial cell cultures from rat cerebral tissue Journal of Cell Biology 85 890 902 10.1083/jcb.85.3.890 6248568 Nikolenko V Poskanzer KE Yuste R 2007 Two-photon photostimulation and imaging of neural circuits Nature Methods 4 943 950 10.1038/nmeth1105 17965719 Nikolenko V Watson BO Araya R Woodruff A Peterka DS Yuste R 2008 SLM microscopy: Scanless two-photon imaging and photostimulation with spatial light modulators Frontiers in Neural Circuits 2 e14193 10.3389/neuro.04.005.2008 Oron D Tal E Silberberg Y 2005 Scanningless depth-resolved microscopy Optics Express 13 1468 1476 10.1364/OPEX.13.001468 19495022 Oron D Papagiakoumou E Anselmi F Emiliani V 2012 Two-photon optogenetics Progress in Brain Research 196 119 143 10.1016/B978-0-444-59426-6.00007-0 22341324 Packer AM Yuste R 2011 Dense, unspecific connectivity of neocortical parvalbumin-positive interneurons: a canonical microcircuit for inhibition? Journal of Neuroscience 31 13260 13271 10.1523/JNEUROSCI.3131-11.2011 21917809 Packer AM Peterka DS Hirtz JJ Prakash R Deisseroth K Yuste R 2012 Two-photon optogenetics of dendritic spines and neural circuits Nature Methods 9 1202 1205 10.1038/nmeth.2249 23142873 Packer AM Russell LE Dalgleish HW Häusser M 2015 Simultaneous all-optical manipulation and recording of neural circuit activity with cellular resolution in vivo Nature Methods 12 140 146 10.1038/nmeth.3217 25532138 Papagiakoumou E de Sars V Oron D Emiliani V 2008 Patterned two-photon illumination by spatiotemporal shaping of ultrashort pulses Optics Express 16 22039 22047 10.1364/OE.16.022039 19104638 Papagiakoumou E de Sars V Emiliani V Oron D 2009 Temporal focusing with spatially modulated excitation Optics Express 17 5391 5401 10.1364/OE.17.005391 19333304 Papagiakoumou E Anselmi F Bègue A de Sars V Glückstad J Isacoff EY Emiliani V 2010 Scanless two-photon excitation of channelrhodopsin-2 Nature Methods 7 848 854 10.1038/nmeth.1505 20852649 Papagiakoumou E Bègue A Leshem B Schwartz O Stell BM Bradley J Oron D Emiliani V 2013 Functional patterned multiphoton excitation deep inside scattering tissue Nature Photonics 7 274 278 10.1038/nphoton.2013.9 Prakash R Yizhar O Grewe B Ramakrishnan C Wang N Goshen I Packer AM Peterka DS Yuste R Schnitzer MJ Deisseroth K 2012 Two-photon optogenetic toolbox for fast inhibition, excitation and bistable modulation Nature Methods 9 1171 1179 10.1038/nmeth.2215 23169303 Rickgauer JP Tank DW 2009 Two-photon excitation of channelrhodopsin-2 at saturation PNAS 106 15025 15030 10.1073/pnas.0907084106 19706471 Rickgauer JP Deisseroth K Tank DW 2014 Simultaneous cellular-resolution optical perturbation and imaging of place cell firing fields Nature Neuroscience 17 1816 1824 10.1038/nn.3866 25402854 Tian L Hires SA Mao T Huber D Chiappe ME Chalasani SH Petreanu L Akerboom J McKinney SA Schreiter ER Bargmann CI Jayaraman V Svoboda K Looger LL 2009 Imaging neural activity in worms, flies and mice with improved GCaMP calcium indicators Nature Methods 6 875 881 10.1038/nmeth.1398 19898485 Trimmer JS 1991 Immunological identification and characterization of a delayed rectifier K+ channel polypeptide in rat brain PNAS 88 10764 10768 10.1073/pnas.88.23.10764 1961744 Trimmer JS 2015 Subcellular localization of K+ channels in mammalian brain neurons: remarkable precision in the midst of extraordinary complexity Neuron 85 238 256 10.1016/j.neuron.2014.12.042 25611506 Wu C Ivanova E Zhang Y Pan ZH 2013 rAAV-mediated subcellular targeting of optogenetic tools in retinal ganglion cells in vivo PLoS One 8 e14193 10.1371/journal.pone.0066332 Yizhar O Fenno LE Prigge M Schneider F Davidson TJ O'Shea DJ Sohal VS Goshen I Finkelstein J Paz JT Stehfest K Fudim R Ramakrishnan C Huguenard JR Hegemann P Deisseroth K 2011 Neocortical excitation/inhibition balance in information processing and social dysfunction Nature 477 171 178 10.1038/nature10360 21796121 Zhang Z Feng J Wu C Lu Q Pan Z-H 2015 Targeted expression of channelrhodopsin-2 to the axon initial segment alters the temporal firing properties of retinal ganglion cells PLOS One 10 e14193 10.1371/journal.pone.0142052 Zhu G van Howe J Durst M Zipfel W Xu C 2005 Simultaneous spatial and temporal focusing of femtosecond pulses Optics Express 13 2153 2159 10.1364/OPEX.13.002153 19495103

PMC005xxxxxx/PMC5001838.txt

==== Front eLifeElifeeLifeeLifeeLife2050-084XeLife Sciences Publications, Ltd 274728981844910.7554/eLife.18449Biophysics and Structural BiologyComputational and Systems BiologyShort ReportAntiparallel protocadherin homodimers use distinct affinity- and specificity-mediating regions in cadherin repeats 1-4 Nicoludis John M 12Vogt Bennett E 2Green Anna G 3Schärfe Charlotta PI 34Marks Debora S 3Gaudet Rachelle http://orcid.org/0000-0002-9177-054X2*1 Department of Chemistry and Chemical Biology, Harvard University, Cambridge, United States2 Department of Molecular and Cellular Biology, Harvard University, Cambridge, United States3 Department of Systems Biology, Harvard Medical School, Boston, United States4 Applied Bioinformatics, Department of Computer Science, University of Tübingen, Tübingen, GermanyWeis William I Reviewing editorStanford University, United Statesgaudet@mcb.harvard.edu29 7 2016 2016 5 e1844906 6 2016 28 7 2016 © 2016, Nicoludis et al2016Nicoludis et alThis article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.Protocadherins (Pcdhs) are cell adhesion and signaling proteins used by neurons to develop and maintain neuronal networks, relying on trans homophilic interactions between their extracellular cadherin (EC) repeat domains. We present the structure of the antiparallel EC1-4 homodimer of human PcdhγB3, a member of the γ subfamily of clustered Pcdhs. Structure and sequence comparisons of α, β, and γ clustered Pcdh isoforms illustrate that subfamilies encode specificity in distinct ways through diversification of loop region structure and composition in EC2 and EC3, which contains isoform-specific conservation of primarily polar residues. In contrast, the EC1/EC4 interface comprises hydrophobic interactions that provide non-selective dimerization affinity. Using sequence coevolution analysis, we found evidence for a similar antiparallel EC1-4 interaction in non-clustered Pcdh families. We thus deduce that the EC1-4 antiparallel homodimer is a general interaction strategy that evolved before the divergence of these distinct protocadherin families. DOI: http://dx.doi.org/10.7554/eLife.18449.001 eLife digest As the brain develops, nerve cells or neurons connect with one another to form complex networks. These connections form between branch-like structures, called dendrites, that project from the cell body of each neuron. To prevent unneeded connections from forming, dendrites that belong to the same neuron need a way to recognize and avoid one another. A family of proteins called protocadherins supports this process of self-avoidance. Protocadherins have three main parts or domains: an extracellular domain that faces outwards away from the cell, a transmembrane domain that sits within the cell’s surface membrane and an intracellular domain that faces into the cell’s interior. There are two major groups of protocadherins – clustered and non-clustered – and the former are responsible for the self-avoidance behavior between dendrites. Clustered protocadherins in turn comprise three subfamilies, each of which consists of multiple variants with slightly different structures (known as isoforms). The particular set of protocadherin isoforms that a neuron displays on its surface distinguishes that neuron from all others, a little like a barcode. When two dendrites meet, the protocadherins in their membranes come into contact with one another. If both dendrites come from the same neuron and therefore possess identical sets of protocadherins, then all protocadherins can form two-subunit complexes containing one copy of the same isoform from each dendrite. These complexes are called homodimers and their formation acts as a signal that informs the cell that it has encountered one of its own dendrites and should therefore not establish a connection. By using X-rays to determine the structure of a crystallized protocadherin fragment down to the level of its individual atoms, Nicoludis et al. now reveal exactly how clustered protocadherins form homodimers. The results show that each protocadherin subfamily uses a slightly different type of interaction due to differences in the structure of their extracellular domains. The next challenge is to identify the signaling cascade that is triggered by the formation of clustered protocadherin homodimers, and to work out how activation of this cascade prevents a permanent connection from forming. In addition, the results of Nicoludis et al. predict that some non-clustered protocadherins form dimers with a similar architecture to that of clustered protocadherins. This possibility should also be tested experimentally. DOI: http://dx.doi.org/10.7554/eLife.18449.002 Author Keywords protocadherinscell adhesionsequence coevolutioncrystallographyResearch Organism Humanhttp://dx.doi.org/10.13039/100000001National Science FoundationDGE1144152Green Anna G http://dx.doi.org/10.13039/100000002National Institutes of HealthNCRR 1S10RR028832-01Marks Debora S http://dx.doi.org/10.13039/100000057National Institute of General Medical SciencesGM106303Marks Debora S National Defense Science and Engineering Graduate FellowshipNicoludis John M http://dx.doi.org/10.13039/100000057National Institute of General Medical SciencesP41 GM103403Gaudet Rachelle http://dx.doi.org/10.13039/100000002National Institutes of HealthS10 RR029205Gaudet Rachelle The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.elife-xml-version2.5Author impact statementClustered and non-clustered protocadherins form antiparallel homodimers in which distinct regions of the extended interface demonstrate a division of labor between driving affinity and determining specificity. ==== Body Introduction Protocadherins (Pcdhs) encompass about 70% of the cadherin superfamily in mammals, and are involved in cell adhesion in the nervous system of higher animals (Hulpiau and van Roy, 2011; Hulpiau et al., 2013; Keeler et al., 2015; Sotomayor et al., 2014). Pcdhs segregate into two groups: clustered (expressed from a large gene cluster) and non-clustered. Clustered Pcdhs, comprising the α, β, and γ subfamilies, mediate neuronal survival, self-avoidance and self/nonself discrimination, and promote dendritic arborization in neuronal cells (Emond and Jontes, 2008; Garrett et al., 2012; Kostadinov and Sanes, 2015; Ledderose et al., 2013; Lefebvre et al., 2012; Molumby et al., 2016; Suo et al., 2012; Wang et al., 2002; Weiner et al., 2005). Clustered Pcdhs thus function analogously to insect Dscam isoforms to mediate neuronal identity (Zipursky and Sanes, 2010). Clustered Pcdhs are also broadly involved in synapse formation and maintenance, neuronal connectivity and neuronal survival (Hayashi and Takeichi, 2015; Keeler et al., 2015). Non-clustered Pcdhs play key roles in neuronal development (Keeler et al., 2015; Kim et al., 2011). For example, Pcdh7 is involved in germ layer differentiation (Rashid et al., 2006; Yoshida, 2003), and Pcdh1 and Pcdh8 mediate cell sorting and migration during gastrulation (Kim et al., 1998; Kuroda et al., 2002). Both clustered and non-clustered Pcdhs control these phenotypes through homophilic interactions of their extracellular cadherin (EC) repeat domains (Hirano et al., 1999; Hoshina et al., 2013; Kim et al., 1998; Kuroda et al., 2002; Schreiner and Weiner, 2010; Thu et al., 2014; Yamagata et al., 1999; Yoshida, 2003). Clustered Pcdh subfamilies show distinct phenotypes. In zebrafish, α and γ Pcdhs express in overlapping but distinct brain regions (Biswas et al., 2012). In mammals, α Pcdhs regulate sorting of olfactory sensory neuron axons into glomeruli, serotonergic axon maturation, and dendritic patterning in CA1 pyramidal neurons (Hasegawa et al., 2012, 2008; Katori et al., 2009; Suo et al., 2012). The γ subfamily is important for self/non-self discrimination in retinal starburst amacrine cells and Purkinje neurons (Kostadinov and Sanes, 2015; Lefebvre et al., 2012). Thus, available data suggest that the different Pcdh subfamilies may function independently or cooperatively, perhaps depending on the brain region and/or neuronal cell type. Our recent PcdhγA1 and PcdhγC3 EC1-3 structures revealed dimer interactions between EC2 and EC3 (Nicoludis et al., 2015), consistent with previous biochemical and bioinformatics data (Schreiner and Weiner, 2010; Wu, 2005). Using sequence co-evolution analysis, we predicted intersubunit EC1-EC4 interactions, and proposed that clustered Pcdhs form extended antiparallel homodimers engaging EC1-4. A complementary biochemical and structural study arrived at a very similar docking model (Rubinstein et al., 2015), which was recently confirmed for α and β clustered Pcdhs (Goodman et al., 2016). We determined the crystal structure of PcdhγB3 EC1-4, the first full antiparallel dimer for a γ isoform. We analyzed the clustered Pcdhs structures in light of biological, biochemical and evolutionary data to further resolve how clustered Pcdhs encode specificity. We describe how structural differences between the α, β and γ subfamilies generate distinct modes of specificity encoding. We also provide evidence that the EC1/EC4 and EC2/EC3 interfaces are functionally different: EC1/EC4 provides nonselective dimerization affinity while EC2/EC3 is generally responsible for enforcing specificity. Finally, we extend our sequence coevolution analysis to the non-clustered Pcdhs and provide evidence that the EC1-4 interaction is broadly used by Pcdhs. Results and discussion Structure of the PcdhγB3 EC1-4 extended antiparallel dimer The in vitro-refolded recombinantly-expressed PcdhγB3 EC1-4 (47 kDa) yielded two peaks on size exclusion chromatography (SEC; Figure 1—figure supplement 1). Based on multi-angle light scattering (MALS; Figure 1—figure supplement 1), peak 1 was wide and polydisperse (~200–300 kDa). Peak 2 was monodisperse at 80 kDa – consistent with a dimer – and readily yielded a crystal structure (Figure 1—source data 1). As expected, each EC forms a seven-stranded Greek key β-sandwich motif (Figure 1A), similarly to other clustered Pcdh structures (Goodman et al., 2016; Nicoludis et al., 2015; Rubinstein et al., 2015). Notably, EC4 has a unique β-strand arrangement compared to EC1-EC3 (Figure 1B,C) and all known cadherin repeat structures. Strand β1a is extended by 4–5 residues, while β1b is correspondingly shortened. Additionally, while in EC1-3 strand β2 splits into β2a and β2b, interacting with strands β5 and β1a, respectively, in EC4 it forms a continuous strand interacting with both simultaneously. This distinct structural feature contributes to intersubunit EC1/EC4 interactions (see below).10.7554/eLife.18449.003Figure 1. PcdhγB3 EC1-4 extended antiparallel dimer relies on unusual EC4 β-strand arrangement and is similar to other clustered Pcdh dimers. (A) Structure of the PcdhγB3 EC1-4 antiparallel dimer, with each EC a different shade of blue and the Ca2+ ions in grey. (B) Superposition of PcdhγB3 EC2 and EC4 highlighting the differences in β-strands 1 and 2. (C) Comparison of the canonical cadherin (top) and EC4 (bottom) β-strand arrangement. (D) The structures of Pcdh dimers α4 EC1-4, α7 EC1-5, β6 EC1-4, and β8 EC1-4 (grey) were superimposed using the dimeric EC2-3 region onto γB3 EC1-4 (blue), illustrating variations in twist/corkscrew. (E) The EC4 domains of clustered Pcdh structures (colored as labeled) were superimposed, highlighting EC1 position differences that correlate with subfamily. Point of view (eye symbol) shown in (D). See Figure 1—figure supplements 1–4. DOI: http://dx.doi.org/10.7554/eLife.18449.003 10.7554/eLife.18449.004Figure 1—source data 1. Statistics for PcdhγB3 EC1-4 structure. DOI: http://dx.doi.org/10.7554/eLife.18449.004 10.7554/eLife.18449.005Figure 1—figure supplement 1. PcdhγB3 refolding yields two species, one of which corresponds to monodisperse dimeric protein. (A) SEC profile of refolded PcdhγB3 run on a Superdex 200 16/60 column, with the two collected peak fractions indicated. (B) SEC-MALS profile of Peak 1 run on a Superdex S200 10/300 column was broad and polydisperse. (C) SEC-MALS profile of Peak 2 run on a Superdex S200 10/300 column was monodisperse at a molecular weight of ~80 kDa, consistent with a dimer (monomeric molecular weight 47 kDa). DOI: http://dx.doi.org/10.7554/eLife.18449.005 10.7554/eLife.18449.006Figure 1—figure supplement 2. Protocadherins and non-classical cadherins have a distribution of orientation between repeat pairs that is distinct from classical cadherins. Distribution of tilt and azimuthal angles from adjacent EC repeat pairs of classical (black), Pcdh15 and Cdh23 (blue) and clustered Pcdh (red). (Inset) The orientation of adjacent EC repeats was defined by the tilt and azimuthal rotation of the EC domain principal axes. DOI: http://dx.doi.org/10.7554/eLife.18449.006 10.7554/eLife.18449.007Figure 1—figure supplement 3. EC1 and EC3 use the same face for intersubunit contacts, as do EC2 and EC4. (A) Sequence alignment of clustered Pcdhs for which dimer interface structures are available. EC1 and EC4 of PcdhγA1 are grey because their interaction interface is unknown. Residues highlighted orange have a BSA > 10 Å2 in those respective structures. We selected as conserved interface residues (boxed) those that have a BSA > 10 Å2 in 5 of 6 structures for EC2/EC3 or 4 of 5 structures for EC1/EC4. The interface regions are notably similar in EC1 and EC3, as well as in EC2 and EC4. (B) Superposition of the first half of PcdhγB3 EC1-4 with the second half yields an RMSD of 2.24 Å over 142 Cα atoms. (C) The two superimposed chains at the top (B) are rotated 90°, highlighting how similar surfaces of EC1 and EC3, and EC2 and EC4, form the extended antiparallel interface of clustered Pcdhs. In (B) and (C), the conserved interface positions identified in (A) are marked by Cα atom spheres. DOI: http://dx.doi.org/10.7554/eLife.18449.007 10.7554/eLife.18449.008Figure 1—figure supplement 4. HEPES molecule near the EC2/EC3 interface. Final 2Fo-Fc electron density contoured at 1σ, with the final structural model shown as sticks. Nearby side chains from EC2 (lighter blue) and EC3 (darker blue) are labeled. DOI: http://dx.doi.org/10.7554/eLife.18449.008 Although the asymmetric unit contains a single PcdhγB3 molecule, a crystallographic two-fold axis generates an antiparallel dimer with intersubunit EC1/EC4 and EC2/EC3 interactions (Figure 1A). This dimer is consistent with the PcdhγA1 EC1-3 crystal structure, validating the previously predicted interface (Nicoludis et al., 2015; Rubinstein et al., 2015), and with recent α and β Pcdhs structures (Goodman et al., 2016), confirming that this interaction mechanism is conserved among all clustered Pcdh subfamilies (Figure 1D). The structures do differ noticeably in overall twist, including subfamily-specific differences in relative EC1/EC4 orientation (Figure 1E). The linear architecture of clustered Pcdhs enables extended antiparallel dimer interfaces. Overall, the tilt and azimuthal angles between adjacent clustered Pcdh repeats are distinct from those of classical cadherins (Figure 1—figure supplement 2) (Nicoludis et al., 2015). Classical cadherins, which typically dimerize through EC1/EC1 interfaces, exhibit smaller tilt angles and thus an overall curved structure (Boggon et al., 2002). Notably, the clustered Pcdh repeat orientation is such that EC1 and EC3 use the same face for intersubunit contacts, as do EC2 and EC4 (Figure 1—figure supplement 3), suggesting that longer cadherins could readily form even more extended interfaces. Clustered protocadherin subfamilies have distinct specificity mechanisms dictated by structural differences Clustered Pcdh subfamilies control different phenotypes in vivo and have discrete expression patterns (Biswas et al., 2012; Keeler et al., 2015), suggesting that they encode specificity using distinct modes, which may relate to subfamily-specific structural features. To investigate this hypothesis, we calculated the isoform conservation ratio (ICR) within individual subfamilies, which quantifies the extent to which individual residue positions are conserved among orthologs (same isoform in different species) and diversified in paralogs (different isoforms in the same species) (Nicoludis et al., 2015), resulting in three ICR value sets for the α, β and γ subfamilies, respectively (Figure 2—figure supplement 1). To account for subfamily differences in sequence conservation, we normalized the ICR values by dividing by the subfamily average. We then mapped them onto the Pcdhα7 EC1-5, Pcdhβ8 EC1-4 and PcdhγB3 EC1-4 structures (Figure 2A). Comparing the structures and isoform-specific conservation in the different subfamilies allowed us to identify key specificity determinant regions for individual subfamilies. We illustrate three examples of how the subfamilies have encoded specificity using unique structural features.10.7554/eLife.18449.009Figure 2. Isoform-specific conservation and structural differences reveal subfamily differences in diversity generation. (A) Subfamily-specific ICR values mapped onto the surfaces of Pcdhα7 (top, green), Pcdhβ8 (middle, magenta) and PcdhγB3 (bottom, blue). The black outline marks the dimer interface footprint. (B, C, D) Unique structural features of the α (left), β (center), and γ (right) structures (colored according to Figure 1). ICR values for highlighted residues shown below and normalized amino acid frequencies for these positions shown on the right. See Figure 2—figure supplement 1. DOI: http://dx.doi.org/10.7554/eLife.18449.009 10.7554/eLife.18449.010Figure 2—figure supplement 1. Clustered Pcdh subfamilies have distinct patterns of isoform-specific conservation. (A) Subfamily-specific normalized ICR values as a sliding average with a 5-residue window size (α = green, β = magenta, γ = cyan). Loop regions illustrated in Figure 2B–D are indicated at the bottom. (B) ICR values for interface residues (BSA > 10 Å2) for Pcdh α4 EC1-4 (yellow), α7 EC1-5 (green), β6 EC1-4 (salmon), β8 EC1-4 (magenta), γB3 EC1-4 (cyan), and γA1 EC1-3 (dark blue). In black are the average and standard error. Residues with high isoform-specific conservation localize to EC2 and EC3 surfaces. DOI: http://dx.doi.org/10.7554/eLife.18449.010 In α isoforms, the EC2 β4-β5 loop is enriched in high-ICR and chemically diverse residues, and differs in conformation in the Pcdhα4 and Pcdhα7 structures (Figure 2B): the Pcdhα4 EC2 β4-β5 loop contacts β1b of EC3, while the corresponding loop in Pcdhα7 does not, suggesting variable interactions in other isoforms. In comparison, the EC2 β4-β5 loop residues in both β and γ isoforms have lower ICR values, more similar loop structure, and do not contact β1b of EC3. Thus, this loop may have evolved to generate diversity within α isoforms, but not in other subfamilies. In β isoforms, the Phe-X10-Phe loop between β3 and β4 of EC3 has limited diversity compared to α and γ isoforms and wedges between the EC2 β4-β5 loop and β2b strand (Figure 2C). In contrast, the Phe-X10-Phe loop of α and γ isoforms has a helical conformation, and has residues with higher ICR values and greater chemical diversity. Therefore alterations in secondary structure can affect how specificity is encoded within the subfamilies. The short loop following the extended β1a strand in EC4 contacts the EC1 β6-β7 loop (Figure 2D), and there are large structural differences in the EC1/EC4 interaction between subfamilies (Figure 1E). α Isoforms have low-ICR residues at this interface, whereas β and γ isoforms have higher ICR value residues. This thus suggests that the large structural differences drive inter-subfamily specificity, on which may be layered additional isoform specificity. In all cases, sequence regions with high isoform-specific conservation correlate with interface contacts, revealing the interplay between dimer structure and how subfamilies encode specificity. Diversity in the composition and conformation of loop regions provides distinct specificity mechanisms to subfamilies. Phylogenetic analysis indicates that isoforms are more similar within than across subfamilies (Sotomayor et al., 2014; Wu, 2005), and the available structures show that the interface architecture is more similar within subfamilies as well (Figure 1E) (Goodman et al., 2016; Nicoludis et al., 2015). With this insight, the dimer interface seen in the PcdhγC3 EC1-3 crystal structure may represent a unique dimer architecture for C-type isoforms (Nicoludis et al., 2015), as these isoforms are transcriptionally, functionally and evolutionarily distinct from the subfamilies in which they reside (Chen et al., 2012; Frank et al., 2005; Kaneko et al., 2006). Distinct expression of the clustered Pcdh subfamilies in different tissues and at different developmental stages supports the necessity for intra-subfamily specificity (Biswas et al., 2012; Frank et al., 2005). Differences in subfamily structure and isoform-specific conservation suggest that homophilic specificity mechanisms emerged independently in each subfamily through diversification of subfamily-specific interface contacts. The EC1/EC4 interaction provides affinity of dimerization The EC2/EC3 interaction is integral to clustered Pcdh dimerization specificity, as evidenced by bioinformatics and cell-aggregation assays (Nicoludis et al., 2015; Rubinstein et al., 2015; Schreiner and Weiner, 2010; Thu et al., 2014; Wu, 2005). We sought to understand the functional purpose of the EC1/EC4 interaction, and made three observations. First, for all isoforms with available structures, fewer EC1/EC4 interface residues have high isoform-specific conservation compared to the EC2/EC3 interface residues (Figure 2A, Figure 2—figure supplement 1). Second, interface residues shared by most isoforms are more hydrophobic in EC1/EC4 than in EC2/EC3 (Figure 3A). Third, the PcdhγB3 EC1/EC4 interface is much larger (BSA = 976 Å2 per protomer) than the EC2/EC3 interface (555 Å2 per protomer). The lack of isoform specificity, the hydrophobic nature, and large interface area together suggest that the EC1/EC4 interface promotes binding with little specificity.10.7554/eLife.18449.011Figure 3. The EC1/EC4 interface is enriched in affinity-driving hydrophobic residues, while the EC2/EC3 interface contains high-ICR residues driving specificity. (A) Amino acid frequencies in clustered Pcdhs of conserved interface residues (see Figure 1—figure supplement 2). (B) Plot of ICR value and ∆∆Gcalc of interface residues of Pcdh α4 EC1-4 (yellow), α7 EC1-5 (green), β6 EC1-4 (salmon), β8 EC1-4 (magenta), γB3 EC1-4 (blue). Two subsets of interface residues segregate from the main cluster: high-∆∆Gcalc and low-ICR residues (‘affinity’; black box) and low-∆∆Gcalc and high-ICR residues (‘specificity’; crimson box). Residue F86 from PcdhγB3 EC1-4 is labeled. (C) High-∆∆Gcalc and low-ICR residues (black) map primarily to EC1 and EC4, while low-∆∆Gcalc and high-ICR residues (crimson) primarily map to EC2 and EC3. N253 (*) is found in the ‘specificity’ region for γB3 and in the ‘affinity’ region for β6 and β8. (D) The EC1/EC4 interface features a hydrophobic cluster, with EC1 F86 near its center. (E) SEC-MALS profiles of WT PcdhγB3 EC1-4 (blue; molecular weight 82 kDa) and F86A (black; molecular weight 52 kDa) run on a Superdex S200 10/300 column, are consistent with dimeric and monomeric proteins, respectively. DOI: http://dx.doi.org/10.7554/eLife.18449.011 To probe this hypothesis, we used the Computational Interface Alanine Scanning Server to assess each interface residue’s contribution to the free energy of complex formation (ΔΔGcalc) when computationally mutated to alanine (Kortemme and Baker, 2002; Kortemme et al., 2004). Using the five available EC1-4 interfaces, two residue groups emerged from comparing the ICR values to ΔΔGcalc: one group with low ICR values and high ΔΔGcalc, the other with high ICR values and low ΔΔGcalc (Figure 3B). These residue groups can be regarded as contributing to the affinity and specificity of the complex, respectively. When mapped on PcdhγB3, predicted affinity residues concentrated on EC1 and EC4, and predicted specificity residues on EC2 and EC3, corroborating the distinction between EC1/EC4 and EC2/EC3 interactions (Figure 3C). In the PcdhγB3 EC1/EC4 interface, F86, one of the predicted affinity-driving residues from EC1, wedges into a cavity created by hydrophobic EC4 residues (Figure 3D). A PcdhγB3 EC1-4 F86A mutant indeed disrupted dimerization, resulting in a monomeric protein as measured by MALS (Figure 3E). Thus, the hydrophobic interactions between EC1 and EC4 are crucial to dimerization. Analogously, purified EC1-3 constructs failed to dimerize in vitro whereas EC1-4 constructs did (Nicoludis et al., 2015; Rubinstein et al., 2015), and K562 cells expressing ΔEC1 or ΔEC4-6 constructs did not aggregate while cells expressing chimeras in which EC1 and EC4 derived from different paralogs did (Schreiner and Weiner, 2010; Thu et al., 2014). Together, these results indicate that the EC1/EC4 interaction is not strictly required for the specificity of dimerization but it drives dimerization affinity through non-specific hydrophobic interactions. Antiparallel EC1-4 interaction is predicted in non-clustered Pcdhs The antiparallel EC1-4 interaction architecture can encode diverse specificities within the clustered Pcdh family. Is this architecture unique to clustered Pcdhs or is it ancestral, and thus also found in non-clustered Pcdhs? These include the δ-1 (Pcdh1, Pcdh7, Pcdh9, Pcdh11) and δ-2 (Pcdh8, Pcdh10, Pcdh17, Pcdh18, Pcdh19) families that are integral to the development and maintenance of the nervous system (Keeler et al., 2015; Kim et al., 2011). We used sequence coevolution analysis, which successfully predicted the clustered Pcdh interface (Nicoludis et al., 2015) (Figure 4—figure supplement 1),to look for evidence of an antiparallel interface in non-clustered Pcdhs (Figure 4A). As in clustered Pcdhs, most covarying residue pairs in non-clustered Pcdhs were intra-domain structural contacts of the well-conserved cadherin fold. Additionally, several covarying pairs are found between EC2 and EC3, or EC1 and EC4, similar to those observed for the clustered Pcdhs. When mapped onto the PcdhγB3 dimer, these covarying pairs are somewhat further apart than true interface contacts (Figure 4B), which could be due to differences in dimerization interfaces, as we observe between the clustered Pcdh families, or in the δ-1 or δ-2 Pcdhs secondary structure, for which there are no available structures. This analysis thus predicts an antiparallel EC1-4 interaction in members of the non-clustered Pcdhs. Notably, we cannot determine whether all members or only a subset – and if so, which – likely use this architecture. However maximum parsimony suggests that the ancestral Pcdh used the antiparallel EC1-4 dimer interaction, and Pcdh members which do not show this interaction mechanism either diverged before it evolved or lost it subsequently.10.7554/eLife.18449.012Figure 4. Evolutionary couplings in non-clustered Pcdhs predict an antiparallel interface engaging EC1-EC4. (A) The top 38 covarying pairs are shown in black, and include a number of EC1-EC4 and EC2-EC3 covarying residue pairs. The intramolecular contact maps of PcdhγB3 EC1-4, Pcdhα4 EC1-4, Pcdhα7 EC1-4, Pcdhβ6 EC1-4, Pcdhβ8 EC1-4 and PcdhγA1 EC1-3 are in gray for reference. The observed interface contact residues are also mapped (α4, yellow; α7, green; β6, salmon; β8, magenta; γB3, blue; γA1, dark blue). (B) Covarying residue pairs across EC1-EC4 or EC2-EC3 are mapped onto the PcdhγB3 EC1-4 structure with a line between coupled residues. Alignments and evolutionary couplings in Figure 4—source data 1 and 2. (C) Amino acid frequencies at non-clustered Pcdh alignment positions corresponding to the conserved interface residue positions identified in clustered Pcdhs (Figure 1—figure supplement 2). See Figure 4—figure supplements 1 and 2. DOI: http://dx.doi.org/10.7554/eLife.18449.012 10.7554/eLife.18449.013Figure 4—source data 1. Alignment of non-clustered Pcdhs EC1-4. DOI: http://dx.doi.org/10.7554/eLife.18449.013 10.7554/eLife.18449.014Figure 4—source data 2. Evolutionary couplings from the non-clustered Pcdh alignment. DOI: http://dx.doi.org/10.7554/eLife.18449.014 10.7554/eLife.18449.015Figure 4—source data 3. Alignment of clustered Pcdhs EC1-4. DOI: http://dx.doi.org/10.7554/eLife.18449.015 10.7554/eLife.18449.016Figure 4—source data 4. Evolutionary couplings from the clustered Pcdh alignment. DOI: http://dx.doi.org/10.7554/eLife.18449.016 10.7554/eLife.18449.017Figure 4—figure supplement 1. Evolutionary couplings in clustered Pcdhs are consistent with all available EC1-EC4 antiparallel homodimeric interfaces. The top 83 covarying pairs are shown in black. The intramolecular contact maps of PcdhγB3 EC1-4, Pcdhα4 EC1-4, Pcdhα7 EC1-4, Pcdhβ6 EC1-4, Pcdhβ8 EC1-4 and PcdhγA1 EC1-3 are in gray for reference. The observed interface contact residues are also mapped (α4, yellow; α7, green; β6, salmon; β8, magenta; γB3, blue; γA1, dark blue). Alignments and evolutionary couplings in Figure 4—source data 3 and 4. DOI: http://dx.doi.org/10.7554/eLife.18449.017 10.7554/eLife.18449.018Figure 4—figure supplement 2. Phylogenetic tree distinguishes clustered from non-clustered Pcdhs. Based on this phylogeny, evolutionary couplings were obtained for the two groups labeled. The clustered and non-clustered Pcdh alignments had effective sequences numbers of 2660 and 405.5, respectively. DOI: http://dx.doi.org/10.7554/eLife.18449.018 Finally, we looked at the composition of a predicted non-clustered Pcdh interface, by selecting residues homologous to those found at clustered Pcdh interfaces. The predicted EC1/EC4 interface residues are predominantly hydrophobic, while EC2/EC3 residues have more polar and ionic character (Figure 4C). Notably, positions 41 and 77 in EC1 and 320, 321, 371 and 373 in EC4 are more hydrophobic in non-clustered than clustered Pcdhs, indicating that these may form contacts in some non-clustered Pcdhs. Thus, like in the clustered Pcdhs, the EC1/EC4 interaction may promote dimer affinity while the EC2/EC3 interaction provides specificity. Conclusions Recently, we and others predicted that clustered Pcdhs form homophilic antiparallel EC1-EC4 complexes based on crystal structures, mutagenesis and bioinformatics (Nicoludis et al., 2015; Rubinstein et al., 2015). Our structure of PcdhγB3 EC1-4 confirms our sequence coevolution analysis, demonstrating the robustness of the analysis and revealing the molecular details of this interaction. Here we extended this prediction to other non-clustered Pcdhs using sequence coevolution analysis. Analysis of the PcdhγB3 EC1-4 structure in comparison to α or β subfamily dimers revealed structural differences that correlated with differences in isoform-specific conservation, indicating distinct specificity mechanisms. Unlike the Dscams, where isoforms vary at specific alternatively-spliced regions (Li et al., 2016; Meijers et al., 2007; Sawaya et al., 2008; Wojtowicz et al., 2007), the clustered Pcdh subfamilies are structurally diverse, and thus can encode specificity in different ways. We identified a hydrophobic interaction between EC1 and EC4 that contributes to dimerization affinity, whereas its conservation among clustered Pcdh isoforms suggests that this interaction is not a driver of specificity. Overall, our data support a general role for conserved hydrophobic EC1/EC4 interactions in affinity, and for highly diversified polar EC2/EC3 contacts in specificity, and sequence analyses suggest that this is conserved in at least some non-clustered Pcdhs. Materials and methods Expression, purification and crystallization of PcdhγB3 EC1-4 Human PcdhγB3 EC1-4 (residues 1–414, not counting the signal peptide) was cloned into pET21 with a C-terminal hexahistidine tag, expressed in BL21 Gold (DE3) Escherichia coli cells in terrific broth. Cells were induced at OD600 = 0.8 with 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) at 37°C for 4 hr, harvested and lysed by sonication in 8 M guanadinium hydrochloride (GuHCl), 50 mM HEPES pH 7.5, 2 mM CaCl2, and 20 mM imidazole. Cell lysates were diluted to 5 M GuHCl and loaded onto Ni-Sepharose, washed with 50 mM HEPES pH 7.5, 250 mM NaCl, 10 mM CaCl2, and 25 mM imidazole and eluted with 250 mM imidazole. Eluted protein was refolded at 1 mg/mL in 12 hr dialysis steps reducing the GuHCl concentration from 2.5 M to 1.25 M and finally 0 M in refolding buffer (100 mM Tris pH 8.5, 10 mM CaCl2, 1 mM EDTA, 5 mM dithiothreitol (DTT), and 0.5 M L-arginine). Concentrated refolded protein was purified by size-exclusion chromatography (SEC) on a Superdex 200 16/60 column (GE Healthcare, Pittsburgh, PA) in 20 mM Tris pH 8.5, 200 mM NaCl, and 2 mM CaCl2 (SEC buffer). Two peaks were isolated and each peak was run again separately by SEC before being concentrated for crystallization. Multi-angle light scattering (MALS) Approximate molecular mass of PcdhγB3 EC1-4 protein (WT or F86A mutant) was determined using a Superdex S200 10/300 column (GE Healthcare, Pittsburgh, PA) with in-line Wyatt Dawn Heleos II and Optilab T-rex refractive index detectors. Protein (100 μL at 4 mg/mL) was injected and run at 0.4 mL/min in SEC buffer. Signals were aligned, normalized and band-broadened using bovine serum albumin as a standard. Crystallization, data collection, and structure determination and analysis Crystals were obtained by vapor diffusion at room temperature in 0.1 M HEPES pH 7, 4% ethylene glycol, and 5% polyethylene glycol monomethyl ether 500 in a 0.3 μL protein (12 mg/mL) to 0.3 μL reservoir drop, then cryoprotected with reservoir with 20% glycerol before flash cooling in liquid N2. Diffraction data (Figure 1—source data 1) were processed in HKL2000 (Otwinowski and Minor, 1997). The PcdhγB3 EC1-4 structure was determined by an iterative molecular replacement search with single domains of the PcdhγA1 EC1-3 structure (PDBID 4zi9) in PHENIX (Adams et al., 2010). Model building was done in Coot (Emsley and Cowtan, 2004) and refinement in PHENIX (Adams et al., 2010). We analyzed the physicochemical properties of the dimer interface using PISA (Krissinel and Henrick, 2007). In the structure, we found a HEPES molecule near the EC2/EC3 interface that forms a salt bridge with N253 and N155 (Figure 1—figure supplement 4). MALS data were collected with Tris as the buffer (Figure 1—figure supplement 1), indicating that HEPES is not required for dimerization. ICR value calculations Overall percent identity of the most common residue at each position was used to calculate ICR values, dividing the percent identity across subfamily orthologs by the percent identity across subfamily paralogs. ICR values were then normalized by dividing by the whole sequence ICR average within each subfamily. The alignment and identity data are provided here (Nicoludis et al., 2015). Computational Interface Alanine Scanning Server Pcdhα4 EC1-4 (5dzw), Pcdhα7 EC1-5 (5dzv), Pcdhβ6 EC1-4 (5dzx), Pcdhβ8 EC1-4 (5dzy), and PcdhγB3 EC1-4 (5k8r) dimer structures were submitted to the Computational Interface Alanine Scanning Server using default settings (Kortemme and Baker, 2002; Kortemme et al., 2004). Covariation analyses Previously, we generated an alignment of clustered Pcdhs using mouse PcdhγC3 and manually filtered by phylogeny, using FastTree 2.1 (Price et al., 2010), to eliminate non-clustered Pcdhs (Figure 4—figure supplement 2) (Nicoludis et al., 2015). Both this clustered Pcdh and the now separated non-clustered Pcdh alignment were filtered to remove sequences with more than 50% gaps and columns with more than 30% gaps, and trimmed to contain only EC1-EC4. The clustered and non-clustered Pcdh alignments had 2660 and 405.5 effective non-redundant sequences, respectively; sequences were considered redundant and downweighted when more than 90% identical over their full length (Hopf et al., 2014). Evolutionary couplings (Hopf et al., 2014; Marks et al., 2011) were computed using the updated ‘PLMC’ algorithm (Weinreb et al., 2015) available on https://github.com/debbiemarkslab/plmc, which uses a pseudo maximum likelihood approximation (Balakrishnan et al., 2011; Ekeberg et al., 2013; Kamisetty et al., 2013). Alignments and all EC scores are provided (Figure 4—source data 1–4). We used the precision of the intra-domain evolutionary couplings to determine whether the inter-domain evolutionary couplings are likely to be true. For the clustered Pcdh alignment, 83 non-local (more than five residues apart) contacts fall above a threshold of 80% precision of the intra-domain evolutionary couplings. Intra-domain evolutionary couplings are considered true if they correspond to structural contact (minimum atom distance < 8 Å) in any structure (Figure 4—figure supplement 1). Based on the same 80% precision threshold, the top 38 non-local ECs are significant in the non-clustered Pcdh alignment. We exclude couplings between residues greater than 400 in this analysis due to the false signal from gaps in this region. Funding Information This paper was supported by the following grants: http://dx.doi.org/10.13039/100000001National Science Foundation DGE1144152 to Anna G Green. http://dx.doi.org/10.13039/100000002National Institutes of Health NCRR 1S10RR028832-01 to Debora S Marks. http://dx.doi.org/10.13039/100000057National Institute of General Medical Sciences GM106303 to Debora S Marks. National Defense Science and Engineering Graduate Fellowship to John M Nicoludis. http://dx.doi.org/10.13039/100000057National Institute of General Medical Sciences P41 GM103403 to Rachelle Gaudet. http://dx.doi.org/10.13039/100000002National Institutes of Health S10 RR029205 to Rachelle Gaudet. Acknowledgements We thank Kelly Arnett, director of the Center for Macromolecular Interactions, Harvard Medical School) for help in collecting SEC-MALS data. We thank the beamline staff of NE-CAT at the Advanced Photon Source (Argonne, IL, USA) for help with data collection. NE-CAT is funded by NIH (P41 GM103403 and S10 RR029205) and the Advanced Photon Source by the US Department of Energy (DE-AC02-06CH11357). Evolutionary couplings analysis was conducted on the Orchestra High Performance Compute Cluster at Harvard Medical School, which is funded by the NIH (NCRR 1S10RR028832-01). Financial support (to JMN) was provided by the National Defense Science and Engineering Graduate Fellowship. AGG was supported by the National Science Foundation Graduate Research Fellowship (DGE1144152). DSM was supported by National Institutes of Health (GM106303). Additional information Competing interests The authors declare that no competing interests exist. Author contributions JMN, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. BEV, Conception and design, Acquisition of data, Drafting or revising the article. AGG, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. CPIS, Drafting or revising the article, Contributed unpublished essential data or reagents. DSM, Conception and design, Analysis and interpretation of data, Drafting or revising the article. RG, Conception and design, Analysis and interpretation of data, Drafting or revising the article. Additional files Major datasets The following datasets were generated: Nicoludis JM ,Vogt BE ,Gaudet R ,2016 ,Structure of human clustered protocadherin gamma B3 EC1-4 ,http://www.rcsb.org/pdb/explore.do?structureId=5K8R,Publicly available at the RCSB Protein Data Bank (accession no: 5K8R) Nicoludis JM ,Vogt BE ,Gaudet R ,2016 ,Structure of human clustered protocadherin gamma B3 EC1-4 ,https://data.sbgrid.org/dataset/325,Publicly available at Structural Biology Data Grid (accession no: 325) 10.7554/eLife.18449.023Decision letter Weis William I Reviewing editorStanford University, United StatesIn the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included. Thank you for submitting your article "Antiparallel protocadherin homodimers use distinct affinity- and specificity-mediating regions in cadherin repeats 1-4" for consideration by eLife. Your article has been reviewed by two peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Arup Chakraborty as the Senior Editor. The reviewers have opted to remain anonymous. The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission. The reviewers and editors are in agreement that the work is worthy of publication and only have a few minor concerns that would need to be addressed before acceptance. Summary: The authors report the structure of the first four extracellular cadherin repeats of PcdhgB3, the first of the γ subfamily of clustered protocadherins. The structure reveals an antiparallel dimer in which the four domains mediate dimerization. Using the previously known α- and β-subfamily dimer structures, the authors provide a detailed structural analysis of dimerization specificity throughout the clustered protocadherins: it appears that the EC2-EC3 interaction contributes to the specificity of binding within each subfamily, while the EC1-EC4 interface provides for a more general nonselective binding affinity. They also predict that the non-clustered protocadherins likely work through similar mechanisms. This provides insights into the binding mechanisms of the entire clustered protocadherin family that supports and extends previous functional studies. Overall it represents a significant advance in understanding this large family of adhesion proteins. Minor points: 1) First paragraph, Results and Discussion and Figure 1C: "Topology" refers to the connectivity of secondary structure elements that interact with one another. The topology of EC4 is not different from other EC domains – it simply has a split β strand but it hydrogen bonds interacts with the same neighboring strands. It is fine to point out this difference but is not a topological difference. 2) Materials and methods section, subsection “Covariation analyses”: "…405.5 non-redundant sequences…" – what is a 0.5 sequence? 3) Figure 1—source data 1 (Crystallographic statistics):a) The number of significant figures is inappropriate, please fix (true of almost every entry, e.g. unit cell lengths, R values, B factors); b) "Ligands" – please specify, this presumably means Ca2+, HEPES, anything else? 10.7554/eLife.18449.024Author response Summary: The authors report the structure of the first four extracellular cadherin repeats of PcdhgB3, the first of the γ subfamily of clustered protocadherins. The structure reveals an antiparallel dimer in which the four domains mediate dimerization. Using the previously known α- and β-subfamily dimer structures, the authors provide a detailed structural analysis of dimerization specificity throughout the clustered protocadherins: it appears that the EC2-EC3 interaction contributes to the specificity of binding within each subfamily, while the EC1-EC4 interface provides for a more general nonselective binding affinity. They also predict that the non-clustered protocadherins likely work through similar mechanisms. This provides insights into the binding mechanisms of the entire clustered protocadherin family that supports and extends previous functional studies. Overall it represents a significant advance in understanding this large family of adhesion proteins. Minor points: 1) First paragraph, Results and Discussion and Figure 1C: "Topology" refers to the connectivity of secondary structure elements that interact with one another. The topology of EC4 is not different from other EC domains – it simply has a split β strand but it hydrogen bonds interacts with the same neighboring strands. It is fine to point out this difference but is not a topological difference. Thank you for pointing this out. We have now rephrased our descriptions: Within the Results and Discussion section, we now state (new wording italicised): “Notably, EC4 has a unique β-strand arrangement compared to EC1-EC3 (Figure 1B,C) […] This distinct structural feature contributes to intersubunit EC1/EC4 interactions (see below).” In the legend to Figure 1, two instances of the word “topology” were replaced with “β-strand arrangement”. 2) Materials and methods section, subsection “Covariation analyses”: "…405.5 non-redundant sequences…" – what is a 0.5 sequence? The numbers listed actually correspond to the effective number of sequences in the alignment. To calculate the effective number of sequences, each sequence is weighted by dividing by one plus its number of neighbors, which we define as sequences that are more than 90% identical. For example, a sequence that has one neighbor in the alignment counts as 0.5 sequence. Because each sequence will have a different number of neighbors, this can result in non-integer totals when the weighted numbers of sequences are summed. We have now clarified this in the methods, in the following reworded sentence: “The clustered and non-clustered Pcdh alignments had 2660 and 405.5 effective non-redundant sequences, respectively; sequences were considered redundant and downweighted when more than 90% identical over their full length (Hopf et al., 2014).” 3) Figure 1—source data 1 (Crystallographic statistics):a) The number of significant figures is inappropriate, please fix (true of almost every entry, e.g. unit cell lengths, R values, B factors); We have edited the table to reduce the number of significant digits so as to be more in line with typical statistics provided in publications. b) "Ligands" – please specify, this presumably means Ca2+, HEPES, anything else? We have added the list of ligands in the first column, and the corresponding number of non-hydrogen atoms in the second column, as follows: Ligands (Ca2+, Cl-, Na+, HEPES, ethylene glycol) 31 (9, 2, 1, 15, 4) ==== Refs References Adams PD Afonine PV Bunkóczi G Chen VB Davis IW Echols N Headd JJ Hung LW Kapral GJ Grosse-Kunstleve RW McCoy AJ Moriarty NW Oeffner R Read RJ Richardson DC Richardson JS Terwilliger TC Zwart PH 2010 PHENIX: a comprehensive Python-based system for macromolecular structure solution Acta Crystallographica Section D Biological Crystallography 66 213 221 10.1107/S0907444909052925 20124702 Balakrishnan S Kamisetty H Carbonell JG Lee SI Langmead CJ 2011 Learning generative models for protein fold families Proteins 79 1061 1078 10.1002/prot.22934 21268112 Biswas S Emond MR Jontes JD 2012 The clustered protocadherins Pcdhα and Pcdhγ form a heteromeric complex in zebrafish Neuroscience 219 280 289 10.1016/j.neuroscience.2012.05.058 22659564 Boggon TJ Murray J Chappuis-Flament S Wong E Gumbiner BM Shapiro L 2002 C-cadherin ectodomain structure and implications for cell adhesion mechanisms Science 296 1308 1313 10.1126/science.1071559 11964443 Chen WV Alvarez FJ Lefebvre JL Friedman B Nwakeze C Geiman E Smith C Thu CA Tapia JC Tasic B Sanes JR Maniatis T 2012 Functional significance of isoform diversification in the protocadherin gamma gene cluster Neuron 75 402 409 10.1016/j.neuron.2012.06.039 22884324 Ekeberg M Lövkvist C Lan Y Weigt M Aurell E 2013 Improved contact prediction in proteins: Using pseudolikelihoods to infer Potts models Physical Review E 87 1 16 10.1103/PhysRevE.87.012707 Emond MR Jontes JD 2008 Inhibition of protocadherin-alpha function results in neuronal death in the developing zebrafish Developmental Biology 321 175 187 10.1016/j.ydbio.2008.06.011 18602383 Emsley P Cowtan K 2004 Coot: model-building tools for molecular graphics Acta Crystallographica Section D Biological Crystallography 60 2126 2132 10.1107/S0907444904019158 15572765 Frank M Ebert M Shan W Phillips GR Arndt K Colman DR Kemler R 2005 Differential expression of individual gamma-protocadherins during mouse brain development Molecular and Cellular Neuroscience 29 603 616 10.1016/j.mcn.2005.05.001 15964765 Garrett AM Schreiner D Lobas MA Weiner JA 2012 γ-protocadherins control cortical dendrite arborization by regulating the activity of a FAK/PKC/MARCKS signaling pathway Neuron 74 269 276 10.1016/j.neuron.2012.01.028 22542181 Goodman KM Rubinstein R Thu CA Bahna F Mannepalli S Ahlsén G Rittenhouse C Maniatis T Honig B Shapiro L 2016 Structural basis of diverse homophilic recognition by clustered α- and β-protocadherins Neuron 90 709 723 10.1016/j.neuron.2016.04.004 27161523 Hasegawa S Hamada S Kumode Y Esumi S Katori S Fukuda E Uchiyama Y Hirabayashi T Mombaerts P Yagi T 2008 The protocadherin-alpha family is involved in axonal coalescence of olfactory sensory neurons into glomeruli of the olfactory bulb in mouse Molecular and Cellular Neuroscience 38 66 79 10.1016/j.mcn.2008.01.016 18353676 Hasegawa S Hirabayashi T Kondo T Inoue K Esumi S Okayama A Hamada S Yagi T 2012 Constitutively expressed Protocadherin-α regulates the coalescence and elimination of homotypic olfactory axons through its cytoplasmic region Frontiers in Molecular Neuroscience 5 e18449 10.3389/fnmol.2012.00097 Hayashi S Takeichi M 2015 Emerging roles of protocadherins: from self-avoidance to enhancement of motility Journal of Cell Science 128 1 10 10.1242/jcs.166306 25556248 Hirano S Yan Q Suzuki ST 1999 Expression of a novel protocadherin, OL-protocadherin, in a subset of functional systems of the developing mouse brain Journal of Neuroscience 19 995 1005 9920663 Hopf TA Schärfe CP Rodrigues JP Green AG Kohlbacher O Sander C Bonvin AM Marks DS 2014 Sequence co-evolution gives 3D contacts and structures of protein complexes eLife 3 e18449 10.7554/eLife.03430 Hoshina N Tanimura A Yamasaki M Inoue T Fukabori R Kuroda T Yokoyama K Tezuka T Sagara H Hirano S Kiyonari H Takada M Kobayashi K Watanabe M Kano M Nakazawa T Yamamoto T 2013 Protocadherin 17 regulates presynaptic assembly in topographic corticobasal Ganglia circuits Neuron 78 839 854 10.1016/j.neuron.2013.03.031 23684785 Hulpiau P Gul SI van Roy F 2013 New insights into the evolution of metazoan cadherins and catenins Roy Van F The Molecular Biology of Cadherins San Diego, CA Elsevier 71 94 10.1016/b978-0-12-394311-8.00004-2 Hulpiau P van Roy F 2011 New insights into the evolution of metazoan cadherins Molecular Biology and Evolution 28 647 657 10.1093/molbev/msq233 20817718 Kamisetty H Ovchinnikov S Baker D 2013 Assessing the utility of coevolution-based residue-residue contact predictions in a sequence- and structure-rich era PNAS 110 15674 15679 10.1073/pnas.1314045110 24009338 Kaneko R Kato H Kawamura Y Esumi S Hirayama T Hirabayashi T Yagi T 2006 Allelic gene regulation of Pcdh-alpha and Pcdh-gamma clusters involving both monoallelic and biallelic expression in single Purkinje cells Journal of Biological Chemistry 281 30551 30560 10.1074/jbc.M605677200 16893882 Katori S Hamada S Noguchi Y Fukuda E Yamamoto T Yamamoto H Hasegawa S Yagi T 2009 Protocadherin-alpha family is required for serotonergic projections to appropriately innervate target brain areas Journal of Neuroscience 29 9137 9147 10.1523/JNEUROSCI.5478-08.2009 19625505 Keeler AB Molumby MJ Weiner JA 2015 Protocadherins branch out: Multiple roles in dendrite development Cell Adhesion & Migration 9 214 226 10.1080/19336918.2014.1000069 25869446 Kim SY Yasuda S Tanaka H Yamagata K Kim H 2011 Non-clustered protocadherin Cell Adhesion & Migration 5 97 105 10.4161/cam.5.2.14374 21173574 Kim SH Yamamoto A Bouwmeester T Agius E Robertis EM 1998 The role of paraxial protocadherin in selective adhesion and cell movements of the mesoderm during Xenopus gastrulation Development 125 4681 4690 9806917 Kortemme T Baker D 2002 A simple physical model for binding energy hot spots in protein-protein complexes PNAS 99 14116 14121 10.1073/pnas.202485799 12381794 Kortemme T Kim DE Baker D 2004 Computational alanine scanning of protein-protein interfaces Science Signaling 2004 pl2 10.1126/stke.2192004pl2 Kostadinov D Sanes JR 2015 Protocadherin-dependent dendritic self-avoidance regulates neural connectivity and circuit function eLife 4 e18449 10.7554/eLife.08964 Krissinel E Henrick K 2007 Inference of macromolecular assemblies from crystalline state Journal of Molecular Biology 372 774 797 10.1016/j.jmb.2007.05.022 17681537 Kuroda H Inui M Sugimoto K Hayata T Asashima M 2002 Axial protocadherin is a mediator of prenotochord cell sorting in Xenopus Developmental Biology 244 267 277 10.1006/dbio.2002.0589 11944936 Ledderose J Dieter S Schwarz MK 2013 Maturation of postnatally generated olfactory bulb granule cells depends on functional γ-protocadherin expression Scientific Reports 3 e18449 10.1038/srep01514 Lefebvre JL Kostadinov D Chen WV Maniatis T Sanes JR 2012 Protocadherins mediate dendritic self-avoidance in the mammalian nervous system Nature 488 517 521 10.1038/nature11305 22842903 Li S-A Cheng L Yu Y Chen Q Li S 2016 Structural basis of Dscam1 homodimerization: Insights into context constraint for protein recognition Science Advances 2 e1501118 1501119 10.1126/sciadv.1501118 27386517 Marks DS Colwell LJ Sheridan R Hopf TA Pagnani A Zecchina R Sander C 2011 Protein 3D structure computed from evolutionary sequence variation PLoS One 6 e18449 10.1371/journal.pone.0028766 21479177 Meijers R Puettmann-Holgado R Skiniotis G Liu JH Walz T Wang JH Schmucker D 2007 Structural basis of Dscam isoform specificity Nature 449 487 491 10.1038/nature06147 17721508 Molumby MJ Keeler AB Weiner JA 2016 Homophilic protocadherin cell-cell interactions promote dendrite complexity Cell Reports 15 1037 1050 10.1016/j.celrep.2016.03.093 27117416 Nicoludis JM Lau SY Schärfe CP Marks DS Weihofen WA Gaudet R 2015 Structure and sequence analyses of clustered protocadherins reveal antiparallel interactions that mediate homophilic specificity Structure 23 2087 2098 10.1016/j.str.2015.09.005 26481813 Otwinowski Z Minor W 1997 Processing of x-ray diffraction data collected in oscillation mode Carter Jr CW Sweet RM Methods in Enzymology, Volume 276: Macromolecular Crystallography, Part A New York Academic Press 307 326 10.1016/s0076-6879(97)76066-x Price MN Dehal PS Arkin AP 2010 FastTree 2--approximately maximum-likelihood trees for large alignments PLoS One 5 e18449 10.1371/journal.pone.0009490 Rashid D Newell K Shama L Bradley R 2006 A requirement for NF-protocadherin and TAF1/Set in cell adhesion and neural tube formation Developmental Biology 291 170 181 10.1016/j.ydbio.2005.12.027 16426602 Rubinstein R Thu CA Goodman KM Wolcott HN Bahna F Mannepalli S Ahlsen G Chevee M Halim A Clausen H Maniatis T Shapiro L Honig B 2015 Molecular logic of neuronal self-recognition through protocadherin domain interactions Cell 163 629 642 10.1016/j.cell.2015.09.026 26478182 Sawaya MR Wojtowicz WM Andre I Qian B Wu W Baker D Eisenberg D Zipursky SL 2008 A double S shape provides the structural basis for the extraordinary binding specificity of Dscam isoforms Cell 134 1007 1018 10.1016/j.cell.2008.07.042 18805093 Schreiner D Weiner JA 2010 Combinatorial homophilic interaction between gamma-protocadherin multimers greatly expands the molecular diversity of cell adhesion PNAS 107 14893 14898 10.1073/pnas.1004526107 20679223 Sotomayor M Gaudet R Corey DP 2014 Sorting out a promiscuous superfamily: towards cadherin connectomics Trends in Cell Biology 24 524 536 10.1016/j.tcb.2014.03.007 24794279 Suo L Lu H Ying G Capecchi MR Wu Q 2012 Protocadherin clusters and cell adhesion kinase regulate dendrite complexity through Rho GTPase Journal of Molecular Cell Biology 4 362 376 10.1093/jmcb/mjs034 22730554 Thu CA Chen WV Rubinstein R Chevee M Wolcott HN Felsovalyi KO Tapia JC Shapiro L Honig B Maniatis T 2014 Single-cell identity generated by combinatorial homophilic interactions between α, β, and γ protocadherins Cell 158 1045 1059 10.1016/j.cell.2014.07.012 25171406 Wang X Weiner JA Levi S Craig AM Bradley A Sanes JR a. L Craig S 2002 Gamma protocadherins are required for survival of spinal interneurons Neuron 36 843 854 10.1016/S0896-6273(02)01090-5 12467588 Weiner JA Wang X Tapia JC Sanes JR 2005 Gamma protocadherins are required for synaptic development in the spinal cord PNAS 102 8 14 10.1073/pnas.0407931101 15574493 Weinreb C Riesselman AJ Ingraham JB Gross T Sander C Marks DS 2016 3D RNA and functional interactions from evolutionary couplings Cell 165 963 975 10.1016/j.cell.2016.03.030 27087444 Wojtowicz WM Wu W Andre I Qian B Baker D Zipursky SL 2007 A vast repertoire of Dscam binding specificities arises from modular interactions of variable Ig domains Cell 130 1134 1145 10.1016/j.cell.2007.08.026 17889655 Wu Q 2005 Comparative genomics and diversifying selection of the clustered vertebrate protocadherin genes Genetics 169 2179 2188 10.1534/genetics.104.037606 15744052 Yamagata K Andreasson KI Sugiura H Maru E Dominique M Irie Y Miki N Hayashi Y Yoshioka M Kaneko K Kato H Worley PF 1999 Arcadlin is a neural activity-regulated cadherin involved in long term potentiation Journal of Biological Chemistry 274 19473 19479 10.1074/jbc.274.27.19473 10383464 Yoshida K 2003 Fibroblast cell shape and adhesion in vitro is altered by overexpression of the 7a and 7b isoforms of protocadherin 7, but not the 7c isoform Cellular & Molecular Biology Letters 8 735 741 12949613 Zipursky SL Sanes JR 2010 Chemoaffinity revisited: dscams, protocadherins, and neural circuit assembly Cell 143 343 353 10.1016/j.cell.2010.10.009 21029858

PMC005xxxxxx/PMC5001839.txt

==== Front eLifeElifeeLifeeLifeeLife2050-084XeLife Sciences Publications, Ltd 0954110.7554/eLife.09541Research ArticleBiophysics and Structural BiologyComputational and Systems BiologyOverall energy conversion efficiency of a photosynthetic vesicle Sener Melih 12*Strumpfer Johan 13Singharoy Abhishek 1Hunter C Neil 4Schulten Klaus http://orcid.org/0000-0001-7192-9632123*1 Beckman Institute for Advanced Science and Technology, University of Illinois at Urbana-Champaign, Urbana, United States2 Department of Physics, University of Illinois at Urbana-Champaign, Urbana, United States3 Center for Biophysics and Computational Biology, University of Illinois at Urbana-Champaign, Urbana, United States4 Department of Molecular Biology and Biotechnology, University of Sheffield, Sheffield, United KingdomHummer Gerhard Reviewing editorThe Max Planck Institute of Biophysics, Germanymelih@ks.uiuc.edu (MS);kschulte@ks.uiuc.edu (KS)26 8 2016 2016 5 e0954119 6 2015 11 7 2016 © 2016, Sener et al2016Sener et alThis article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.The chromatophore of purple bacteria is an intracellular spherical vesicle that exists in numerous copies in the cell and that efficiently converts sunlight into ATP synthesis, operating typically under low light conditions. Building on an atomic-level structural model of a low-light-adapted chromatophore vesicle from Rhodobacter sphaeroides, we investigate the cooperation between more than a hundred protein complexes in the vesicle. The steady-state ATP production rate as a function of incident light intensity is determined after identifying quinol turnover at the cytochrome bc1 complex (cytb⁢c1) as rate limiting and assuming that the quinone/quinol pool of about 900 molecules acts in a quasi-stationary state. For an illumination condition equivalent to 1% of full sunlight, the vesicle exhibits an ATP production rate of 82 ATP molecules/s. The energy conversion efficiency of ATP synthesis at illuminations corresponding to 1%–5% of full sunlight is calculated to be 0.12–0.04, respectively. The vesicle stoichiometry, evolutionarily adapted to the low light intensities in the habitat of purple bacteria, is suboptimal for steady-state ATP turnover for the benefit of protection against over-illumination. DOI: http://dx.doi.org/10.7554/eLife.09541.001 eLife digest Photosynthesis, or the conversion of light energy into chemical energy, is a process that powers almost all life on Earth. Plants and certain bacteria share similar processes to perform photosynthesis, though the purple bacterium Rhodobacter sphaeroides uses a photosynthetic system that is much less complex than that in plants. Light harvesting inside the bacterium takes place in up to hundreds of compartments called chromatophores. Each chromatophore in turn contains hundreds of cooperating proteins that together absorb the energy of sunlight and convert and store it in molecules of ATP, the universal energy currency of all cells. The chromatophore of primitive purple bacteria provides a model for more complex photosynthetic systems in plants. Though researchers had characterized its individual components over the years, less was known about the overall architecture of the chromatophore and how its many components work together to harvest light energy efficiently and robustly. This knowledge would provide insight into the evolutionary pressures that shaped the chromatophore and its ability to work efficiently at different light intensities. Sener et al. now present a highly detailed structural model of the chromatophore of purple bacteria based on the findings of earlier studies. The model features the position of every atom of the constituent proteins and is used to examine how energy is transferred and converted. Sener et al. describe the sequence of energy conversion steps and calculate the overall energy conversion efficiency, namely how much of the light energy arriving at the microorganism is stored as ATP. These calculations show that the chromatophore is optimized to produce chemical energy at low light levels typical of purple bacterial habitats, and dissipate excess energy to avoid being damaged under brighter light. The chromatophore’s architecture also displays robustness against perturbations of its components. In the future, the approach used by Sener et al. to describe light harvesting in this bacterial compartment can be applied to more complex systems, such as those in plants. DOI: http://dx.doi.org/10.7554/eLife.09541.002 Author Keywords bacterial photosynthesisexcitation transferATP productionenergy conversion efficiencyRhodobacter sphaeroidesResearch Organism Otherhttp://dx.doi.org/10.13039/501100000268Biotechnology and Biological Sciences Research CouncilBB/M000265/1Hunter C Neil http://dx.doi.org/10.13039/100000015U.S. Department of EnergyDE-SC0001035Hunter C Neil Schulten Klaus http://dx.doi.org/10.13039/501100000781European Research Council338895Hunter C Neil http://dx.doi.org/10.13039/100000001National Science FoundationPHY0822613Schulten Klaus http://dx.doi.org/10.13039/100000002National Institutes of HealthNIH 9P41GM104601Schulten Klaus http://dx.doi.org/10.13039/100000001National Science FoundationMCB1157615Schulten Klaus The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.elife-xml-version2.5Author impact statementThe overall energy conversion efficiency is calculated for a bacterial vesicle that harvests solar energy for ATP production on the basis of an atomic-detail structural model. ==== Body Introduction Energy for most life on Earth is provided by sunlight harvested by photosynthetic organisms, which have evolved a wide variety of mechanisms for utilizing light energy to drive cellular processes (Blankenship, 2014). These organisms absorb sunlight and subsequently utilize the Förster mechanism (Sener et al., 2011) and quantum coherence (Strümpfer et al., 2012; Panitchayangkoon et al., 2010; Scholes, 2010) for efficient excitation energy transfer, followed by conversion of light energy into chemical energy (Feniouk and Junge, 2009). The light harvesting system of purple bacteria (Hu et al., 2002; Cartron et al., 2014) is claimed to be the earliest of the current photosynthetic lineages (Xiong et al., 2000) and exhibits, at the supra-molecular level as well as at the level of individual proteins, less complexity than the thylakoid membranes of the more ubiquitous cyanobacteria and plants (Kirchhoff et al., 2002). In the purple bacterium Rhodobacter (Rba.) sphaeroides the basic photosynthetic unit is the chromatophore (Cogdell et al., 2006; Strümpfer et al., 2011; Cartron et al., 2014), a 50–70 nm diameter vesicle, as shown in Figure 1, formed through invagination of the intracytoplasmic membrane (Tucker et al., 2010; Gubellini et al., 2007) and comprising over a hundred protein complexes (Jackson et al., 2012; Woronowicz and Niederman, 2010; Woronowicz et al., 2013). The proteins that constitute the chromatophore are primarily the light harvesting (LH) complexes, photosynthetic reaction centers (RCs), cytb⁢c1 complexes, and ATP synthases, which cooperate to harvest light energy for photophosphorylation. The architecture of the chromatophore, reported in (Şener et al., 2007, 2010; Cartron et al., 2014), has been determined by combining atomic force microscopy (AFM) (Bahatyrova et al., 2004; Olsen et al., 2008), cryo-electron microscopy (cryo-EM) (Qian et al., 2005; Cartron et al., 2014), crystallography (Koepke et al., 1996; McDermott et al., 1995; Papiz et al., 2003; Jamieson et al., 2002), optical spectroscopy (Hunter et al., 1985; Sener et al., 2010), mass spectroscopy (Cartron et al., 2014), and proteomics (Jackson et al., 2012; Woronowicz and Niederman, 2010; Woronowicz et al., 2013) data. The composition of the chromatophore depends on growth conditions such as light intensity (Adams and Hunter, 2012; Woronowicz et al., 2011b, 2011a) and can also be influenced by mutations (Siebert et al., 2004; Hsin et al., 2010b).10.7554/eLife.09541.003Figure 1. Atomic structural model of a low-light-adapted chromatophore vesicle from Rba. sphaeroides. The model is based on AFM, EM, crystallography, mass spectroscopy, proteomics, and optical spectroscopy data (Cartron et al., 2014). The inner diameter of the vesicle is 50 nm. The model considered in this study is a variant of the one reported in (Cartron et al., 2014) , which features 63 LH2 complexes (green), 11 dimeric and 2 monomeric RC-LH1-PufX complexes (LH1:red; RC:blue; PufX:lime), 4 cytb⁢c1 (magenta), and 2 ATP synthases (orange), as well as 2469 BChls and 1542 carotenoids. Proteins are shown in surface representation. (A) Proteins and BChls of the chromatophore. Some of the light harvesting proteins are rendered transparent to reveal their BChl pigments. BChls are represented by their porphyrin rings only. See Video 1 presenting the vesicle. (B) Close-up of chromatophore showing its lipid membrane (transparent) along with its proteins colored as in (A). The membrane of 16,000 lipids contains the quinone/quinol pool of about 900 molecules. Energy conversion in the chromatophore proceeds in three stages: (I) light harvesting and electron transfer reducing the quinone pool; (II) quinone/quinol diffusion and exchange of quinols for quinones at cytb⁢c1 (thereby generating a proton gradient across the vesicle membrane) as well as diffusive motion of cytochrome c2 inside the chromatophore shuttling single electrons from cytb⁢c1 to RC; (III) utilization of the proton gradient for ATP synthesis. DOI: http://dx.doi.org/10.7554/eLife.09541.003 Video 1. Chromatophore structural model. A movie that shows a detailed structural model for the low-light adapted chromatophore vesicle as displayed in Figure 1 (Cartron et al., 2014). Presented is a rotating view of the vesicle comprising LH2 complexes (green), dimeric RC-LH1-PufX complexes (red-blue-lime green), dimeric cytb⁢c1 complexes (magenta), and ATP synthases (orange). For half of the model, proteins are shown in solid surface representation, and for the other half, proteins are shown as transparent surfaces with bacteriochlorophylls (BChls), represented by their porphyrin rings, shown as solid surfaces. DOI: http://dx.doi.org/10.7554/eLife.09541.004 10.7554/eLife.09541.004 The chromatophore displays organizational principles for the integration of multiple processes (Sener et al., 2010). Evolutionary competition at the organism level has driven photosynthetic subsystems toward optimal and robust function (Noy et al., 2006; Noy, 2008; Scholes et al., 2011) that can guide the design of artificial light harvesting devices such as biohybrid antennas (Harris et al., 2013) and nanopatterned light harvesting (LH) complex arrays (Reynolds et al., 2007; Vasilev et al., 2014; Patole et al., 2015). The development and improvement of such artificial, biological, or biohybrid light harvesting systems may alleviate mankind’s future energy demand (Blankenship et al., 2011). The functional principles displayed by the chromatophore and prevalent also in other photosynthetic systems include efficient excitonic coupling between components (Hu et al., 1997, 1998; van Grondelle and Novoderezhkin, 2006b; Olaya-Castro et al., 2008; Sener et al., 2011), the utilization of quantum coherence (Ishizaki and Fleming, 2009a; Strümpfer et al., 2012), photoprotection by carotenoids (Damjanović et al., 1999), accommodation of thermal fluctuations, studied through experimental (Visscher et al., 1989; van Grondelle et al., 1994; Pullerits et al., 1994; Gobets et al., 2001; Janusonis et al., 2008; Freiberg et al., 2009) as well as theoretical (Damjanović et al., 2002; Şener and Schulten, 2002; Ishizaki and Fleming, 2009b; van Grondelle and Novoderezhkin, 2006b; Strümpfer and Schulten, 2011, 2012a) methods. The chromatophore exhibits also the features of modularity, repair, and assembly of components (Hsin et al., 2010a), high quantum yield of organelle-scale pigment networks (Şener et al., 2007, 2010; Cartron et al., 2014), isolation of the electron transfer chains (Şener and Schulten, 2008), co-accommodation of competing functions such as efficient energy transfer and diffusion in the quinone/quinol pool (Lavergne et al., 2009; Sener et al., 2010), as well as adaptation to changing external conditions (Adams and Hunter, 2012; Woronowicz et al., 2011a; Niederman, 2013; Woronowicz et al., 2013). Energy conversion in the chromatophore proceeds in three stages as discussed below: (i) light harvesting and charge separation, converting quinone into quinol at a RC; (ii) diffusion of quinone/quinol in the chromatophore membrane and cytochrome c2 diffusion inside the chromatophore vesicle, resulting, at cytb⁢c1, in the generation of a proton gradient as well as a transmembrane electrochemical gradient across the chromatophore membrane (henceforth referred to collectively as proton gradient); (iii) utilization of the proton gradient, culminating in ADP binding and ATP release at ATP synthase. The quinone/quinol as well as the generated proton-motive force function as energy buffers between light harvesting and ATP synthesis stages. The proton gradient along with the redox states of the quinone/quinol pool are influenced by the enzymes succinate dehydrogenase, NADH dehydrogenase, cytochrome c oxidase, and ubiquinol oxidase (Bowyer et al., 1985; Klamt et al., 2008). A summary for the energy conversion processes in the chromatophore can be found in (Klamt et al., 2008; Sener et al., 2014). In addition to ATP synthesis, the chromatophore utilizes the generated proton motive force also for NADH production via NADH dehydrogenase (Klamt et al., 2008) and, thereby, for control of the quinone/quinol pool redox state. Other channels for proton gradient depletion are flagellar motility (Kojadinovic et al., 2013) and proton leak across the vesicle membrane. In the present study, we focus on the overall energy conversion characteristics of the molecular components identified in the current structural model (Cartron et al., 2014), namely LH2, RC-LH1, cytb⁢c1, and ATP synthase, where NADH dehydrogenase plays an indirect role. Efficient energy conversion requires some degree of robustness with respect to supramolecular organization, since no two chromatophores are likely to be identical. Though chromatophore vesicles share structural motifs (Bahatyrova et al., 2004; Cartron et al., 2014) that vary gradually with growth conditions, inevitable irregularities in the distribution of their constituent proteins and their quinone/quinol pools render chromatophores heterogeneous, requiring energy conversion processes to be insensitive to structural inhomogeneity. Robustness in photosynthetic systems had been demonstrated computationally for the excitation transfer step of light harvesting at the single protein level with respect to loss or rearrangement of pigments (Şener et al., 2002) as well as against fluctuations of pigment site-energies (Damjanovic et al., 2002) and at the vesicle level against deformations of the pigment network (Sener et al., 2010). Efficiency of energy conversion in a photosynthetic system is not straightforward to define, since it involves multiple interrelated subprocesses spanning both quantum mechanical and classical domains over timescales ranging from picoseconds to milliseconds (Blankenship, 2014; van Amerongen et al., 2000). A simple measure of conversion efficiency at the light harvesting stage is provided by the quantum yield, q, defined as the probability, upon the absorption of a photon by any pigment of the chromatophore, of charge separation at any RC ready for excitation-induced electron transfer to quinone. The quantum yield is solely a function of pigment network geometry, is independent of incident light intensity, and is found to be close to unity (Strümpfer et al., 2012; Sener et al., 2011) for initial chlorophyll light absorption; in case of carotenoid light absorption the quantum yield can be lower due to so-called covalent electronic excitation as argued in (Ritz et al., 2000b). Since excitation transfer does not constitute a rate-limiting step of photosynthetic energy conversion, the quantum yield is not a major limiting factor for the overall efficiency of the chromatophore. A comprehensive measure of chromatophore efficiency that also permits a limited comparison with photovoltaic systems is the conversion efficiency of captured solar energy to the chemical energy of the final photoproduct, namely ATP. Earlier studies of photosynthetic membrane systems include percolation theory-based models of quinone diffusion in Rhodospirillum (Rsp.) photometricum membranes (Scheuring et al., 2006), plastoquinone diffusion in thylakoid membranes (Kirchhoff et al., 2002), models of dissipative photoprotective behavior in Rsp. photometricum membranes (Caycedo-Soler et al., 2010), and stoichiometry-based rate kinetics (Geyer et al., 2007, 2010). In fact, long before any structural information of the light harvesting apparatus of purple bacteria was available, Vredenberg and Duysens (Vredenberg and Duysens, 1963) postulated that the total fluorescence yield can be expressed in terms of the ratio of closed and open RCs, after which random-walk models of excitation transfer were developed using a master equation formalism (Den Hollander et al., 1983). Prior to the availability of AFM imaging data (Bahatyrova et al., 2004), the supramolecular organization of chromatophores was suggested to feature RCs partially surrounded by LH-complexes facilitating efficient shuttling of quinones (Joliot et al., 1990, 1996; Jungas et al., 1999). The aim of the present study is to determine, based on a supramolecular structural model (Cartron et al., 2014), for the chromatophore of Rba. sphaeroides the ATP production rate as a function of illumination and vesicle stoichiometry along with the corresponding energy conversion efficiency. A low-light adapted chromatophore vesicle model is considered (Cartron et al., 2014), since low-light illumination, namely ≲10% of full sunlight, is typical for the habitat of purple bacteria (Woronowicz and Niederman, 2010; Blankenship, 2014). The quantum yield of excitation transfer for the pigment network geometry shown in Figure 1 is determined in terms of an effective Hamiltonian formulation. The processes subsequent to charge separation and the corresponding rate kinetics of ATP production are described in terms of chromatophore vesicle stoichiometry, instead of at atomic detail, by identifying rate-limiting steps. The organizational optimization of the chromatophore is considered in terms of the dependence of energy conversion on vesicle composition and illumination conditions. Results Based on the theoretical framework discussed in the Materials and methods section below, one can quantify how well the chromatophore performs in converting light energy into ATP synthesis and compare its performance characteristics, such as energy conversion efficiency, to the characteristics of other biological and artificial energy conversion systems. In particular, we examine below ATP turnover of the chromatophore as a function of light intensity and vesicle composition. The reader is urged to read the Materials and methods (Section 4) before proceeding further with the present section. ATP turnover rate as a function of illumination Previous studies showed that the quantum yield of excitation transfer, q, computed through Equation 8 below and discussed in greater detail in Supplementary Materials, is very high, namely, 85–94%, varying gradually with LH2:RC stoichiometry (Şener et al., 2007, 2010, 2011). For the vesicle presented in Figure 1 the quantum yield, q, has a value of 0.91, consistent with earlier studies (Şener et al., 2007, 2010; Cartron et al., 2014). Such high value for the quantum yield, close to the ideal limit of 1, is achieved because loss due to internal conversion and fluorescence arises much more slowly (rates about (1 ns)−1) than excitation transfer or charge separation at RC (rates about (10 ps)−1). Clearly, the quantum yield does not constitute a limiting factor for the overall energy conversion efficiency in the chromatophore. At very low light intensity, nearly all electronic excitation delivered to RCs contribute to the generation of a proton gradient across the membrane and to eventual ATP synthesis. With increasing light intensity, the cycling time of quinones at the RC, τR⁢C⁢(I) as given by Equation 19, increases; fewer RCs are found in a state available to receive photoexcitation, described by the probability, pRC⁢(I), given by Equation 13, and resulting in a corresponding loss of electronic excitation. The time-scale with which the quinone/quinol pool redox state adapts to a change in light conditions is reported to be about 0.5 s (Woronowicz et al., 2011a). The ATP turnover rate, kATP, calculated according to Equation (20), and the energy conversion efficiency, ηATP, calculated according to Equation (21), for a low-light adapted chromatophore vesicle (Figure 1) under steady-state illumination are presented in Figure 2. At light intensities equivalent to 1% and 3% of full sunlight, the vesicle is found to produce ATP molecules at a rate of 82 s−1 and 118 s−1, respectively. At the high-light limit, the ATP synthesis rate approaches 158 molecules s−1. These rates are consistent with experimental observations for continuous light-induced photophosphorylation, reported to be in the range of 0.017 molecules per BChl per second (Saphon et al., 1975) and 0.05 ATP molecules per BChl per second (Clark et al., 1983), corresponding to ~43 ATP molecules s−1 and ~130 ATP molecules s−1, respectively, for the vesicle shown in Figure 1. We note that the Clark estimate was reported for Rhodopseudomonas capsulata. The corresponding energy conversion efficiency, ηA⁢T⁢P, at the stated low-light intensities of 1% and 3% of full sunlight, calculated in the present study, are 12% and 7%, respectively. Notably, an upper-limit of 30% was estimated in (Hellingwerf et al., 1993) for the conversion efficiency of photosynthesis in Rba. sphaeroides. In comparison, the efficiency value, ηA⁢T⁢P, computed for the recently established model (Cartron et al., 2014) of the chromatophore (Figure 2) ranges between 0%–17%.10.7554/eLife.09541.005Figure 2. ATP production rate and energy conversion efficiency. (A) Steady-state ATP production rate, kATP, calculated according to Equation (20), and (B) energy conversion efficiency, ηATP, calculated according to Equation (21), of a chromatophore vesicle as a function of incident light intensity Φ. Solid curves correspond to the vesicle shown in Figure 1; dashed curves represent a similar vesicle with only a single cytb⁢c1. The vertical lines denote the light intensities corresponding to (I) 3% of full sunlight (30 W/m2), a typical growth condition for purple bacteria, and (II) full sunlight (1 kW/m2), respectively. Thus, for light intensities typical for the habitat of purple bacteria (1–5% of full sunlight; shaded area) the energy conversion efficiency ηATP of a chromatophore vesicle is between 0.12–0.04. DOI: http://dx.doi.org/10.7554/eLife.09541.005 The lower efficiency values, ηA⁢T⁢P, for the chromatophore at higher light intensities (Figure 2B) does not indicate a failing, since the chromatophore does actually produce slightly more ATP in high-light than in low-light illumination (Figure 2A), but rather reflects the optimization of purple bacteria for a low-light intensity habitat. The saturation of ATP synthesis with increasing light intensity seen in Figure 2 arises because quinol turnover capacity at the cytb⁢c1 complex becomes rate limiting at higher light intensities. The rate limiting property of cytb⁢c1 complexes was suggested by earlier studies (Lavergne et al., 2009; Geyer et al., 2010) and is discussed below in connection with Equation (14) in Materials and methods. The maximal electron processing capacity of all cytb⁢c1 complexes is estimated (see Equation (14) in Materials and methods) to be 2×nB⁢τB-1=320⁢s-1, where nB=4 is the number of cytb⁢c1 dimers and τB = 25 ms is the quinol turnover time at cytb⁢c1 (Crofts, 2004) and the prefactor 2 accounts for every quinol transferring two electrons. The electron processing capacity at cytb⁢c1 becomes equal to the total RC electron turnover rate, I⁢q, at a light intensity of 9 W/m2, i.e., at approximately 1% of full sunlight. As illumination exceeds this low-light value, RC electron turnover is limited by the electron processing capacity of cytb⁢c1, leading to a gradual saturation of proton gradient formation and ATP turnover, as seen in Figure 2A, thereby reducing the efficiency of ATP synthesis (Figure 2B). Rate limitation of ATP synthesis by the cytb⁢c1 turnover capacity can be related also to the availability of quinones at the RC. In the absence of bound quinone, excitations delivered to a RC are wasted, except if the excitation escapes from the RC and reaches another RC ready for quinone reduction, especially within the same RC-LH1 dimer. However, with increasing illumination the likelihood of nearby RCs having quinones available also diminishes, excitation energy is lost, and energy conversion efficiency is reduced. The probability of a RC being ready for quinone reduction, pRC⁢(I), is given by Equation (13) in the Materials and methods section. With increasing illumination, pRC⁢(I) decreases, thereby reducing the overall conversion efficiency, ηATP. At 1% of full sunlight, pRC⁢(I) assumess the value of 0.73, which, according to Equation (13), drops to 0.23 at 5% of full sunlight. Remarkably, the role of closed and open RCs in determining the overall efficiency of the photosynthesic apparatus had been already pointed out long before any structural details were known (Vredenberg and Duysens, 1963). The rate limiting effect of cytb⁢c1 can be further illustrated considering the efficiency of chromatophores with fewer cytb⁢c1 complexes compared to the ones shown in Figure 1. As indicated by the dotted lines in Figure 2A,B, describing ATP synthesis in a chromatophore with a single cytb⁢c1 dimer, a lower number of cytb⁢c1 dimers results in a reduction of the ATP production rate, kATP, and, accordingly, in a lower conversion efficiency, ηATP. A comparison with plant light harvesting efficiencies is not straightforward: efficiency for biomass production is significantly lower than the aforementioned thermodynamic efficiency; in fact, only as little as 1% of total incident solar energy is stored by crop plants as biomass (Blankenship et al., 2011). One might wonder how the chromatophore compares to engineered photovoltaic devices. At peak solar intensity photovoltaic-driven electrolysis is reported to have an energy conversion efficiency of 5–15% (Blankenship et al., 2011). However, comparison of efficiency alone overlooks issues such as stability and reclaimability of the energy stored in the final products. More refined measures of efficiency need to include total integrated cost of components, life expectancy, repair and maintenance. Optimality of vesicle composition for ATP production Evolutionary pressure toward greater fitness at the organism level results in the composition and architecture of photosynthetic systems to display adaptation toward optimal function (Xiong et al., 2000; Şener and Schulten, 2008; Blankenship, 2014). Such adaptation has been reported for the individual protein level; it is not as well understood at the system integration level. For instance, pigment networks of individual light harvesting proteins were reported to display optimality and robustness in their quantum yield with respect to the spatial organization of pigments and the site energy distribution (Şener et al., 2002; Noy et al., 2006; Damjanovic et al., 2002); a similar robustness was reported with respect to size-scaling deformations of an entire vesicle (Sener et al., 2010). Prior studies did not take into account optimization of the complete energy conversion process, including ATP synthesis, the effects of vesicle composition influenced by growth conditions such as light intensity (Niederman, 2013; Woronowicz et al., 2013), the regulation of the redox state of the quinone/quinol pool (Klamt et al., 2008), or the effects of cell-scale concentration and connectivity of chromatophores also influenced by light intensity at growth (Tucker et al., 2010). In the following, the effect of vesicle composition on the ATP turnover rate is examined in order to determine the degree of optimality of the vesicle composition for ATP production. The vesicle shown in Figure 1 is used as a reference point for comparison with chromatophores of alternate composition, As composition variables, the number of dimeric cytb⁢c1 complexes, nB, and the number of dimeric RC-LH1-PufX complexes, nL, are considered for a two-parameter (nB,nL) family of vesicles with the same surface area as the reference vesicle (Figure 1). The dependence of the steady-state ATP turnover rate, kATP⁢(nB,nL;I), on nB and nL is determined according to Equations 17,19,20, where nL=2×nRC. In order to avoid massive computation, vesicles are not constructed explicitly. Instead, the corresponding quantum yield q is estimated by a linear interpolation on the LH2:RC stoichiometry based on earlier reported values (Şener et al., 2007, 2010, 2011) as described in Materials and methods (Equation 10). Since q varies very little with vesicle composition, the dependence of kATP on composition is dominated primarily by the explicit nB and nRC dependence in Equations 19,20. The rate kATP⁢(nB,nL;I) is shown in Figure 3 for light intensities equal to 1% and 3% of full sunlight. The respective ATP synthesis rates for the reference vesicle in Figure 1 under these two illumination conditions are 82 and 118 ATP molecules/s, respectively, (marked by circles in Figure 3) which corresponds to 79% and 50% of the maximum possible rate (marked by crosses in Figure 3) among all possible (nB,nL) values at that illumination. Clearly, steady state ATP synthesis is not optimized by the vesicle composition shown in Figure 1. The turnover rate, kATP, would be improved by an nB:nL ratio that is greater than the native value of 1:3 (Crofts, 2004; Cartron et al., 2014), as suggested also by a comparison of the turnover times at cytb⁢c1 and RC (τB/τL≃8).10.7554/eLife.09541.006Figure 3. Effect of vesicle composition on steady-state ATP production at different light intensities. Vesicle composition is given in terms of the number of cytb⁢c1 dimers (nB) and of RC-LH1-PufX dimers (nL) for vesicles featuring identical surface area; LH2 composition of the vesicle is determined by considering the vesicle shown in Figure 1 as a reference point and adjusting the number of LH2 complexes to compensate for the changes in the number of cytb⁢c1 and RC-LH1-PufX dimers to cover the vesicle surface. ATP production rate, kATP, is shown for (A) 1% of full sunlight (10 W/m2) and (B) 3% of full sunlight (30 W/m2), determined according to Equation (20). The two RC-LH1-PufX monomers of the vesicle in Figure 1 were counted as a single dimer for the purposes of this plot. The reference vesicle (Figure 1) is represented by a circle, corresponding to an ATP production rate of 82 s−1 (118 s−1), i.e. 79% (51%) of the maximum possible rate among all stoichiometries, for 1% (3%) of full sun light. The optimal vesicle composition for each illumination is represented by a cross; the corresponding LH2 count for optimal composition is 93 (74) at 1% (3%) of full sunlight as compared with 63 for the reference vesicle (circle). The ATP production rate is marginally greater for vesicles that contain more cytb⁢c1 and LH2 complexes at the expense of fewer RC-LH1-PufX complexes as compared with the reference vesicle. This increase in ATP production rate results from cytb⁢c1 being the rate-limiting component in the energy conversion process. DOI: http://dx.doi.org/10.7554/eLife.09541.006 A reason for the aforementioned suboptimal (nB,nL) values in native low-light adapted vesicles might be protection against light-induced damage that can arise at high illumination via destruction of the vesicle membrane through overacidification. Though typical illumination levels in habitats of purple bacteria are low, occasional surges in light intensity are inevitable. During sustained (>1 s) high illumination intervals, a proton turnover unhindered by a low (nB=4) cytb⁢c1 stoichiometry can exceed the turnover capacity of the ATP synthases, resulting in overacidification of the vesicle interior, harming the integrity of the chromatophore membrane and its proteins. The observed nB value of 4, apparently suboptimal for most light intensities (Figure 3), ensures that during sustained over-illumination proton turnover is limited by cytb⁢c1 to a rate below the synthesis capacity of ATP synthase (Lavergne et al., 2009; Geyer et al., 2010), thus preventing overacidification. The (nB,nL) value also has an effect on the size of the quinone/quinol pool relevant for intermittent energy storage under fluctuating light conditions, since the number of quinones in the system correlates with the number of RCs (Comayras et al., 2005; Woronowicz et al., 2011a; Cartron et al., 2014). Energy conversion through the quinone/quinol pool also involves electron exchange processes from outside the chromatophore as furnished, for example, through the enzymes NADH dehydrogenase and succinate dehydrogenase (Klamt et al., 2008). The turnover capacities of cytb⁢c1 and ATP synthase are compared in Materials and methods in relation to the rate limitation of energy conversion by the cytb⁢c1. A single ATP synthase is sufficient to take advantage of proton turnover of an entire chromatophore (Etzold et al., 1997). Additional ATP synthases reported in chromatophore vesicles (Cartron et al., 2014) appear to provide necessary redundancy, since an isolated chromatophore without ATP synthase is non-functional. In this regard, it is of interest that vesicles have been found to occasionally fuse through formation of membrane tubes (Tucker et al., 2010) permitting passage of protons between neighboring chromatophore vesicles, thereby sharing their proton gradients with the ATP synthases of multiple vesicles, reducing the need for back-up ATP synthases and even permitting less than one ATP synthase per vesicle. Robustness requirements for protecting the vesicle against damage under environmental strain apparently supersede optimality constraints for steady state conditions. A photosynthetic vesicle adapted for steady-state illumination at higher light intensities than considered in this study would require a larger number of cytb⁢c1 to maximize ATP production, along with more than the 1–2 ATP synthases observed per vesicle (Cartron et al., 2014). Discussion The combined structural and functional model of a low-light adapted chromatophore (Cartron et al., 2014) permits a quantitative description of ATP synthesis at different light intensities. The energy conversion efficiency, ηATP, is determined to be ~12%–4% at the low-light conditions typical for purple bacterial habitats (1%–5% of full sunlight), dropping rapidly to ≲0.1% beyond full sunlight conditions. Moderate levels of illumination saturate the bacterial light harvesting apparatus lowering its efficiency, whereas plants and photovoltaic devices function efficiently at high light intensities. The efficiency curve determined in the present study for the purple bacterial chromatophore (Figure 2B) indicates specialization for low-light intensities. The primary rate-limiting component among the energy conversion steps in the chromatophore appears to be quinol turnover at cytb⁢c1, as discussed in Section 4.2. The rate limitation at cytb⁢c1, as compared with the ATP synthase turnover capacity, prevents the generation of an overly strong proton gradient at sustained high-light conditions, thereby protecting the chromatophore against overacidification of its interior and assuring vesicle integrity. As the light intensity I increases, photoexcitations are more likely to be dissipated as the probability for a RC to have a quinone or semiquinone ready to accept an electron, pR⁢C⁢(I), decreases. The chromatophore composition appears to be suboptimal for ATP production under steady-state illumination. The chromatophore is apparently a highly specialized device that performs its energy conversion function robustly for low average light intensity, while featuring protective measures that dissipate energy at higher light intensity. Robustness against damage, such as overacidification of the membrane due to sustained overillumination, appears to supersede optimality under idealized conditions, such as steady state illumination. The present study focuses on steady state energy harvesting in the chromatophore without explicitly modeling the spatial dynamics of the charge carriers (quinone/quinol and cytochrome c2), the redox states of the proteins (RC and cytb⁢c1), proton leakage through the membrane, or the coupling of NADH dehydrogenase to the proton-motive force. A more complete description of chromatophore function requires placement of NADH dehydrogenase, along with possibly succinate dehydrogenase, cytochrome c oxidase, and ubiquinol oxidase in the chromatophore membrane, the presence of which would also affect the energy conversion efficiency determined in this study. The added enzymes need to be described along with their reactions with redox partners located in the cell’s cytoplasm. In particular, a non-steady state formulation is necessary to account for spatial heterogeneity and light intensity dependence of the redox states of the proteins and the charge carriers in the chromatophore. The present study differs from earlier studies in functional modeling of the chromatophore (Geyer et al., 2007, 2010) in several respects: first, it is based on an explicit atomic-detail structural model; second, instead of employing many (over 30) adjustable parameters, few experimentally determined rate constants are employed to describe the rate determining steps; third, a steady-state description is chosen such that energy conversion steps that are not rate-limiting can be left out of the kinetic model. Nonetheless, earlier and present studies give similar results for the overall ATP synthesis rate at saturation, since this rate is determined largely by the total turnover capacity of cytb⁢c1 complexes as a rate-limiting component. A key role in chromatophore energy conversion involves proton translocation, generating and using proton motive force. The present treatment does not resolve individual translocation steps, but rather assumes that the individual steps taking place at the overall RC, cytb⁢c1, and ATP synthase proton reactions can be treated as a single reaction event. Primary conclusions reached presently would not be affected by a more detailed description, i.e., the model is robust with respect to the neglect of explicit modeling of individual proton translocation steps and proton motive force conversion. Integrative models of organelle function such as the one presented here provide a bridge between experimental methods that do not resolve temporal and spatial detail needed for establishing physical mechanisms and microscopic simulations that span the multiple length and time scales relevant for the function of living cells. Materials and methods In the following, structural organization and energy conversion in the chromatophore are described in terms of a kinetic model. It is highly recommended that the text below is read before Sections 2 and 3. First, the supramolecular organization of a low light-adapted chromatophore vesicle is introduced. Next, the energy conversion processes are characterized: excitation transfer, diffusion of quinones/quinol and of cytochrome c2, and ATP synthesis. The description is based on steady state kinetics. Inhomogeneities of the quinone/quinol and cytochrome c2 pools and of the membrane proton gradient are not modeled; instead, the three attributes are assumed to function as homogeneous buffers of energy storage. The framework outlined is used to define three different measures of efficiency for the chromatophore: (i) quantum yield, q, (ii) quinol conversion (Q→Q⁢H2) probability, ηQ, i.e., the probability that an absorbed photon is successfully utilized for quinol formation at a RC, and (iii) energy conversion efficiency, ηATP, i.e., the ratio of the energy stored in the conversion ADP→ATP to incident solar energy absorbed. Supramolecular organization of a chromatophore vesicle adapted to low-light illumination As already stated, the structural model of the chromatophore considered in the present study is a variation of the model reported in (Cartron et al., 2014). The primary components of chromatophore vesicles in purple bacteria, as depicted in Figure 1, are, in order of energy utililization (Cogdell et al., 2006,Cartron et al., 2014): (i) light harvesting complex 2 (LH2) (Koepke et al., 1996; Papiz et al., 2003); (ii) light harvesting complex 1 (LH1 [Qian et al., 2008; Sener et al., 2009]); (iii) RC (Jamieson et al., 2002; Strümpfer and Schulten, 2012a); (iv) cytb⁢c1(Crofts, 2004; Crofts et al., 2006); and (v) ATP synthase (Feniouk and Junge, 2009; Hakobyan et al., 2012). RC-LH1 complexes typically form dimeric RC-LH1-PufX complexes facilitated by the polypeptide PufX (Qian et al., 2013; Sener et al., 2009), although monomeric complexes are also found in membranes from photosynthetically grown cells at a ratio of approximately 10% (Olsen et al., 2008). The chromatophore in Figure 1 exhibits for the LH2:RC complexes a stoichiometry of 2.6:1 and corresponds to a low-light-adapted vesicle as described in (Sener et al., 2010; Cartron et al., 2014). In a typical vesicle, about a hundred protein complexes, LH2 and RC-LH1-PufX, form an efficient light harvesting network (Şener et al., 2007, 2010) supplying electronic excitation energy for the conversion of quinones to quinols. The quinols produced at the RC are converted back to quinones by cytb⁢c1 to generate a proton gradient across the chromatophore vesicle membrane, which, in turn, is consumed by the ATP synthase for the synthesis of ATP from ADP and phosphate. The electrons from quinol-to-quinone conversion are shuttled back to the RC by cytochrome c2 acting inside the vesicle. These energy conversion processes are illustrated in Figure 4. We note that the experimental data (Saphon et al., 1975; Clark et al., 1983) used to test the present energy conversion model based on (Cartron et al., 2014) were not obtained with chromatophores in vivo, but for a suspension of chromatophores in a pH-buffer; the energy conversion processes as coupled to the entire bacterium are inevitably more complex than portrayed here.10.7554/eLife.09541.007Figure 4. Processes involved in energy conversion in the photosynthetic chromatophore. (A) Energy conversion processes starting after initial light absorption are divided into three stages: (1) quinol production at RC as a result of excitation transfer; (2) diffusion between RC and cytbc1 of quinone/quinol and cytochrome c2, together with quinol-to-quinone conversion resulting in a proton gradient across the vesicle membrane; (3) utilization of proton gradient for ATP synthesis. (B) Chromatophore components, in which stages (1–3) take place, include LH2 (green), LH1(red)-RC(blue), cytbc1(purple), and ATP synthase (brown) complexes as well as the lipid phase (olive; see also Figure 1B). DOI: http://dx.doi.org/10.7554/eLife.09541.007 Atomic level structural models of chromatophores have been presented earlier (Şener et al., 2007, 2010; Hsin et al., 2010b; Sener et al., 2011; Chandler et al., 2014) for Rba. sphaeroides and Rsp. photometricum and their mutants. The supramolecular organization of the vesicles in Rba. sphaeroides was determined primarily by AFM and EM images of intact membrane domains (Bahatyrova et al., 2004; Frese et al., 2004; Scheuring et al., 2007; Olsen et al., 2008; Qian et al., 2008; Scheuring and Sturgis, 2009), whereas the stoichiometry of light harvesting proteins was determined by optical spectroscopy (Sener et al., 2010) and mass spectrometry (Cartron et al., 2014). Vesicle models were subsequently constructed by mapping planar membrane patches viewed through AFM imaging back onto the parent spherical domains (Şener et al., 2007), adjusting for the observed packing density (Olsen et al., 2008) and the spatial arrangement patterns (Hsin et al., 2009; Qian et al., 2008) of the constituting proteins. The chromatophore model shown in Figure 1 comprises, in addition to the aforementioned constituent proteins, 16,000 lipids and 900 quinones, corresponding to a system containing 100 million atoms, including solvent. This system has been equilibrated through a 100 ns MD simulation to test the viability of the model employed. However, molecular dynamics simulations of the chromatophore describing energy conversion processes are not considered in the current study, because the large system size combined with timescales of energy conversion reaching milliseconds render a straightforward simulation prohibitive. Instead, the current study aims to describe key rate limiting components of energy conversion processes, such as quinone diffusion and turnover at cytb⁢c1 as discussed below, to guide future simulation efforts. The atomic detail model is used below for the computation of the quantum yield, but rate kinetics subsequent to charge separation is described in terms vesicle stoichiometry only, with key rate constants taken from experimental studies. Early chromatophore models prior to (Cartron et al., 2014) account only for LH proteins, whereas in proteomics studies, hundreds of different types of non-LH peptides are actually identified, including ATP synthase, cytb⁢c1, membrane assembly factors, as well as proteins of unknown function (Jackson et al., 2012; Woronowicz and Niederman, 2010). Most of these components are notably unresolved in AFM images. Assignment of cytb⁢c1 was recently achieved through EM and AFM studies using gold nanoparticle labeling, revealing separated regions containing one or more cytb⁢c1, suggested to be located within lipid- and quinone-enriched membrane domains (Cartron et al., 2014). It is plausible that cytb⁢c1 induces different curvature profiles in membrane domains compared to the LH-rich constant-curvature regions predominant in AFM images. Such a curvature-induced separation of protein domains is also supported by experimental (Frese et al., 2004; Sturgis and Niederman, 1996) and computational (Frese et al., 2008; Chandler et al., 2009; Hsin et al., 2009, 2010a) studies that established the role of LH2 and RC-LH1-PufX domains in determining membrane shape. Induced curvature profiles are known to exert a segregating force between different types of proteins in the membrane (Frese et al., 2008). Mass spectrometry showed that the RC:cytb⁢c1 stoichiometry is 3:1 (Cartron et al., 2014), consistent with earlier observations (Crofts, 2004; Crofts et al., 2006), corresponding to approximately 4 cytb⁢c1 dimeric complexes per vesicle. Chromatophore vesicles typically contain 1–2 ATP synthases (Feniouk et al., 2002; Cartron et al., 2014). Proteomics studies suggest preferential co-location of ATP synthase with LH2 subunits (Woronowicz and Niederman, 2010). Consequently, ATP synthase locations were assigned to LH2-rich regions of the membrane (Cartron et al., 2014). The low-light adapted vesicle studied here contains 63 LH2 complexes, 11 dimeric and 2 monomeric RC-LH1-PufX complexes, 4 dimeric cytb⁢c1 complexes, and 2 ATP synthases, in a spherical vesicle of 50 nm inner diameter based on a variation of the model reported in (Cartron et al., 2014) and shown in Figure 1 (see also Video 1). Transmembrane proteins beyond those of the light harvesting-cytb⁢c1-ATP synthase model shown in Figure 4, namely NADH dehydrogenase, succinate dehydrogenase, cytochrome c oxidase, and ubiquinol oxidase, are associated with controlling the redox state of the quinone/quinol pool in the chromatophore (Klamt et al., 2008). These proteins, presented schematically in Figure 5, indirectly couple the chromatophore proton gradient to metabolic reactions in the cytoplasmic part of the bacterial cell. Indeed, the chromatophore structure shown in Figure 1 may accommodate, by removal of LH2 complexes near cytb⁢c1 complexes (the latter referred to as complex III in the respirasome of mitochondria [Dudkina et al., 2011]), the placement of adjacent NADH dehydrogenase complexes (referred to as complex I) in an arrangement similar to that in respirasomes as reported in (Dudkina et al., 2011). First simulations, employing the complex I structure reported in (Baradaran et al., 2013), have demonstrated that the chromatophore can adapt to the necessary local shape change.10.7554/eLife.09541.008Figure 5. Regulation of the quinone/quinol-pool redox state in the chromatophore involves transmembrane proteins beyond those included in the light harvesting-RC-cytb⁢c1-ATP synthase system described in (Cartron et al., 2014) and shown in Figure 4. Succinate dehydrogenase and NADH dehydrogenase regulate the quinone/quinol redox state in the chromatophore membrane. This regulation influences the light intensity dependence of the ATP production rate in the chromatophore by changing the likelihood of finding available quinones at the RC. Two further proteins involved in redox kinetics are cytochrome c oxidase and ubiquinol oxidase, which are not shown. DOI: http://dx.doi.org/10.7554/eLife.09541.008 Description of energy conversion in the chromatophore The overall aim of the present study is to determine the ATP synthesis rate as a function of chromatophore vesicle illumination and composition establishing, thereby, the energy conversion efficiency. The three stages of energy conversion in the chromatophore introduced above are summarized in Figure 4. These stages span time scales ranging from femto- and picoseconds (transfer of excitations) to milliseconds (diffusion of quinols, quinones, and cytochrome c2; ATP synthesis), involving both classical and quantum dynamics. The absorbed light power I is given in units of photons absorbed per second for the entire vesicle, i.e., it holds approximately, I=F⁢σt⁢o⁢t⁢a⁢l where F is the flux of useable photons and σt⁢o⁢t⁢a⁢l is the total absorption cross-section of the chromatophore determined via the functional absorption cross-section reported in (Woronowicz et al., 2011a). Key quantities describing energy conversion are the quinone-to-quinol formation rate kQ→QH2⁢(I), the quinol-to-quinone use rate kQH2→Q⁢(I) and the ATP synthesis rate, kATP⁢(I), all functions of the absorbed light power I. For stationary illumination, assumed here, the chromatophore kinetics becomes stationary and, as a result, the rates kQ→QH2⁢(I), kQH2→Q⁢(I), and kATP⁢(I) must be identical, (1) kQ→QH2(I)=kQH2→Q(I)=kATP(I). We note here that every net quinol→quinone conversion event at cytb⁢c1, due to the so-called Q-cycle (Crofts, 2004), results in the release of four protons into the vesicle interior (two at cytb⁢c1 and two at RC, for each quinol passage), which coincidentally happens to be, in the present system, the same number of protons as the ones that have to move back over the membrane to produce one ATP molecule at ATP synthase (based on the assuption of a 12-subunit c-ring of the ATP synthase). Under steady state conditions, the rate, kATP(I), is then equal to kQH2→Q⁢(I) and kQ→QH2⁢(I). The quinone/quinol pool in the lipid phase of the vesicle and the proton gradient across the vesicle membrane act as temporary energy buffers between light harvesting and ATP synthesis (Feniouk and Junge, 2009; Clark et al., 1983). Under the steady-state conditions assumed here, quinone-quinol pool and redox states of RC/cytb⁢c1 are assumed to feature spatially homogeneous distributions. As a result, individual diffusive processes of quinone/quinol, cytochrome c2, and protons do not need to be modeled, and the aforementioned energy buffers are determined solely by incident light intensity and quinol→quinone turnover capacity of cytb⁢c1, the latter constituting the rate limiting conversion process as discussed below. Typical proton diffusion timescales are on the order of microseconds (Agmon, 1995), i.e., not rate limiting compared to turnover at cytb⁢c1 and, therefore, do not affect significantly overall conversion rates. In going beyond steady-state conditions, a simulation of quinone mobility in a lamellar chromatophore membrane has recently been achieved for a 20 million atom, 150 ns simulation (Chandler et al., 2014); however, the time scale covered is not long enough to observe long-range positional relaxation of the quinone/quinol pool. This non-stationary behavior needs to be addressed by a coarser description like the ones employed for cell-scale modeling (Roberts et al., 2013). In purple bacteria there are proton gradient consumption channels other than ATP synthesis. These channels include: flagellar motility (Kojadinovic et al., 2013); NADH/NADPH synthesis (Klamt et al., 2008; Blankenship, 2014) involving respiratory protein complexes in the chromatophore vesicle; leakage across the membrane. These channels are not included in the chromatophore kinetics described below, though the influence of NADH dehydrogenase is implicitly accounted for as explained below. Stage I: Light absorption, excitation energy transfer, and quinol formation The first stage of energy conversion in the chromatophore begins with light absorption by carotenoid and BChl pigments in the light harvesting complexes LH1 and LH2 leading to electronic excitation of individual pigments. Carotenoids transfer excitation within less than a picosecond to a nearby BChl (Damjanović et al., 1999; Berera et al., 2009) and also play a role in quenching triplet states of BChls through reverse excitation transfer (Ritz et al., 2000a). The electronic excitations of BChls embedded in LH1 and LH2 are reviewed in (Hu et al., 1998, 2002; Cogdell et al., 2006; van Grondelle and Novoderezhkin, 2006b; Kosztin and Schulten, 2014). These excitations form so-called exciton states, excitations shared among LH1 or LH2 BChls (Ma et al., 1997; Bradforth et al., 1995) coherently (Strümpfer et al., 2012; Ishizaki and Fleming, 2009b; Rebentrost et al., 2009). Electronic excitation is transferred efficiently in the form of excitons between light harvesting complexes (Hu et al., 1997; Ritz et al., 1998; Janusonis et al., 2008; Ishizaki and Fleming, 2009b). Exciton-based excitation transfer in the chromatophore proceeds within 10–100 ps (Sener et al., 2011), first among the LH2s, then from LH2 to LH1, and finally from LH1 to the four BChls of the RC (Visscher et al., 1989; Beekman et al., 1994; Strümpfer and Schulten, 2012a; Sener et al., 2009). In the RC, the excitation quickly settles onto the so-called special pair BChls (Small, 1995; Damjanović et al., 2000), where it induces the transfer of an electron (Pawlowicz et al., 2008; Jordanides et al., 2004). This transfer proceeds stepwise to reach a quinone molecule, Q, attracted into the RC from the quinone/quinol pool of about 900 molecules (Cartron et al., 2014). The quinol and quinone molecules of the pool are inter-converted at RC and cytb⁢c1 (Crofts, 2004) (see Figure 4). The electron transferred in the RC is joined on the quinone by a proton, turning Q into semi-quinone, Q⁢H. Repeating the reaction turns Q⁢H into quinol, Q⁢H2. In converting Q to Q⁢H2 two electron charges move from near the inside of the chromatophore (where the special pair BChls are located and the electron potential is low) to near the cytoplasmic exterior of the chromatophore (where the quinone is bound and the electron potential is high), i.e., to the cytoplasmic side; the protons are attracted from the exterior of the chromatophore vesicle. Freshly formed quinol is released by the RC into the lipid phase of the chromatophore rejoining the quinone/quinol pool. The efficiency of the 10–100 ps light harvesting step is measured by the so-called quantum yield, q, namely the probability that light absorption leads to electron transfer in a RC with a Q or Q⁢H bound to receive the electron. The quantum yield can be calculated as reported in (Sener et al., 2011, 2010). Electronic excitation energy absorbed directly or indirectly (through carotenoids) by a BChl is rapidly shared between BChls within individual LH1 and LH2 light harvesting complexes (Cory et al., 1998), forming, within about a ps, thermally equilibrated exciton states as established experimentally (Visser et al., 1996; Jimenez et al., 1997; Valkunas et al., 2007) as well as computationally (Strümpfer and Schulten, 2009; Strümpfer et al., 2012). The exciton states of the BChls of each complex are determined as eigenstates of the effective Hamiltonian HI, accounting for the Qy excitations and their coupling inside LH1, LH2 and RC as described in (Strümpfer et al., 2012), (2) HI=∑i=1NIEiI∣i⟩⟨i∣+∑i>j>0N1VijI(∣i⟩⟨j∣+∣j⟩⟨i∣). Here, the index I is employed to label one of the pigment-protein complexes, namely one of 63 LH2s, 24 LH1s, and 24 RC complexes for the vesicle shown in Figure 1, with NI BChls; |i⟩ corresponds to the Qy excitation of BChl i with excited state energy EiI;VijI accounts for the respective Qy-Qy coupling among BChls i and j. Tables S2 and S3 in Supplementary Materials list the BChl coordinates as well as the constants employed in this study and discuss the computation of the quantum yield in greater detail. The coupling Vi⁢jI in Equation (2) can be computed for well separated pigments (ri⁢j>1 nm) using the point-dipole approximation (Ritz et al., 2001; Sener et al., 2011), employing, (3) Vi⁢j=C⁢(𝐝^i⋅𝐝^jri⁢j3-3⁢(𝐝^i⋅𝐫i⁢j)⁢(𝐝^j⋅𝐫i⁢j)ri⁢j5), where 𝐝^i is the transition dipole moment unit vector of pigment i, 𝐫i⁢j is the vector joining pigments i and j; the coupling constant C has the value C=348,000⁢Å3 cm-1 (using wavenumbers as unit of energy) (Şener et al., 2007, 2010). Couplings between closely spaced pigments (ri⁢j<1 nm) require quantum chemical calculations as described in (Damjanović et al., 1999; Tretiak et al., 2000). The exciton states |α)=∑ici⁢α|i⟩ and the associated energies ϵα correspond to the eigenstates defined through Hℐ|α)=ϵα|α). As electronic excitations settle within about 1 ps into the Boltzmann-populated excitons (Strümpfer et al., 2012; Strümpfer and Schulten, 2009), excitation transfer among LH2 and LH1 involves the excitons, not individual chlorophyll or carotenoid excitations. The rate of excitation transfer between a donor complex I and an acceptor complex J is given by (Ritz et al., 2001; Şener et al., 2007, 2011) (4) kIJ=2πℏ∑μ∈I∑ν∈JpμI|(μ|HIJ|ν)|2Jμν, where HI⁢J is the matrix of interactions between the excited states of pigments in complexes I and J, and (5) Jμ⁢ν=∫d⁢E⁢SμD⁢(E)⁢SνA⁢(E), is the spectral overlap between donor exciton state |μ) and acceptor exciton state |ν) in units of (1/energy) (Sener et al., 2011); SμI⁢(E) and SνJ⁢(E) are the normalized (∫d⁢E⁢S⁢(E)= 1) spectra for emission of the donor (D) and absorption of the acceptor (A), respectively; pμI in Equation (S6) are the populations of donor exciton states, which, as stated, become very rapidly (~1 ps) (Strümpfer and Schulten, 2009) Boltzmann-distributed such that pμI are given by (6) pμI=e-β⁢ϵμ∑γ∈Ie-β⁢ϵγ. The above description is known as the generalized Förster theory (Förster, 1948; Novoderezhkin and Razjivin, 1996; Hu et al., 1997; Sumi, 1999; Scholes et al., 2001). For reviews see (van Grondelle and Novoderezhkin, 2006a; Sener et al., 2011; Strümpfer et al., 2012). Excitation transfer kinetics in the chromatophore was reported experimentally in (Woodbury and Parson, 1984; Visscher et al., 1989; Crielaard et al., 1994; Hess et al., 1994, 1995). Exciton migration across the network of light harvesting complexes in the chromatophore can be described by a rate matrix 𝒦 which is constructed from inter-complex exciton transfer rates kI⁢J, the latter given by Equation (S6), as follows (Sener et al., 2010, 2007) (7) (𝒦)I⁢J=kJ⁢I-δI⁢J⁢(∑MkI⁢M+kdiss+kCS⁢δI,RC), where I,J are defined as in Equation (S6); kdiss=1/ns is the rate of excitation loss due to internal conversion; kCS=1/(3ps) is the rate of charge separation at the RC (Ritz et al., 2001); δI,RC assumes the value 1 if complex I is a RC and the value 0 otherwise. The quantum yield, q, is the probability for an absorbed photon to initiate charge transfer at a RC ready for electron transfer; q is given for an initial state vector 𝐏⁢(0) by (Sener et al., 2011, 2007; Ritz et al., 2001) (8) q=-kCS⁢(𝟏𝐑𝐂)T⋅𝒦-1⋅𝐏⁢(0) where the components of the vector (𝟏𝐑𝐂) are (𝟏𝐑𝐂)I=δI,RC; the initial state, 𝐏⁢(0), corresponds to every BChl in the system being equally likely to be excited by photon absorption and accordingly is given by (9) (𝐏⁢(0))I=NI/(∑JNJ), where NI is the number of BChls in complex I as indicated above. The effect of the initial state, 𝐏⁢(0), on the quantum yield, q, arising, for example, due to wavelength-dependent absorption, is considered in (Şener et al., 2007), with the result that q is altered by less than 3%. Therefore, wavelength dependence of q, through corresponding changes in 𝐏⁢(0), is not considered further in the present study. The quantum yield given by Equation (8) for the vesicle shown in Figure 1 is 0.91. For alternate vesicle compositions considered in Figure 3, the quantum yield q is not computed by an explicit construction of vesicles to avoid massive computation; instead, q is approximated as a linear interpolation between the values reported earlier for high LH2:RC and low LH2:RC chromatophore vesicles (Şener et al., 2007, 2010, 2011), namely between q=0.85 and q=0.95. For a vesicle containing nB cytb⁢c1 dimers and nL LH1-RC dimers, the corresponding number of LH2 complexes nLH2⁢(nB,nL) is estimated by the excluded surface resulting from changes in nB and nL with respect to the reference vesicle in Figure 1. The corresponding quantum yield is estimated according to (10) q=0.91+0.0152⁢(s0-s) chosen to reproduce the correct value of q for the reference vesicle (Figure 1) as well as for the low LH2 limit (Sener et al., 2011); here s=nLH2/(2⁢nL) is the LH2:RC stoichiometry, which for the reference vesicle equals s0=2.625; a lower limit of q=0.85 is imposed to account for high LH2:RC vesicles (Şener et al., 2007), where the linear interpolation breaks down. Validity of the generalized Förster formulation, thus outlined, has been demonstrated by excitation transfer calculations employing the so-called hierarchy equation of motion formalism of stochastic quantum mechanics (Ishizaki and Fleming, 2009b; Strümpfer and Schulten, 2012b); the calculations show that photoexcitation of chromatophore BChls relaxes into a Boltzmann occupancy of exciton states within approximately 1 ps, i.e., faster than inter-complex transfer that takes 3–5 ps (Hess et al., 1995; Strümpfer and Schulten, 2012b). Accordingly, the assumption underlying generalized Förster theory, namely that transfer occurs from a thermally relaxed distribution of exciton states, holds in good approximation. The final step in stage I of energy conversion is the formation of quinol from quinone at the RC. The respective formation rate can be expressed (11) kQ→QH2⁢(I)=12⁢I⁢q⁢pRC⁢(I), where the prefactor 12 accounts for every quinol requiring two electron transfer events at the RC. Here q is the quantum yield given by Equation (8) and pRC⁢(I) is the probability for the RC to hold a quinone Q or a semiquinone QH, in either case the RC being ready to accept and convert an electronic excitation. The probability pRC⁢(I) decreases with increasing I, since the quinone/quinol pool becomes quinol rich/quinone poor at increasing light intensities, due in part to coupling (Figure 5) to chromatophore redox factors, succinate dehydrogenase, NADPH dehydrogenase, and cytochrome c oxidase (Klamt et al., 2008). As the quinone/quinol ratio decreases, it becomes less likely for RC to have a quinone/semiquinone available for electron transfer. The stated change in the quinone/quinol pool is crucial for energy conversion control of the chromatophore and comes about through the proton motive force, generated through light harvesting, inducing in the redox factors redox generation of products along with quinone/quinol conversion. The light-condition dependency of the quinone/quinol pool is described in the present model heuristically as explained below in Equation (15–19). Under the assumed steady state conditions, the rate kQ→QH2⁢(I), i.e., the rate at which RCs release QH2 as given by Equation (11), is equal to the rate at which RCs bind fresh quinones. Accordingly holds (12) 12IqpRC(I)=[nRC(1−pRC(I))]/τRC(I), where nRC=2⁢nL is the number of RCs in the chromatophore (24 for the vesicle shown in Figure 1), 1-pRC⁢(I) is the fraction of RCs ready to bind a fresh Q, τRC⁢(I) is the mean time needed for a RC to become available for binding a new Q after it had just accepted a Q (Remy and Gerwert, 2003). Below, we refer to τRC⁢(I) as the cycling time. The probability pRC⁢(I) is assumed, for convenience, to be uniform across all RCs rather than to vary between RCs due to inhomogeneities in the redox state of the quinone/quinol pool. This assumption is strictly valid only when the mixing time of quinols and quinones in the vesicle lipid phase is shorter than the time scales associated with the rates in Equation (1). The spatial inhomogeneity of pRC⁢(I) can be determined only through the simulation of the diffusive processes in the chromatophore, which is currently prohibitive. It had been suggested in (Geyer and Helms, 2006) that the primary rate-limiting step in the chromatophore is quinol turnover at cytb⁢c1 rather than cytochrome c2 diffusion; in (Geyer and Helms, 2006) it had been estimated that each cytochrome c2 is capable of approximately 80 electron transfers per second and that three cytochrome c2’s per vesicle are sufficient to saturate the turnover capacity of an ATP synthase. A chromatophore vesicle is expected to feature 10–20 cytochrome c2 molecules (Geyer and Helms, 2006; Cartron et al., 2014), safely exceeding the necessary number needed for saturation. Therefore, cytochrome c2 kinetics should not be rate limiting for energy conversion in the chromatophore. Using Equation (12), pRC⁢(I) can be expressed in terms of τRC⁢(I), namely, (13) pRC⁢(I)=(1+12⁢I⁢q⁢τRC⁢(I)⁢1nRC)-1. According to Equations (11) and (13), τRC⁢(I) needs to be determined in order to compute the rate kQ→QH2⁢(I) or, equivalently, kATP⁢(I). Stage II: Diffusion of charge carriers and estimate of cycling time τRC⁢(I) The cycling time, τRC⁢(I), arising in Equations (12,13), depends on light intensity. The cycling time is related to the quinone/quinol stoichiometry, i.e., the redox state of the quinone/quinol pool: the fewer quinones are present, the longer is the cycling time. The redox state is affected by not only RC and cytb⁢c1 reactions, but also by transmembrane enzymes succinate dehydrogenase and NADH dehydrogenase (Figure 5). The low-light and high-light limits for the cycling time, τRC, employed below are based on experimental observation (Woronowicz et al., 2011b, 2011a; Crofts, 2004) instead of direct computation; the reported values of τRC implicitly combine the redox reactions of all enzymes interacting with the quinone/quinol pool, including NADH dehydrogenase. In a stationary state, the rate kATP⁢(I) of ATP synthesis is equal to the rate kQ→QH2⁢(I) as stated in Equation (1), which according to Equations (11) and (13) can be expressed through the cycling time, τRC⁢(I). The condition of equilibrium assumed here might not be valid for rapidly fluctuating light intensities, where spatial inhomogeneities of the vesicle and the quinone/quinol pool are expected to play a nontrivial role on the cycling time. The low-light limit, τL, of the cycling time, τRC⁢(I), is observed to range from 0.7 ms for the membrane of an LH2-minus mutant to about 3 ms for the LH2-rich chromatophores adapted to low-light growth conditions (Woronowicz et al., 2011b, 2011a). In the following, we assume τL=3ms for the low-light growth vesicle shown in Figure 1. At the high-light limit, the immediate vicinity of a RC contains mostly quinols and the replacement of the converted quinones at the RC becomes rate limited by the turnover at cytb⁢c1(Woronowicz et al., 2011b, 2011a). The high-light limit, τH, of the cycling time, τR⁢C⁢(I), can be estimated by considering the total turn-over rate at all RCs, namely nRC⁢τH-1. In the stationary high I regime, this rate must be equal to the quinol turnover rate at all cytb⁢c1, namely nB⁢τB-1, i.e., it holds (14) nB⁢τB-1=nRC⁢τH-1, where nB is the number of cytb⁢c1 dimers (4 for the vesicle shown in Figure 1) and τB=25 ms is the quinol turnover time at a cytb⁢c1 (Crofts, 2004). Rate limitation of energy conversion by cytb⁢c1 The estimate of the cycling time, τR⁢C⁢(I), given below is based on the observation that energy conversion in the chromatophore is rate limited by quinol turnover at cytb⁢c1 (Lavergne et al., 2009; Geyer et al., 2010). This rate limitation follows directly from a comparison of turnover capacities, i.e., maximal turnover rates, at each key protein. The turnover rates of proteins are, in general, a function of the chromatophore conditions such as light intensity and redox states. The rate limiting components of the chromatophore can be identified by comparing the maximal values of the turnover rates, i.e., the turnover capacities, at each key protein, namely the quinol turnover capacity at cytb⁢c1, the quinol generation capacity at RC and the proton utilization capacity at ATP synthase. As an illustration of the cytb⁢c1-limited kinetics, we first compare quinol turnover capacities at the cytb⁢c1 and the RC. At a light intensity equivalent to 5% of full sunlight, i.e., 50 W/m2, a chromatophore vesicle absorbs I=1860 photons/s (estimated from the functional absortion cross-section of a chromatophore given in [Woronowicz et al., 2011a]), corresponding to a quinol turnover capacity at the RCs of 12⁢I⁢q=846 s-1. In contrast, the quinol turnover capacity at all cytb⁢c1s, given by Equation (14), is equal to nBτB−1=160s−1. Hence, already at 5% of full sunlight the quinol production capacity at the RCs exceeds the total quinol turnover capacity at cytb⁢c1 by more than five-fold. Consequently, under steady-state conditions the quinol production at the RCs at this illumination becomes limited by quinol turnover at cytb⁢c1. The onset of saturation of the energy conversion rate arising from rate limititation due to cytb⁢c1 at low light intensities is evident in Figure 2A. Next, we compare the proton turnover capacities at the cytb⁢c1 and the ATP synthase. The maximal proton turnover capacity at cytb⁢c1 for the vesicle shown in Figure 1 is 4×nB⁢τB-1=640 s-1. In comparison, total proton utilization capacity of ATP synthases is 4×2×270=2160 s-1, estimated based on the reported ATP synthase turnover capacity of 270 ATP molecules/s (Etzold et al., 1997) with four protons utilized per ATP and 2 ATP synthases present in the vesicle shown in Figure 1. Thus, the proton utilization capacity at ATP synthases exceeds the proton turnover capacity at cytb⁢c1s by more than threefold. In summary, of the three potential kinetic bottlenecks in the chromatophore, cytb⁢c1, RC, and ATP synthase, the lowest total turnover capacity is displayed by cytb⁢c1. Light intensity dependence of the cycling time at RC The I-dependence of the cycling time, τR⁢C⁢(I), needed to evaluate Equation (13), is approximated in terms of the relative populations of a two state system, the two states corresponding to the low-light and high-light limits (15) τR⁢C⁢(I)=cL⁢(I)⁢τL+cH⁢(I)⁢τH, where cL⁢(I) and cH⁢(I) are the probabilities that quinol turnover follows the low-light (limited to RC vicinity) or high-light (cytb⁢c1-limited) kinetics, respectively; it holds (16) cL⁢(I)+cH⁢(I)=1. The high-light limit of the cycling time, τH, can be expressed using Equation (14), (17) τH=nRCnB⁢τB. Since low-light and high-light limits are actually the extremes of a gradual behavior, the assumption of a two-state system appears rather drastic. However, at low light levels τL is diffusion controlled and amounts to the first passage time of the quinone to the RC, while τH is determined by cytb⁢c1 turnover, not diffusion. As a result one expects a distinct transition between τL and τH at some light intensity I corresponding to the saturation of the rate-limiting process in the chromatophore. From Equation (15,16) follows (18) τR⁢C⁢(I)=τL+(τH-τL)⁢(1-cL⁢(I)). The description of the cycling time according to Equation (18) is heuristic only. Future studies need to account for the time-dependent spatial inhomegeneity of the quinone/quinol pool by explicitly modeling the diffusion processes and redox states in the chromatophore. The population of photosynthetic states with respect to light intensity is typically governed by a Poisson distribution in terms of the utilization rate of excitations (Mauzerall, 1986; Peterson et al., 1987). In order to express the relative population of the low-light state, cL⁢(I), we observe that charge separated states are created at the RCs with the rate I⁢q and that the characteristic time for electron turnover at all cytb⁢c1s is given by 1/(2×nB⁢τB-1). The probability that no charge separation events occur during this time, i.e., the probability that the system remains in the low-light state, is given by the zero-event Poisson distribution, employed typically in describing the light intensity-dependence of photoproduct yield (Mauzerall, 1986; Peterson et al., 1987). According to this description holds, cL⁢(I)=exp⁡(-12⁢I⁢q/B), where B=2×nB⁢τB-1 is the total turnover capacity of cytb⁢c1s, which along with Equation (18) permits an estimate of the cycling time τR⁢C⁢(I), namely, (19) τR⁢C⁢(I)=τL+(τH-τL)⁢(1-e12⁢I⁢q⁢B-1). Equation (19), when substituted into Equations (11) and (13), permits an estimate of the quinol turnover rate kQ→QH2⁢(I), employed below for the computation of the ATP synthesis rate. Stage III: ATP synthesis As the last step of energy conversion, the proton gradient, generated at cytb⁢c1 through quinol → quinone conversion, is utilized by ATP synthase for the production of ATP. The ATP turnover rate of the vesicle, kATP⁢(I), under stationary conditions is equal to kQ→QH2⁢(I) given by Equation (11). This equality is based on the assumption that ATP synthase of Rba. sphaeroides has an Fo of 12 c-subunits such that four H+ conducted through the Fo-ring of ATP synthase lead to a 120° rotation of the stalk in the F1 part and, thereby, to synthesis of one ATP. Currently, the structure of the ATP synthase of Rba. sphaeroides and the corresponding number of c-subunits is not known experimentally. If the Fo oligomer were to feature, e.g., 11 or 10 subunits instead, this structural detail would proportionally affect the number of protons required for the rotation of the stalk and subsequent synthesis of each ATP and, therefore, directly influence the estimated energy conversion efficiency of the chromatophore, by 11% or 20%, respectively. Combining Equation (1),(11),(13), the ATP turnover rate can be expressed (20) kATP⁢(I)=12⁢I⁢q⁢(1+12⁢I⁢q⁢τRC⁢(I)⁢1nRC)-1, where the cycling time at the RC, τRC⁢(I), is given by Equation (19). The overall energy conversion efficiency of the chromatophore, ηATP⁢(I), can be defined as the ratio of formation rate of energy in the form of ADP→ATP synthesis to the total absorption rate of photon energy (21) ηATP⁢(I)=EATP⁢kATP⁢(I)Eγ⁢I, where EATP=4197⁢cm-1 is the ATP hydrolysis energy in the cell (Berg et al., 2011) and Eγ is the average energy of an absorbed photon, taken to be the photon energy at 850 nm (11765 cm-1). Not all the energy of an absorbed photon, Eγ, is available for energy harvesting. The fraction of Eγ available for conversion into chemical energy is the so-called Carnot yield (Lavergne, 2009) described by comparing photochemical energy conversion to the function of a heat engine. This limitation in photochemical energy conversion establishes a theoretical upper limit for photosynthetic energy conversion at broad daylight of approximately 0.7 (Lavergne, 2009). The determination of the energy conversion efficiency, ηATP⁢(I), computed through Equation (21), has the shortcoming that the ATP hydrolysis energy, EATP, depends, in principle, on the ADP, ATP, and H+ concentrations in the cytoplasm, which are not modeled explicitly. Nevertheless, Equation (21) permits a comparison with similar measures of efficiency reported for other photosynthetic or photovoltaic systems (Blankenship et al., 2011). Supplementary material Computation of quantum yield and table of BChl properties In the following, we describe the computation of the quantum yield, q, given by Equation (8), in section 2.2 of the main text. The quantum yield is central to the overall efficiency of the chromatophore as it accounts for the efficiency of the primary subsystem, the light harvesting apparatus. The constants employed in the computation of the quantum yield are listed below in Table 1. We adopt the same values as in similar computations reported by us in (Sener et al., 2010, 2007). For the sake of clarity, this section repeats some information in the main text, as indicated.10.7554/eLife.09541.009Table 1. Constants employed in the computation of the quantum yield. DOI: http://dx.doi.org/10.7554/eLife.09541.009 Symbol Value* Description ϵ1LH2ϵ2LH2 12,459 cm−1 12,625 cm−1 BChl site energies for alternating LH2 B850 BChls, used in Equation (S2) ϵLH1 12,344 cm−1 BChl site energies for LH1 B875 BChls, used in Equation (S2) ϵ1RCϵ2RC 12,092 cm−1 12,581 cm−1 BChl site energies for RC special pair and accessory BChls, used in Equation (S2) V1LH2 363 cm−1 nearest neighbor BChl-BChl coupling for LH2 B850 BChls within the same αβ dimer, used for VijI values in Equation (S3) V2LH2 320 cm−1 nearest neighbor BChl-BChl coupling for LH2 B850 BChls across neighboring αβ dimers, used for VijI values in Equation (S3) V1LH1 806 cm−1 nearest neighbor BChl-BChl coupling for LH1 B875 BChls within the same αβ dimer, used for VijI values in Equation (S3) V2LH1 377 cm−1 nearest neighbor BChl-BChl coupling for LH1 B875 BChls across neighboring αβ dimers, used for VijI values in Equation (S3) VRC 500 cm−1 Coupling between RC special pair BChls, used in Equation (S3) C 348,000 Å3 cm−1 coupling constant for transition dipole interactions between non-nearest neighbor LH2 BChls, used in Equation (S4) σLH1 235 cm−1 linewidth of LH1 exciton states, assumed uniform, used for σA in Equation (S8) σL⁢H⁢2 188 cm−1 linewidth of LH2 exciton states, assumed uniform, used for σA in Equation (S8) kLH1,RC (35 ps)−1 excitation transfer rate from LH1 B875 BChls to RC, used for corresponding kIJ values in in Equation (S10) kRC,LH1 (8 ps)−1 excitation transfer rate from RC to LH1 B875 BChls, used for corresponding kIJ values in in Equation (S10) kdiss (1 ns)−1 excitation decay rate due to internal conversion, used in Equation (S10) kCS (3 ps)−1 charge separation rate at reaction center, used in Equation (S10) *: energy units given in wavenumbers (1 eV = 8066 cm−1). In order to describe the electronic excitation transfer in the chromatophore and determine the quantum yield, q, the matrix of transfer rates, kIJ, between BChl clusters, I,J, needs to be constructed, which in turn depends on the effective Hamiltonian, HI, for each BChl cluster, I. In the following, the indices, I,J, label the BChl clusters listed in the previous section: for the vesicle shown in Figure 1, based on (Cartron et al., 2014), there are 63 LH2 B850 BChl clusters, 24 LH1 B875 BChl clusters, and 24 RC BChl clusters. As mentioned in the previous section, LH2 B800 BChls do not form excitonically coupled states, transferring excitation energy, after light absorption, immediately to LH2 B850 BChls. The effective Hamiltonian HI of each BChl cluster I is given, according to (Strümpfer et al., 2012), by Equation (2) in the main text, namely (S1) HI=∑i=1NIEiI|i⟩⟨i|+∑i>j>0NIVijI(|i⟩⟨j|+|j⟩⟨i|), where NI is the number of BChls in cluster I. The site energies EiI in Equation (S1) are (S2) EiI={ϵ1,2LH2,I:LH2B850,ϵLH1,I:LH1B875,ϵ1,2RC,I:RC, with energy constants as listed in Table 1 for corresponding BChl groups, chosen to reproduce corresponding absorption peaks. The couplings Vi⁢jI in Equation (S1) are determined through the point-dipole approximation as described below with the exception of nearest neighbor couplings of the LH2 B850 and LH1 B875 BChl clusters and the RC special pair coupling, (S3) ViI={V1,2LH2,I:LH2B850,nearest  neighbor,V1,2LH1,I:LH1B875,nearest  neighbor,VRC,I:RC,special  pair, which are instead taken from (Damjanović et al., 2000; Şener and Schulten, 2002) following quantum chemistry computations reported in (Tretiak et al., 2000) and listed in Table 1. For non-nearest neighbor BChls, the coupling Vi⁢jI in Equation (S1) is computed according to the point-dipole approximation (Ritz et al., 2001; Sener et al., 2011) (Equation (3) in the main text) (S4) Vi⁢j=C⁢(𝐝^i⋅𝐝^jri⁢j3-3⁢(𝐝^i⋅𝐫i⁢j)⁢(𝐝^j⋅𝐫i⁢j)ri⁢j5), where 𝐝^i is the transition dipole moment unit vector of pigment i, 𝐫i⁢j is the vector joining pigments i and j, in BChl cluster I; the coupling constant C (Şener and Schulten, 2002) is listed in Table 1. The transition dipole moment unit vector 𝐝^k for BChl k is determined from the coordinates listed in the previous section according to (S5) 𝐝^k=(𝐫kD-𝐫kB)|𝐫kD-𝐫kB|, where rkB and rkD are the positions of atoms NB and ND of BChl k; the transition dipole moment is centered at the position, rkM, of the MG atom of BChl k, as labeled in the PDB files. Based on the effective Hamiltonians, Equation (S1), thus constructed, the rate of excitation transfer between a donor complex I and an acceptor complex J, kI⁢J, can be calculated according to the so-called modified Förster theory (Ritz et al., 2001; Şener et al., 2007, 2011) given by Equation (4) in the main text, namely through (S6) kIJ=2πℏ∑μ∈I∑ν∈JpμI|(μ|HIJ|ν)|2Jμν, where (HI⁢J)i⁢j is the matrix of interactions between the excited states of pigments i and j in complexes I and J, computed according to Equation (S4) and (μ⁢|HI⁢J|⁢ν) are the couplings (HI⁢J)i⁢j in the basis of the eigenstates |μ) and |ν) of Hamiltonians HI and HJ, respectively; pμI=e-β⁢ϵμ/∑γ∈Ie-β⁢ϵγ are Boltzman weights for the eigenstates HI|μ)=ϵμ|μ), where ϵμ are the exciton energies defined as the eigenvalues of the Hamiltonian, HI, given in Equation (S1). The overlap integrals Jμ⁢ν are described according to Equation (5) in the main text) (S7) Jμν=∫dESμD(E)SνA(E). Jμ⁢ν is the spectral overlap between donor exciton state |μ) and acceptor exciton state |ν). The donor and acceptor lineshapes, SμD⁢(E) and SνA⁢(E), used in the calculation of Jμ⁢ν, are approximated by normalized Gaussians (S8) SμA(E)=12πσAexp⁡[−(E−ϵμσA)2], (S9) SμD(E)=SμA(E−S), σA is the linewidth of excitons assumed to be uniform across all states (Jimenez et al., 1997) and is given in Table 1 as σL⁢H⁢1 and σL⁢H⁢2; S denotes the spectral shift between donor and acceptor spectra (Small, 1995; Damjanović et al., 2000). The transfer rates, kI⁢J, in Equation (S6) are negligible for any non-neighboring BChl clusters, I,J, and are taken to be zero in those cases. The transfer rates between the B875 and the RC BChls of a LH1-RC complex needs to be determined only once, since the relative pigment geometry is identical within each LH1-RC complex. Accordingly, for the LH1→RC and RC→LH1 transfer rates, the values (35 ps)-1 and (8 ps) -1 are assumed, respectively (Sener et al., 2010, 2009), in Table 1; the LH1→RC value was chosen in (Sener et al., 2010) to match the observed excitation lifetime in the RC-LH1 complex of 50 ps; the value of the LH1→RC transition rate, computed according to Equation (S6), was reported to result in an overestimate of the excitation lifetime (Sener et al., 2009). The remaining transfer rates, kI⁢J, namely between neighboring LH2 B850 and LH1 B875 BChl clusters are determined according to Equation (S6). Exciton migration in the chromatophore is governed by the rate matrix for inter-complex exciton transfers, 𝒦I⁢J (Sener et al., 2010, 2007) Equation (7) in the main text) (S10) (𝒦)IJ=kJI−δIJ(∑MkIM+kdiss+kCSδI,RC), where the dissipation rate kdiss and the charge separation rate kCS are listed in Table 1. The quantum yield, q, is finally expressed in terms of 𝒦I⁢J (Şener et al., 2007, 2011) according to Equation (8) in the main text, namely through (S11) q=−kCS(1RC)T⋅𝒦−1⋅P(0), where (𝐏⁢(0))I=NI/(∑JNJ), corresponds to a uniform probablity for initial excitation. The quantum yield is not strongly dependent on the choice of the initial state (𝐏⁢(0)) (Şener et al., 2007). The quantum yield, q, is computed by substituting the BChl atom coordinates, as listed in the caption of Supplementary file 1, into Equations (S4 and S5) to determine the couplings, Vi⁢j, and subsequently the matrices kI⁢J and 𝒦I⁢J according to Equations (S6 and S10), substituted finally into Equation (S11). The quantum yield of the chromatophore shown in Figure 1, thus computed, is 0.91. Funding Information This paper was supported by the following grants: http://dx.doi.org/10.13039/501100000268Biotechnology and Biological Sciences Research Council BB/M000265/1 to C Neil Hunter. http://dx.doi.org/10.13039/100000015U.S. Department of Energy DE-SC0001035 to C Neil Hunter, Klaus Schulten. http://dx.doi.org/10.13039/501100000781European Research Council 338895 to C Neil Hunter. http://dx.doi.org/10.13039/100000001National Science Foundation PHY0822613 to Klaus Schulten. http://dx.doi.org/10.13039/100000002National Institutes of Health NIH 9P41GM104601 to Klaus Schulten. http://dx.doi.org/10.13039/100000001National Science Foundation MCB1157615 to Klaus Schulten. Acknowledgements The reviewers are thanked for extensive suggestions to improve the manuscript, particularly regarding comments on processes in the chromatophore beyond the light harvesting-cytb⁢c1-ATP synthase components. The authors would like to also thank Antony Crofts, Robert Niederman, and Donald Bryant for insightful discussions on chromatophore function. This work was supported by the National Science Foundation (MCB1157615 and PHY0822613) (to KS), the National Institutes of Health (NIH 9P41GM104601) (to KS). CNH acknowledges research grant BB/M000265/1 from the Biotechnology and Biological Sciences Research Council (UK). CNH was also supported by an Advanced Award 338895 from the European Research Council. This research used resources of the Oak Ridge Leadership Computing Facility at the Oak Ridge National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC05-00OR22725. The study reported was funded also by the Photosynthetic Antenna Research Center (PARC), an Energy Frontier Research Center supported by the US Department of Energy, Office of Science, and Office of Basic Energy Sciences under Award Number DE-SC0001035 (to CNH and KS). The molecular image in Figure 1 was generated with VMD (Humphrey et al., 1996). Additional information Competing interests The authors declare that no competing interests exist. Author contributions MS, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. JS, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. AS, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. CNH, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. KS, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. Additional files 10.7554/eLife.09541.010Supplementary file 1. BChl groups of the chromatophore and the corresponding transition dipole moments needed to determine the effective Hamiltonian: LH2 complexes. (file: Vesicle30BCLdipoleLH2.pdb) Here we provide the transition dipole moments of all 2577 chromatophore BChls defining the effective Hamiltonian in Equations (2,3) of the main text, needed for evaluation of the excitation transfer rates through Equation (4). The reader is referred to (Cartron et al., 2014; Şener et al., 2007, 2010) for the construction of the underlying chromatophore structural model.The three PDB files listed here contain the coordinates for the MG, NB, ND atoms of the BChls in the chromatophore model (Cartron et al., 2014) that define the transition dipole moment unit vectors as discussed in the text and given according to Equation (S1) below. These coordinate files correspond to the following protein complexes: LH2 : [Vesicle30BCLdipoleLH2.pdb], dimeric RC-LH1 : [Vesicle30BCLdipoleLH1RCdimer.pdb], monomeric RC-LH1 : [Vesicle30BCLdipoleLH1RCdimer.pdb], where coordinates of multiple complexes of the same type are concatenated into one file for each complex type.The BChls of the chromatophore can be divided into the following groups: (i) B800 BChls of LH2 : labeled as [resid=307] in the file [Vesicle30BCLdipoleLH2.pdb] (resid is used here as an abbreviation for residue sequence number in the PDB file format); named after the absorption peak of 800 nm; (ii) B850 BChls of LH2 : labeled as [resid=301, 302] in the file [Vesicle30BCLdipoleLH2.pdb]; named after the absorption peak of 850 nm; (iii) B875 BChls of LH1-RC (both monomer and dimer) : labeled as [resid=100-155] in the files [Vesicle30BCLdipoleLH1RCdimer.pdb] and [Vesicle30BCLdipoleLH1RCmonomer.pdb]; named after the absorption peak of 875 nm; (iv) reaction center (RC) BChls of LH1-RC (both monomer and dimer) : labeled as [resid=301, 302, 303, 304] in the files [Vesicle30BCLdipoleLH1RCdimer.pdb] and [Vesicle30BCLdipoleLH1RCmonomer.pdb]; the so-called special pair BChls where electron transport is initiated are labeled by [resid = 302, 303]. Of the aforementioned BChl clusters, the three, namely, LH2 B850, LH1 B875, and RC BChls, form strongly coupled excitonic states; the LH2 B800 BChls do not share excitation energy between themselves, transfering it rapidly (within 0.5 ps) (Şener et al., 2007; Ritz et al., 2001) to the B850 ring of the same LH2. The theory of excitation transfer between the BChl clusters listed above is described in Supplementary materials. DOI: http://dx.doi.org/10.7554/eLife.09541.010 10.7554/eLife.09541.011Supplementary file 2. BChl groups of the chromatophore and the corresponding transition dipole moments needed to determine the effective Hamiltonian: RC-LH1 dimer complexes. (file: Vesicle30BCLdipoleLH1RCdimer.pdb) As explained for Supplementary file 1, but for BChls belonging to RC-LH1 dimers. DOI: http://dx.doi.org/10.7554/eLife.09541.011 10.7554/eLife.09541.012Supplementary file 3. BChl groups of the chromatophore and the corresponding transition dipole moments needed to determine the effective Hamiltonian: RC-LH1 monomer complexes. (file: Vesicle30BCLdipoleLH1RCmonomer.pdb) As explained for Supplementary file 1, but for BChls belonging to RC-LH1 monomers. DOI: http://dx.doi.org/10.7554/eLife.09541.012 10.7554/eLife.09541.013Decision letter Hummer Gerhard Reviewing editorThe Max Planck Institute of Biophysics, GermanyIn the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included. Thank you for submitting your work entitled "Overall energy conversion efficiency of a photosynthetic vesicle" for peer review at eLife. Your submission has been favorably evaluated by Michael Marletta (Senior editor), a Reviewing editor, and three reviewers. We apologize for the delay in rendering a decision, but it was necessary to find multiple independent reviews on this manuscript. The reviewers have discussed the reviews with one another and the Reviewing editor has drafted this decision to help you prepare a revised submission. Summary: The authors present a model of the Rhodobacter sphaeroides chromatophore. The model employs a detailed quantum mechanical description for the light harvesting and charge separation processes, and a kinetic description for the subsequent reactions. With this model, the solar energy capture efficiency is determined, as measured in terms of ATP production. Efficiencies are estimated to be on the order of 10%. The authors conclude that the vesicle stoichiometry is sub-optimal for steady-state ATP production and instead appears to be adapted to "robustness of function". Essential revisions: All three reviewers are overall positive, but raise a number of issues concerning the model, its limitations, and its interpretation. I find the issues raised by the referees to be relevant. Addressing them in a revision should result in a stronger paper. Following is a brief summary of their reports that emphasizes the main points as seen from my perspective. Reviewer #1 challenges the conclusion that suboptimality of ATP production is associated with robustness of function. First, the reviewer points out that the build-up of a QH2 pool would only work if an equal amount of cytochrome c2 were present (in essence because the electron cycle with the chromatophore is closed). Since the number of cyt c is much smaller, something seems to be amiss, either in the model or its interpretation. Second, the reviewer argues that evolution acts on the survival of the organism, not at the level of its chromatophores. Reviewer #1 is also concerned about the absence of leak reactions in the model, in particular for electrons between quinol and the Rieske protein, and for protons through the chromatophore membrane. Reviewer #2 is concerned that the model lacks physiological relevance, despite overall being positive and recognizing the value of a reduced model. In particular, the Reviewer notes that chromatophore operation is connected to NADH production. If this link were included, the reviewer argues that the buildup of the ubiquinol pool would not occur. At the least, the authors should discuss the more complex situation of chromatophore function in phototropic bacteria. Concerning the model it is assumed that the Fo part of the ATP synthase has a c-ring with 12 subunits, but to the best of knowledge, this is only a guess. If so, it should be stated as such. Additional major points are the precise definitions of efficiency metrics (Reviewer #1) and a discussion as to what the atomistic description of the chromatophore really adds. Moreover, all reviewers point to the need for a more thorough discussion of the relevant literature. After the reviews were returned, the editors discussed your work in relation to the broader audience of eLife. We urge you to include language in the Discussion section that expands the scope of your presentation to include a description of how your finding may relate to other systems. In light of the reviewers comments, changes will be necessary before the paper can be considered for publication. Following are their full reports. These should be consulted and addressed point by point. Reviewer #1: In this work the authors calculate the solar energy capture efficiency of a bacterial photosynthetic vesicle based on an explicit model for the structure of the photosynthetic apparatus. While a detailed quantum mechanical model is used for the light harvesting and charge separation process, for all the other subsequent reactions kinetic models are utilized. At solar intensities of 1-5% of full sunlight the energy conversion efficiency is calculated to range from 0.12 to 0.04. The vesicle stoichiometry appears to be sub-optimal for steady-state ATP production and appears to be adapted to "robustness of function". In general this is a well-written paper, however there are some serious problems in the interpretation that the authors must account for. 1) The authors suggest that the suboptimality of the ATP-production of the Rba. Sphaeroides vesicle might play a role in the robustness of its function. According to their interpretation reduction of the Q-pool might play a role in storing the redox energy accumulated during short periods of exposure to high light. This would only work if there is an equal amount of cytochrome c present. However, there is one to two orders of magnitude more quinone than cytochrome c so where do the electrons come from that reduce the Q-pool? 2) Evolution did not act at the level of individual chromatophores but on the survival fitness of the intact organism. 3) The authors use efficiency values without specifying exactly what they mean. I presume that they have had a certain numerical value for "delta-Gp" in mind (i.e. the standard free energy of hydrolysis of ATP), and then calculate efficiency as the ratio of the caloric value of the photons over the free energy of the corresponding ATP hydrolysis. This should however be made explicit. Furthermore, they now and then shift back and forth from oxygenic to anoxygenic photosynthesis (e.g. Discussion, first paragraph), which is not proper because in oxygenic photosynthesis the majority of the energy conservation goes into redox energy (i.e. NADPH). Estimates that anoxygenic photosynthesis has a free-energy efficiency in the order of 10% have already been available for a long time (Hellingwerf, K. J., Crielaard, W., & Westerhoff, H. V. (1993). Comparison of retinal-based and chlorophyll-based photosynthesis: A biothermokinetic description of photochemical reaction centers. Modern trends in biothermokinetics (pp. 45-52) Springer). 4) Beyond the LH/RC itself, nowhere in the modelling has any leak-reaction taken into account. Two examples that would have been very relevant to take into account are: (a) the short-cut in electron transfer in the cytochrome bc1 complex that may occur if the second electron of quinol is also directly transferred to the Rieske protein and (b) passive proton leakage through the chromatophore membrane (which may be very much dependent of the protein composition of the vesicles, due to curvature effects). Related to this is the fact that the authors seem to suggest that proton transport merely leads to a pH change (see discussion on acidification), whereas presumably the energetic coupling between the proton pumping system and the ATP synthase is through a transmembrane electrical potential gradient. 5) What precisely does the atomistic description of the light harvesting and trapping process contribute to the calculation of the efficiency? Could the energy transfer and trapping process not be captured by a single effective rate constant that is a function of the fraction of open and closed traps and the ratio of LH2 to RC-LH1. See many earlier papers on the subject: Vredenberg and Duysens, Duysens CIBA foundation, Den Hollander et al., Trissl et al., Joliot cs etc. Reviewer #2: The manuscript by Sener and co-workers presents a model of a Rhodobacter sphaeroides chromatophore. The model describes how the illumination of a chromatophore suspension results in ATP production. The ultimate goal of the modeling is to understand how the ATP yield is coupled with the intensity of light. The main virtue of the model is that it couples the quantum physics of light absorption/procession with the chemistry of ATP synthesis. The model gives reasonable values of ATP yield in response to low light, which is physiological for these bacteria. Development of such a global model, which covers events over a time span of 14 orders of magnitude, required certain simplifications, which is understandable. Unfortunately, from the opinion of this reviewer, the simplifications preclude from making physiologically relevant conclusions from the modeling, at least, at the given stage. Still the model itself is a great achievement and the manuscript deserves publication albeit after certain amendments. 1) Since the physiological relevance of the results obtained is questionable, the main result of this study seems to be the model itself. Therefore, the description of the model should be moved from the Materials and methods into the Results. Otherwise the manuscript is incomprehensive, taking into account that the format of eLife implies Materials and methods after Results. The Materials and methods section should then contain only technical/purely methodical points; supplementary materials, at least partly, could be also moved into Materials and methods. 2) The current simplified model lacks two important modules, which, hopefully, would be added in the future. These are the protonmotive force and reversible NADH dehydrogenase. The authors admit the absence of these modules, but did not consider the consequences of their absence in full extent. This should be done in the Discussion. 3) In the absence of these two modules, the authors conclude that "energy conversion in the chromatophore appears to be rate-limited primarily by cytbc1" and that "cytbc1-limited kinetics prevents an overproduction of protons at sustained high-light conditions,…thereby protecting the chromatophore against overacidification of its interior and assuring vesicle integrity at high-light conditions". In real phototropic bacteria, the situation is more complex. Obviously, the function of a photosynthetic apparatus, such as bacterial chromatophore or a chloroplast thylakoid, is to produce ATP and NAD(P)H. In thylakoids, NAD(P)H molecules are produced via direct reduction of NAD(P)+ by the Photosystem I. In Rb. sphaerides, the redox potential of the RC is too high for NAD(P)H generation. Therefore, NAD(P)H is produced via reversing the NAD(P)H dehydrogenase. The reversion requires (1) high protonmotive force and (2) reduced ubiquinol pool. Therefore, the overreduction of the ubiquinol pool in Rb. sphaeroides is a precondition for the NAD(P)H production; it should increase the overall efficiency of the system. Furthermore, the overacidification of the interior is prevented not by the overreduction of the ubiquinone pool, but by the high sensitivity of the cytbc1to the membrane potential. High voltage, needed to drive the NAD(P)+ reduction (see above), slows the turnover of the cytbc1dramatically, essentially blocking its turnover at voltages above 150 mV. Hence at high, physiological values of protonmotive force the turnover of the cytbc1is essentially limited by the proton discharge, specifically via the ATP synthase and the reverse NAD(P)H dehydrogenase. The operation of the cytbc1close to equilibrium additionally increases the efficiency of energy conversion. The reviewer admits that the data on synthesis of NAD(P)H in Rb. sphaeroides are too scarce to be used for verification of the model. The authors, accordingly, have tested their model by using more abundant data that were obtained with chromatophore suspensions in the absence of NAD(P)+, in experiments where the function of the RC, cytbc1and/or ATP synthase were studied. And still, the discussion of physiological consequences from the given model should be truncated dramatically as misleading. The physiology could be discussed in further publications, after the membrane potential and NAD(P)H synthesis would be explicitly included in the model. Instead, the Discussion section should include comparison of the given model with the earlier models of chromatophore systems (e.g. from the group of Helms) and the approaches used. Reviewer #3: The authors present an important theoretical investigation of the performance and limitations of the photosynthetic chromatophore complex of purple bacteria. This light-harvesting system is of great interest as a model system that contains numerous characteristic properties inherent to all natural light-harvesting systems. This work is a continuation of a long line of first-rate papers published by the authors on this topic, is extremely well written, and merits publication in its current form. There appears to be a minor inconsistency in the units of intensity employed. In the second paragraph of section 4.2 the authors define light intensity, I, in units of [photons absorbed per second], whereas light intensity is often defined in units of [photons absorbed per meter squared per second]. As defined, this intensity is in fact a power, which is also called a 'radiant flux' in radiometry. This in contrast with the more conventional definition of intensity, in the caption of Figure 2, where units of [Watts per meter squared] are used. Clarification of these different definitions would be useful. In the third paragraph of section 4.2, the authors switch to defining 'I' as 'illumination,' which may further confuse the reader. Ideally 'radiant flux' or 'absorbed light power' or similar terminology would be employed throughout, or alternatively 'illumination' would be defined. The explanation provided in the fourth and fifth paragraphs of section 2.2 for the suboptimal (nB, nRC) values in low-light adapted vesicles is appealing, but it does not appear to be supported by calculations. Despite the fact that the model only applies in steady-state, is it possible to perform a simple simulation to support this interpretation, even qualitatively? The choice of initial condition assumed in Equation 9 may possibly be over-simplifying because each complex (LH2, LH1 and RC) does not necessarily have the same absorption spectrum, since at a given wavelength, the probability of absorption of a photon is not the same for each complex. In this scenario, the initial condition should then be wavelength dependent. While the authors state that the quantum yield is not strongly dependent on the choice of P(0), a specific example to demonstrate this would be helpful. [Editors' note: further revisions were requested prior to acceptance, as described below.] Thank you for resubmitting your work entitled "Overall energy conversion efficiency of a photosynthetic vesicle" for further consideration at eLife. Your revised article has been favorably evaluated by Michael Marletta (Senior editor), a Reviewing editor, and two reviewers. On the basis of the referee reports on the revised manuscript, only relatively minor additional changes are required. Reviewer #2 is concerned that some issues have not been fully addressed in the revision. In the following I list the main points. 1) The reviewer maintains that "the model does not consider the production of NADH at all." Therefore, "no statements on the energy conversion efficiency could be done with the current model." The reviewer is concerned that the issue of NADH production and its impact on the model and its assessment are not discussed adequately, and advises "to read the primary literature on the mechanism of NADH production and provide references to this literature (currently they refer only to the paper of Klamt et al. where bacterial photosynthesis was modeled by using several dozen parameters that were more or less freely variable). The primary literature describes how the NADH-dehydrogenase makes NADH by driving the reverse (uphill) electron transfer from the ubiquinone pool to NAD+ with the help of membrane potential. This mechanism differs from the NAD(P)H production in green plants where NAD(P)+ is directly reduced by photoexcitated electrons in a reaction that is independent of membrane potential. The dependence of the light-driven NADH production on membrane potential, specific for Rhodobacter and other purple phototrophic bacteria, should have shaped the properties of their energy-converting machinery." Recommended action: Whereas I do not see the lack of NADH production as a fundamental flaw in the model (after all, all models are incomplete), this point should be made clearer to the reader. Conclusions drawn from the model should be qualified accordingly. In particular, in the discussion of the energy conversion efficiency, relevant limitations of the model should be stated. 2) The reviewer also asks to point out clearly that "the experimental data that were used for testing the model were not obtained with chromatophores in vivo," but for "an artificial system – a suspension of chromatophores in a pH-buffer, in the presence of electron donors and redox mediators, but in the absence of physiological compounds such as NAD+/NADH (a Dutton/Crofts system)." Recommended action: Point out the distinction between the in-vitro experiments and the more complex in-vivo situation. 3) The reviewer also argues that the effect of a transmembrane proton gradient, which is not included explicitly in the model, should be discussed. The reviewer emphasizes that the issue is not "proton migration from the bc1 complex to ATP synthase" but protonmotive force. Recommended action: Discuss the issue of protonmotive force. Examine which conclusions would likely be impacted by the explicit treatment of PMF in the model, and qualify them accordingly. If none of the conclusions would be affected, state that the model is robust with respect to the neglect of explicit modeling of PMF. 4) The reviewer suggests adding a description of the model to the Results section. 10.7554/eLife.09541.014Author response Reviewer #1: In this work the authors calculate the solar energy capture efficiency of a bacterial photosynthetic vesicle based on an explicit model for the structure of the photosynthetic apparatus. While a detailed quantum mechanical model is used for the light harvesting and charge separation process, for all the other subsequent reactions kinetic models are utilized. At solar intensities of 1-5% of full sunlight the energy conversion efficiency is calculated to range from 0.12 to 0.04. The vesicle stoichiometry appears to be sub-optimal for steady-state ATP production and appears to be adapted to "robustness of function". In general this is a well-written paper, however there are some serious problems in the interpretation that the authors must account for. 1) The authors suggest that the suboptimality of the ATP-production of the Rba. Sphaeroides vesicle might play a role in the robustness of its function. According to their interpretation reduction of the Q-pool might play a role in storing the redox energy accumulated during short periods of exposure to high light. This would only work if there is an equal amount of cytochrome c present. However, there is one to two orders of magnitude more quinone than cytochrome c so where do the electrons come from that reduce the Q-pool? The energy buffering capacity of the Q-pool, arising between light absorption and ATP synthesis, is limited, as the reviewer states, by the number of electrons available in the photosynthetic apparatus that can shift the quinone-quinol balance. However, the quinone-quinol balance can be shifted also through redox reactions coupling the chromatophore to the redox pools in the cell's cytoplasm by means of membrane-bound enzymes listed below and in the text. The current structural model (as presented in Figure 1) does not include such enzymes and, therefore, cannot account for sufficient buffering capacity of the Q-pool. As the storage of energy in case of light fluctuations is not an essential outcome of our study, we follow the reviewer's suggestion and removed from the manuscript the interpretation regarding the buffering effect (last sentence of the Abstract; Introduction; Results; Discussion; Materials and methods. As the reviewer states, the redox buffering capacity of the Q-pool is dependent on the electrons available in the chromatophore for quinone-to-quinol reduction. When considering the chromatophore components shown in Figure 1, only a limited number of electrons are maximally available: 24 from RC, 20 from cyt. c2, and from the bc1 complex with already loaded quinols 16, altogether 60. These electrons can contribute to forming at most 30 quinols that become available at the bc1 complex for formation of the proton gradient. As the reviewer points out, it is not possible to explain the free energy buffer capacity of the chromatophore relevant for changes in light intensity from the present model without additional quinols. However, in an actual chromatophore, the mechanism with which the redox state of the Q-pool is regulated involves also exchange of electrons with the succinate/fumarate pool in the cell's cytoplasm via the enzyme succinate dehydrogenase; additionally NADH dehydrogenase and cytochrome c oxidase facilitate electron exchange reactions with the Q-pool. The contribution of these reactions to the redox regulation of the Q-pool is currently not well understood and the corresponding enzymes are not resolved in the current structural model of the chromatophore shown in Figure 1. Therefore, in the current description of chromatophore processes the redox state of the Q-pool is accounted for, according to Equations 14–18, in an heuristic way. For an explanation, an additional schematic figure (Figure 5) is added along with corresponding text (section 4.1, last paragraph) in the Materials and methods to introduce the missing enzymes. The role of these enzymes are discussed in: Introduction, fifth and sixth paragraphs; section 2.2, fifth paragraph; Discussion, fourth paragraph; section 4.1, last paragraph; Figure 5; subsection “Stage I: Light absorption, excitation energy transfer, and quinol formation”, eleventh paragraph; subsection “Stage II: Diffusion of charge carriers and estimate of cycling time τRC(I)”, fifth paragraph. This issue is further discussed in a note following response #1.4 below. 2) Evolution did not act at the level of individual chromatophores but on the survival fitness of the intact organism. Of course, we agree with the reviewer and apologize for the “loose” language used in our original manuscript. As the reviewer states, evolution acts on the reproductive fitness of the entire organism. Yet, increased survival fitness for an organism frequently results from increased efficiency of individual cellular components (c.f. citations in Introduction, third paragraph; section 2.2, last paragraph), even though the direct relation between component efficiency and reproductive fitness is usually difficult to establish. This distinction is clarified in the text at the aforementioned paragraphs (Introduction, third paragraph; section 2.2, first paragraph). 3) The authors use efficiency values without specifying exactly what they mean. I presume that they have had a certain numerical value for "delta-Gp" in mind (i.e. the standard free energy of hydrolysis of ATP), and then calculate efficiency as the ratio of the caloric value of the photons over the free energy of the corresponding ATP hydrolysis. This should however be made explicit. Furthermore, they now and then shift back and forth from oxygenic to anoxygenic photosynthesis (e.g. Discussion, first paragraph), which is not proper because in oxygenic photosynthesis the majority of the energy conservation goes into redox energy (i.e. NADPH). Estimates that anoxygenic photosynthesis has a free-energy efficiency in the order of 10% have already been available for a long time (Hellingwerf, K. J., Crielaard, W., & Westerhoff, H. V. (1993). Comparison of retinal-based and chlorophyll-based photosynthesis: A biothermokinetic description of photochemical reaction centers. Modern trends in biothermokinetics (pp. 45-52) Springer). The various definitions of energy conversion efficiency are compared and discussed now in: Materials and methods, first paragraph; subsection “Stage III: ATP synthesis”, second paragraph, particularly in the paragraph following the definition of overall conversion efficiency η _ATP in Equation (20). In following the reviewer's suggestion, we have removed the misleading comparison to plant photosynthesis from Discussion. We also cite now the work of (Hellingwerf, 1993) that has provided an early estimate for a maximal value for the conversion efficiency of photosynthesis in Rba. sphaeroides (section 2.1, third paragraph). The main point of our study is not only to establish the efficiency value, but (i) to demonstrate that the efficiency value can be achieved on account of the overall molecular structure of the chromatophore; (ii) to discuss the dependence of ATP production rate and efficiency on light conditions; (iii) to determine the degree of optimization for energy conversion as a function of stoichiometry. 4) Beyond the LH/RC itself, nowhere in the modelling has any leak-reaction taken into account. Two examples that would have been very relevant to take into account are: (a) the short-cut in electron transfer in the cytochrome b/c1 complex that may occur if the second electron of quinol is also directly transferred to the Rieske protein and (b) passive proton leakage through the chromatophore membrane (which may be very much dependent of the protein composition of the vesicles, due to curvature effects). Related to this is the fact that the authors seem to suggest that proton transport merely leads to a pH change (see discussion on acidification), whereas presumably the energetic coupling between the proton pumping system and the ATP synthase is through a transmembrane electrical potential gradient. Leak reactions mentioned by the reviewer must inevitably exist, but in the absence of sufficient data we do not account for them in the current model. This omission does not alter the primary conclusions that deal with the structure-function relationship of the chromatophore vesicle given the structurally resolved components (LH2, LH1, RC, bc1, ATP synthase) and cyt c2. For the description of steady-state kinetics based on rate-limiting effects in the system, it is not necessarily beneficial or informative to try to account for every subprocess in the system. Nonetheless, we discuss now for the benefit of the reader the presence of leak reactions as the reviewer suggests (Introduction, sixth paragraph; Discussion, fifth paragraph). Additionally, we also stress once more that a proton gradient involves an electrical potential difference. However, referring throughout the text to “proton gradient and electrical potential difference” or the like makes the reading of the text cumbersome; readers will grasp what is meant, if after a first more complete description we refer in the following part of the text solely to "proton gradient". Rerouting of electrons in the bc1 complex, mentioned by the referee as a leakage, is actually a more complex process leading to unwanted redox products, for example into the transformation of oxygen to superoxide, that needs to be controlled, e.g., through superoxide dismutase, and brings about more problems than just energy leakage (Rottenberg et al., J. Biol. Chem., 2009; 284: 19203-19210). Such rerouting is clearly beyond the scope of the present study that benefits from avoiding the discussion of numerous side reactions and keeping the focus on processes arising through the structurally resolved photosynthetic components of the chromatophore. Since more than one reviewer brought up the possible inclusion of peripheral cellular processes in the present model of chromatophore energy conversion, we would like to note here that we consider the simplifications due to ignoring such processes a benefit of our model, rather than a detriment. The present focus on photosynthetic processes, grounded in known structural detail, leads to a clear interpretation. Nevertheless, the peripheral processes coupled to various trans-membrane enzymes are now listed for the reader in several portions of the text, including in a new figure (Figure 5). Specifically, the processes of flagellar motility, NAD(P)H synthesis, coupling to the succinate/fumarate pool, as well as leak-reactions are not considered. Of these processes, coupling to flagellar motility probably constitutes the largest consumption channel when active. Furthermore, a steady-state model is considered so as not to explicitly simulate the diffusion of quinone/quinol, cytochrome c2, and protons, but rather representing the respective concentrations through their averages. Simulation efforts are beginning to address these diffusion processes, but the timescales currently reachable (~150 ns) are insufficient to account for the diffusive regime (Chandler, et al., 2014. Light harvesting by lamellar chromatophores in Rhodospirillum photometricum. Biophys. J., 106:2503). The steady-state approach permits us to successfully account for the primary function of the chromatophore, determining ATP synthesis rates, overall energy conversion efficiency, and saturation of light harvesting capacity, as depicted in Figures 2 and 3. To address reviewers' concerns for the benefit of the reader, we explicitly discuss, as mentioned already in response #1.1, the missing components of the chromatophore energy conversion absent from the current model in Introduction, fifth and sixth paragraphs; section 2.2, fifth paragraph; Discussion, fourth paragraph; section 4.1, sixth paragraph; subsection “Stage I: Light absorption, excitation energy transfer, and quinol formation”, eleventh paragraph; subsection “Stage II: Diffusion of charge carriers and estimate of cycling time τRC(I)”, fifth paragraph, as well as in an additional figure (Figure 5). 5) What precisely does the atomistic description of the light harvesting and trapping process contribute to the calculation of the efficiency? Could the energy transfer and trapping process not be captured by a single effective rate constant that is a function of the fraction of open and closed traps and the ratio of LH2 to RC-LH1. See many earlier papers on the subject: Vredenberg and Duysens, Duysens CIBA foundation, Den Hollander et al., Trissl et al., Joliot cs etc. The referee brings up an important point: what value does structural cell biology bring to the understanding of cells; did past knowledge, like that achieved over five decades ago by Vredenberg and Duysens and mentioned by the referee (Nature 197: 355-357, 1963; doi:10.1038/197355a0) not suffice today? In general, it seems to be widely accepted that one re-interprets biological knowledge in terms of the more recently discovered atomic level structure of cellular components as they provide a further resolved and more fundamental understanding of cell biology. So, the authors feel well justified in describing photosynthesis, including light harvesting, in the present case on the basis of the atomic level structure of the apparatus established in recent work. They also are convinced that given the fundamental role of light harvesting, the great advances in structural biology and in finite temperature quantum physics of fluorescent systems, it seems a natural goal to develop from ground up the kinetics of the light harvesting apparatus, rather than employ heuristic descriptions from a period when structural details of light harvesting proteins and their arrangement were not known. Nevertheless, we gladly follow the reviewer's suggestion and cite some of the original discoveries now also directly (Introduction, ninth paragraph; section 2.1, sixth paragraph). The authors are grateful to the reviewer for the insight in linking the earlier work of the Dutch photosynthesis masters and the modern highly resolved picture of photosynthetic apparatus structure and function. Reviewer #2: The manuscript by Sener and co-workers presents a model of a Rhodobacter sphaeroides chromatophore. The model describes how the illumination of a chromatophore suspension results in ATP production. The ultimate goal of the modeling is to understand how the ATP yield is coupled with the intensity of light. The main virtue of the model is that it couples the quantum physics of light absorption/procession with the chemistry of ATP synthesis. The model gives reasonable values of ATP yield in response to low light, which is physiological for these bacteria. Development of such a global model, which covers events over a time span of 14 orders of magnitude, required certain simplifications, which is understandable. Unfortunately, from the opinion of this reviewer, the simplifications preclude from making physiologically relevant conclusions from the modeling, at least, at the given stage. Still the model itself is a great achievement and the manuscript deserves publication albeit after certain amendments. 1) Since the physiological relevance of the results obtained is questionable, the main result of this study seems to be the model itself. Therefore, the description of the model should be moved from Materials and methods into the Results. Otherwise the manuscript is incomprehensive, taking into account that the format of eLife implies Materials and methods after Results. The Materials and methods section should then contain only technical/purely methodical points; supplementary materials, at least partly, could be also moved into Materials and methods. We thank the reviewer for the suggestion to moving material of our text between Results, Materials and methods, and supplementary materials. Though this could make for better reading in case of readers following the text each time from the beginning to end, readers are familiar today with the Results/Materials and methods/supplementary materials division of articles in scientific journals like eLife and know to jump between these sections for information of methodological details. We decided for the present division since a reader can find the information of the actual conclusions of our study more easily in the present format. The reader is also asked now to read the Materials and methods section prior to proceeding with the Results section (Results, first paragraph; Materials and methods, first paragraph). We also like to emphasize that the chromatophore structure, the ‘model’, is not the only result of our study; the other key result is the structure-function relationship of the whole chromatophore. This relationship links the overall structure of the chromatophore to ATP synthesis rate and energy conversion efficiency. 2) The current simplified model lacks two important modules, which, hopefully, would be added in the future. These are the protonmotive force and reversible NADH dehydrogenase. The authors admit the absence of these modules, but did not consider the consequences of their absence in full extent. This should be done in the Discussion. The particular short-comings of the current model, including, in particular, the missing NADH dehydrogenase component, are now discussed at greater length in the text, as mentioned above following comments by reviewer #1, in (Introduction, fifth and sixth paragraphs; section 2.2, fifth paragraph; Discussion, fourth paragraph; section 4.1, sixth paragraph; subsection “Stage I: Light absorption, excitation energy transfer, and quinol formation”, eleventh paragraph; subsection “Stage II: Diffusion of charge carriers and estimate of cycling time τRC(I)”, fifth paragraph) as well as in an additional figure (Figure 5). The proton gradient is left out of the analysis as proton migration from the bc1 complex to ATP synthase is not rate determining in chromatophore vesicles; we comment on the expected, fast proton migration rate in chromatophores in the text (section 4.2, fourth paragraph). 3) In the absence of these two modules, the authors conclude that "energy conversion in the chromatophore appears to be rate-limited primarily by cytbc1" and that "cytbc1-limited kinetics prevents an overproduction of protons at sustained high-light conditions,…thereby protecting the chromatophore against overacidification of its interior and assuring vesicle integrity at high-light conditions". In real phototropic bacteria, the situation is more complex. As mentioned for issue #2.2, we now discuss at length in the text the complexity of the coupling between the chromatophore energy conversion processes and cytoplasmic processes in the cell facilitated through NADH dehydrogenase, succinate dehydrogenase and other enzymes (Introduction, fifth and sixth paragraphs; section 2.2, fifth paragraph; Discussion, fourth paragraph; section 4.1, sixth paragraph; subsection “Stage I: Light absorption, excitation energy transfer, and quinol formation”, eleventh paragraph; subsection “Stage II: Diffusion of charge carriers and estimate of cycling time τRC(I)”, fifth paragraph; newly added figure, Figure 5). In particular, the rate-limiting role of bc1 and its relation to missing modules of the chromatophore is emphasized in the Discussion section (fourth paragraph; section 4.1, last paragraph). Obviously, the function of a photosynthetic apparatus, such as bacterial chromatophore or a chloroplast thylakoid, is to produce ATP and NAD(P)H. In thylakoids, NAD(P)H molecules are produced via direct reduction of NAD(P)+ by the Photosystem I. In Rb. sphaerides, the redox potential of the RC is too high for NAD(P)H generation. Therefore, NAD(P)H is produced via reversing the NAD(P)H dehydrogenase. The reversion requires (1) high protonmotive force and (2) reduced ubiquinol pool. Therefore, the overreduction of the ubiquinol pool in Rb. sphaeroides is a precondition for the NAD(P)H production; it should increase the overall efficiency of the system. Furthermore, the overacidification of the interior is prevented not by the overreduction of the ubiquinone pool, but by the high sensitivity of the cytbc1 to the membrane potential. High voltage, needed to drive the NAD(P)+ reduction (see above), slows the turnover of the cytbc1 dramatically, essentially blocking its turnover at voltages above 150 mV. Hence at high, physiological values of protonmotive force the turnover of the cytbc1 is essentially limited by the proton discharge, specifically via the ATP synthase and the reverse NAD(P)H dehydrogenase. The operation of the cytbc1 close to equilibrium additionally increases the efficiency of energy conversion. The reviewer raises great points about NAD(P)H production. We have taken the liberty to incorporate this comment in the text (Introduction: sixth paragraph; Results: section 2.2, fifth and sixth paragraphs; Discussion: fourth paragraph) and have added a corresponding Acknowledgment. As pointed out already in comments to referee #1 above, we emphasize the role of cytoplasmic redox reactions in the new text in all sections of the text and point to the important function of these reactions in regulating the redox state of the chromatophore Q-pool. The reviewer admits that the data on synthesis of NAD(P)H in Rb. sphaeroides are too scarce to be used for verification of the model. The authors, accordingly, have tested their model by using more abundant data that were obtained with chromatophore suspensions in the absence of NAD(P)+, in experiments where the function of the RC, cytbc1 and/or ATP synthase were studied. And still, the discussion of physiological consequences from the given model should be truncated dramatically as misleading. The physiology could be discussed in further publications, after the membrane potential and NAD(P)H synthesis would be explicitly included in the model. As pointed out above, we have added extensive discussions, based on reviewer comments, regarding the complexity of chromatophore energy conversion processes; the reader is made well aware of the presence of related cytoplasmic processes in the cell; and the interpretation about the buffering capacity in the system has been removed (see issue #1.1). Specifically, the following passages were modified: last sentence of the Abstract; Introduction, fifth paragraph; section 2.2, fourth and fifth paragraphs; Discussion, second and third paragraphs; section 4.2, fourth paragraph. We insist, however, that the consequences of the current structural-functional model be discussed in detail while informing the reader at length of the limitations of the current study. Instead, the Discussion section should include comparison of the given model with the earlier models of chromatophore systems (e.g. from the group of Helms) and the approaches used. The work of Geyer and Helms is already cited and discussed abundantly in the text, particularly, their conclusion that the primary rate-limiting step in the chromatophore stems from cytbc1turnover. We add a further comparison in the Discussion section (fifth paragraph). Reviewer #3: The authors present an important theoretical investigation of the performance and limitations of the photosynthetic chromatophore complex of purple bacteria. This light-harvesting system is of great interest as a model system that contains numerous characteristic properties inherent to all natural light-harvesting systems. This work is a continuation of a long line of first-rate papers published by the authors on this topic, is extremely well written, and merits publication in its current form. There appears to be a minor inconsistency in the units of intensity employed. In the second paragraph of section 4.2 the authors define light intensity, I, in units of [photons absorbed per second], whereas light intensity is often defined in units of [photons absorbed per meter squared per second]. As defined, this intensity is in fact a power, which is also called a 'radiant flux' in radiometry. This in contrast with the more conventional definition of intensity in the caption of Figure 2, where units of [Watts per meter squared] are used. Clarification of these different definitions would be useful. We thank the reviewer for pointing out a potential source of confusion (and also for the related next comment). The text has been altered to clarify the definition of I (photons/s absorbed by entire vesicle) as discussed in (section 4.2, second and third paragraphs). In the third paragraph of section 4.2, the authors switch to defining 'I' as 'illumination,' which may further confuse the reader. Ideally 'radiant flux' or 'absorbed light power' or similar terminology would be employed throughout, or alternatively 'illumination' would be defined. Following the reviewer's suggestion, we refer to the quantity 'I' in the text as 'absorbed light power'. The explanation provided in the fourth and fifth paragraphs of section 2.2 for the suboptimal (nB, nRC) values in low-light adapted vesicles is appealing, but it does not appear to be supported by calculations. Despite the fact that the model only applies in steady-state, is it possible to perform a simple simulation to support this interpretation, even qualitatively? We are grateful to the reviewer for pointing out this issue, also discussed in response to issue #1.1. We concede that our current steady-state model is not able to generate quantitative conclusions about the response of the system to fluctuating light conditions and the corresponding energy storage & buffering effect for intermittent dark periods. The related discussion (section 2.2, fourth and fifth paragraphs) is altered to reflect this point. Please see our extended response to reviewer 1, issue #1. However, a non-steady state simulation is currently prohibitive computationally, since explicit spatial dynamics of charge carriers and quinones/quinols in the chromatophore need to be modeled, which goes beyond the technically feasible limit. The choice of initial condition assumed in Equation 9 may possibly be over-simplifying because each complex (LH2, LH1 and RC) does not necessarily have the same absorption spectrum, since at a given wavelength, the probability of absorption of a photon is not the same for each complex. In this scenario, the initial condition should then be wavelength dependent. While the authors state that the quantum yield is not strongly dependent on the choice of P(0), a specific example to demonstrate this would be helpful. A discussion is added after Equations (8) and (9), defining the quantum yield q in terms of the initial state P(0) and explaining that variations in P(0) affect q by only about 3%. The text states that this result was demonstrated in (Seneret al., 2007, PNAS, 104:15723) and, therefore, wavelength dependence of q is not further considered in the present study. [Editors' note: further revisions were requested prior to acceptance, as described below.] On the basis of the referee reports on the revised manuscript, only relatively minor additional changes are required. Reviewer #2 is concerned that some issues have not been fully addressed in the revision. In the following I list the main points. 1) The reviewer maintains that "the model does not consider the production of NADH at all." Therefore, "no statements on the energy conversion efficiency could be done with the current model." The reviewer is concerned that the issue of NADH production and its impact on the model and its assessment are not discussed adequately, and advises "to read the primary literature on the mechanism of NADH production and provide references to this literature (currently they refer only to the paper of Klamt et al. where bacterial photosynthesis was modeled by using several dozen parameters that were more or less freely variable). The primary literature describes how the NADH-dehydrogenase makes NADH by driving the reverse (uphill) electron transfer from the ubiquinone pool to NAD+ with the help of membrane potential. This mechanism differs from the NAD(P)H production in green plants where NAD(P)+ is directly reduced by photoexcitated electrons in a reaction that is independent of membrane potential. The dependence of the light-driven NADH production on membrane potential, specific for Rhodobacter and other purple phototrophic bacteria, should have shaped the properties of their energy-converting machinery." Recommended action: Whereas I do not see the lack of NADH production as a fundamental flaw in the model (after all, all models are incomplete), this point should be made clearer to the reader. Conclusions drawn from the model should be qualified accordingly. In particular, in the discussion of the energy conversion efficiency, relevant limitations of the model should be stated. We were surprised about the reviewer response, since the presence of NADH production as well as the need for modeling it in future studies was stated repeatedly in the revised manuscript, while also discussing why it cannot be accounted for in the current study due to the absence of the relevant proteins from current structural data. In particular, we stated that our model, through Equations (15–19), describes the light intensity dependence of the quinone/quinol pool redox state that comes about through quinone-quinol conversion involving action of NADH dehydrogenase alongside other redox factors; thus, the influence of the redox factors are currently accounted for heuristically in the absence of structural data. We have now further emphasized this issue and the resulting shortcomings of the model in the main text at the following locations: Introduction, sixth paragraph; Discussion, fourth paragraph; section 4.1 last paragraph; section 4.2, last paragraph; subsection “Stage I: Light absorption, excitation energy transfer, and quinol formation”, ninth paragraph; subsection “Stage II: Diffusion of charge carriers and estimate of cycling time τRC(I)”, first paragraph. 2) The reviewer also asks to point out clearly that "the experimental data that were used for testing the model were not obtained with chromatophores in vivo," but for "an artificial system – a suspension of chromatophores in a pH-buffer, in the presence of electron donors and redox mediators, but in the absence of physiological compounds such as NAD+/NADH (a Dutton/Crofts system)." Recommended action: Point out the distinction between the in-vitro experiments and the more complex in-vivo situation. This distinction and the increased complexity of the in vivo situation is now emphasized in the text at: section 4.1, end of first paragraph. 3) The reviewer also argues that the effect of a transmembrane proton gradient, which is not included explicitly in the model, should be discussed. The reviewer emphasizes that the issue is not "proton migration from the bc1 complex to ATP synthase" but protonmotive force. Recommended action: Discuss the issue of protonmotive force. Examine which conclusions would likely be impacted by the explicit treatment of PMF in the model, and qualify them accordingly. If none of the conclusions would be affected, state that the model is robust with respect to the neglect of explicit modeling of PMF. We have added additional text in Discussion (sixth paragraph) and Methods (subsection “Stage I: Light absorption, excitation energy transfer, and quinol formation”, eleventh paragraph) to emphasize that the energy conversion model does not require the explicit treatment of proton translocation steps and is robust with respect to the neglect of explicit modeling of PMF in its prediction of ATP turnover and efficiency. 4) The reviewer suggests adding a description of the model to the Results section. As already argued in our last submission, regarding the fourth issue we would rather keep the division between Results, Methods, and supplementary information in the present form. ==== Refs References Adams PG Hunter CN 2012 Adaptation of intracytoplasmic membranes to altered light intensity in Rhodobacter sphaeroides Biochimica et Biophysica Acta 1817 1616 1627 10.1016/j.bbabio.2012.05.013 22659614 Agmon N 1995 The Grotthuss mechanism Chemical Physics Letters 244 456 462 10.1016/0009-2614(95)00905-J Bahatyrova S Frese RN Siebert CA Olsen JD Van Der Werf KO Van Grondelle R Niederman RA Bullough PA Otto C Hunter CN 2004 The native architecture of a photosynthetic membrane Nature 430 1058 1062 10.1038/nature02823 15329728 Baradaran R Berrisford JM Minhas GS Sazanov LA 2013 Crystal structure of the entire respiratory complex I Nature 494 443 448 10.1038/nature11871 23417064 Beekman LM van Mourik F Jones MR Visser HM Hunter CN van Grondelle R 1994 Trapping kinetics in mutants of the photosynthetic purple bacterium Rhodobacter sphaeroides : influence of the charge separation rate and consequences for the rate-limiting step in the light-harvesting process Biochemistry 33 3143 3147 10.1021/bi00177a001 8136347 Berera R van Grondelle R Kennis JT 2009 Ultrafast transient absorption spectroscopy: principles and application to photosynthetic systems Photosynthesis Research 101 105 118 10.1007/s11120-009-9454-y 19578970 Berg JM Tymoczko JL Stryer L 2011 Biochemistry 7th edition Palgrave MacMillan Blankenship RE Tiede DM Barber J Brudvig GW Fleming G Ghirardi M Gunner MR Junge W Kramer DM Melis A Moore TA Moser CC Nocera DG Nozik AJ Ort DR Parson WW Prince RC Sayre RT 2011 Comparing photosynthetic and photovoltaic efficiencies and recognizing the potential for improvement Science 332 805 809 10.1126/science.1200165 21566184 Blankenship RE 2014 Molecular Mechanisms of Photosynthesis 2nd Ed Wiley Blackwell Bowyer JR Hunter CN Ohnishi T Niederman RA 1985 Photosynthetic membrane development in Rhodopseudomonas sphaeroides . Spectral and kinetic characterization of redox components of light-driven electron flow in apparent photosynthetic membrane growth initiation sites Journal of Biological Chemistry 260 3295 3304 2982855 Bradforth SE Jimenez R van Mourik F van Grondelle R Fleming GR 1995 Excitation transfer in the core light-harvesting complex (LH-1) of Rhodobacter sphaeroides : An ultrafast fluorescence depolarization and annihilation study Journal of Physical Chemistry 99 16179 16191 10.1021/j100043a071 Cartron ML Olsen JD Sener M Jackson PJ Brindley AA Qian P Dickman MJ Leggett GJ Schulten K Hunter CN 2014 Integration of energy and electron transfer processes in the photosynthetic membrane of Rhodobacter sphaeroides Biochimica et Biophysica Acta 1837 1769 1780 10.1016/j.bbabio.2014.02.003 24530865 Caycedo-Soler F Rodríguez FJ Quiroga L Johnson NF 2010 Interplay between excitation kinetics and reaction-center dynamics in purple bacteria New Journal of Physics 12 095008 10.1088/1367-2630/12/9/095008 Chandler DE Gumbart J Stack JD Chipot C Schulten K 2009 Membrane curvature induced by aggregates of LH2s and monomeric LH1s Biophysical Journal 97 2978 2984 10.1016/j.bpj.2009.09.007 19948127 Chandler DE Strümpfer J Sener M Scheuring S Schulten K 2014 Light harvesting by lamellar chromatophores in Rhodospirillum photometricum Biophysical Journal 106  2503 2510 10.1016/j.bpj.2014.04.030 Clark AJ Cotton NPJ Jackson JB 1983 The relation between membrane ionic current and ATP synthesis in chromatophores from Rhodopseudomonas capsulata Biochimica et Biophysica Acta 723 440 453 10.1016/0005-2728(83)90051-8 Cogdell RJ Gall A Köhler J 2006 The architecture and function of the light-harvesting apparatus of purple bacteria: from single molecules to in vivo membranes Quarterly Reviews of Biophysics 39 227 324 10.1017/S0033583506004434 17038210 Comayras F Jungas C Lavergne J 2005 Functional consequences of the organization of the photosynthetic apparatus in Rhodobacter sphaeroides . I. Quinone domains and excitation transfer in chromatophores and reaction center.antenna complexes Journal of Biological Chemistry 280 11203 11213 10.1074/jbc.M412088200 15632164 Cory MG Zerner MC Hu X Schulten K 1998 Electronic excitations in aggregates of bacteriochlorophylls Journal of Physical Chemistry B 102 7640 7650 10.1021/jp980966l Crielaard W Visschers RW Fowler GJS van Grondelle R Hellingwerf KJ Hunter CN 1994 Probing the B800 bacteriochlorophyll binding site of the accessory light-harvesting complex from Rhodobacter sphaeroides using site-directed mutants. I. Mutagenesis, effects on binding, function and electrochromic behaviour of its carotenoids Biochimica et Biophysica Acta 1183 473 482 10.1016/0005-2728(94)90074-4 Crofts AR Lhee S Crofts SB Cheng J Rose S 2006 Proton pumping in the bc1 complex: A new gating mechanism that prevents short circuits Biochimica et Biophysica Acta 1757 1019 1034 10.1016/j.bbabio.2006.02.009 16600173 Crofts AR 2004 The cytochrome bc1 complex: function in the context of structure Annual Review of Physiology 66 689 733 10.1146/annurev.physiol.66.032102.150251 Damjanović A Kosztin I Kleinekathöfer U Schulten K 2002 Excitons in a photosynthetic light-harvesting system: A combined molecular dynamics, quantum chemistry, and polaron model study Physical Review E 65 031919 10.1103/PhysRevE.65.031919 Damjanović A Ritz T Schulten K 1999 Energy transfer between carotenoids and bacteriochlorophylls in light-harvesting complex II of purple bacteria Physical Review E 59 3293 3311 10.1103/PhysRevE.59.3293 Damjanovi A Ritz T Schulten K 2000 Excitation energy trapping by the reaction center of Rhodobacter Sphaeroides International Journal of Quantum Chemistry 77 139 151 10.1002/(SICI)1097-461X(2000)77:1<139::AID-QUA13>3.0.CO;2-S Damjanović A Vaswani HM Fromme P Fleming GR 2002 Chlorophyll Excitations in Photosystem I of Synechococcus elongatus Journal of Physical Chemistry B 106 10251 10262 10.1021/jp020963f Den Hollander WTF Bakker JGC Van Grondelle R 1983 Trapping, loss and annihilation of excitations in a photosynthetic system. I. Theoretical aspects Biochimica et Biophysica Acta 725 492 507 10.1016/0005-2728(83)90190-1 Dudkina NV Kudryashev M Stahlberg H Boekema EJ 2011 Interaction of complexes I, III, and IV within the bovine respirasome by single particle cryoelectron tomography PNAS 108 15196 15200 10.1073/pnas.1107819108 21876144 Etzold C Deckers-Hebestreit G Altendorf K 1997 Turnover number of Escherichia coli FO F1 ATP synthase for ATP synthesis in membrane vesicles European Journal of Biochemistry 243 336 343 10.1111/j.1432-1033.1997.0336a.x 9030757 Feniouk BA Cherepanov DA Voskoboynikova NE Mulkidjanian AY Junge W 2002 Chromatophore vesicles of Rhodobacter capsulatus  contain on average one FO F1 -ATP synthase each Biophysical Journal 82 1115 1122 10.1016/S0006-3495(02)75470-2 11867431 Feniouk BA Junge W 2009 Proton translocation and ATP synthesis by the FFATPase of purple bacteria Hunter C N Daldal, F Thurnhauer M Beatty J T The Purple Phototropic Bacteria 475 493 10.1007/978-1-4020-8815-5_24 Freiberg A Rätsep M Timpmann K Trinkunas G 2009 Excitonic polarons in quasi-one-dimensional LH1 and LH2 bacteriochlorophyll a antenna aggregates from photosynthetic bacteria: A wavelength-dependent selective spectroscopy study Chemical Physics 357 102 112 10.1016/j.chemphys.2008.10.043 Frese RN Pàmies JC Olsen JD Bahatyrova S van der Weij-de Wit CD Aartsma TJ Otto C Hunter CN Frenkel D van Grondelle R 2008 Protein shape and crowding drive domain formation and curvature in biological membranes Biophysical Journal 94 640 647 10.1529/biophysj.107.116913 17827217 Frese RN Siebert CA Niederman RA Hunter CN Otto C van Grondelle R 2004 The long-range organization of a native photosynthetic membrane PNAS 101 17994 17999 10.1073/pnas.0407295102 15601770 Förster T 1948 Zwischenmolekulare energiewanderung und fluoreszenz Annalen Der Physik 437 55 75 10.1002/andp.19484370105 Geyer T Helms V 2006 A spatial model of the chromatophore vesicles of Rhodobacter sphaeroides and the position of the Cytochrome bc1 complex Biophysical Journal 91 921 926 10.1529/biophysj.105.078501 16714339 Geyer T Lauck F Helms V 2007 Molecular stochastic simulations of chromatophore vesicles from Rhodobacter sphaeroides Journal of Biotechnology 129 212 228 10.1016/j.jbiotec.2006.12.018 17276535 Geyer T Mol X Blass S Helms V 2010 Bridging the gap: linking molecular simulations and systemic descriptions of cellular compartments PLoS One 5 e09541 10.1371/journal.pone.0014070 Gobets B Kennis JTM Ihalainen JA Brazzoli M Croce R van Stokkum IHM Bassi R Dekker JP van Amerongen H Fleming GR van Grondelle R Fleming GR van Grondelle R 2001 Excitation energy transfer in dimeric light harvesting complex I: A combined streak-camera/fluorescence upconversion study Journal of Physical Chemistry B 105 10132 10139 10.1021/jp011901c Gubellini F Francia F Turina P Lévy D Venturoli G Melandri BA 2007 Heterogeneity of photosynthetic membranes from Rhodobacter capsulatus : size dispersion and ATP synthase distribution Biochimica et Biophysica Acta 1767 1340 1352 10.1016/j.bbabio.2007.08.007 17961501 Hakobyan L Gabrielyan L Trchounian A 2012 Relationship of proton motive force and the F(0)F (1)-ATPase with bio-hydrogen production activity of Rhodobacter sphaeroides : effects of diphenylene iodonium, hydrogenase inhibitor, and its solvent dimethylsulphoxide Journal of Bioenergetics and Biomembranes 44 495 502 10.1007/s10863-012-9450-3 22689145 Harris MA Parkes-Loach PS Springer JW Jiang J Martin EC Qian P Jiao J Niedzwiedzki DM Kirmaier C Olsen JD Bocian DF Holten D Hunter CN Lindsey JS Loach PA 2013 Integration of multiple chromophores with native photosynthetic antennas to enhance solar energy capture and delivery Chemical Science 4 3924 3933 10.1039/c3sc51518d Hellingwerf KJ Crielaard W Westerhoff HV 1993 Comparison of retinal-based and chlorophyll-based photosynthesis: A Biothermokinetic Description of Photochemical Reaction Centers Schuster S Rigoulet M Ouhabi R Mazat JP Modern Trends in Biothermokinetics Plenum Press 45 92 10.1007/978-1-4615-2962-0_9 Hess S Chachisvilis M Timpmann K Jones MR Fowler GJ Hunter CN Sundstrom V 1995 Temporally and spectrally resolved subpicosecond energy transfer within the peripheral antenna complex (LH2) and from LH2 to the core antenna complex in photosynthetic purple bacteria PNAS 92 12333 12337 10.1073/pnas.92.26.12333 11607622 Hess S Visscher KJ Pullerits T Sundström V Fowler GJ Hunter CN 1994 Enhanced rates of subpicosecond energy transfer in blue-shifted light harvesting LH2 mutants of Rhodobacter sphaeroides Biochemistry 33 8300 8305 10.1021/bi00193a017 8031762 Hsin J Chandler DE Gumbart J Harrison CB Sener M Strumpfer J Schulten K 2010a Self-Assembly of Photosynthetic Membranes ChemPhysChem 11 1154 1159 10.1002/cphc.200900911 20183845 Hsin J Gumbart J Trabuco LG Villa E Qian P Hunter CN Schulten K 2009 Protein-induced membrane curvature investigated through molecular dynamics flexible fitting Biophysical Journal 97  321 329 10.1016/j.bpj.2009.04.031 Hsin J Strümpfer J Şener M Qian P Hunter CN Schulten K 2010b Energy transfer dynamics in an RC–LH1–PufX tubular photosynthetic membrane New Journal of Physics 12 e09541 10.1088/1367-2630/12/8/085005 Hu X Damjanovic A Ritz T Schulten K 1998 Architecture and mechanism of the light-harvesting apparatus of purple bacteria PNAS 95 5935 5941 10.1073/pnas.95.11.5935 9600895 Hu X Ritz T Damjanović A Autenrieth F Schulten K Damjanovic A Schulten K 2002 Photosynthetic apparatus of purple bacteria Quarterly Reviews of Biophysics 35  1 62 10.1017/S0033583501003754 Hu X Ritz T Damjanović A Schulten K 1997 Pigment organization and transfer of electronic excitation in the photosynthetic unit of purple bacteria Journal of Physical Chemistry B 101 3854 3871 10.1021/jp963777g Humphrey W Dalke A Schulten K 1996 VMD: Visual molecular dynamics Journal of Molecular Graphics 14 33 38 10.1016/0263-7855(96)00018-5 8744570 Hunter CN Kramer HJM van Grondelle R 1985 Linear dichroism and fluorescence emission of antenna complexes during photosynthetic unit assembly in Rhodopseudomonas sphaeroides Biochimica et Biophysica Acta 807 44 51 10.1016/0005-2728(85)90051-9 Ishizaki A Fleming GR 2009a Theoretical examination of quantum coherence in a photosynthetic system at physiological temperature PNAS 106 17255 17260 10.1073/pnas.0908989106 19815512 Ishizaki A Fleming GR 2009b Unified treatment of quantum coherent and incoherent hopping dynamics in electronic energy transfer: reduced hierarchy equation approach Journal of Chemical Physics 130 234111 10.1063/1.3155372 19548715 Jackson PJ Lewis HJ Tucker JD Hunter CN Dickman MJ 2012 Quantitative proteomic analysis of intracytoplasmic membrane development in Rhodobacter sphaeroides Molecular Microbiology 84 1062 1078 10.1111/j.1365-2958.2012.08074.x 22621241 Jamieson SJ Wang P Qian P Kirkland JY Conroy MJ Hunter CN Bullough PA 2002 Projection structure of the photosynthetic reaction centre-antenna complex of Rhodospirillum rubrum at 8.5 Å resolution EMBO Journal 21 3927 3935 10.1093/emboj/cdf410 12145194 Janusonis J Valkunas L Rutkauskas D van Grondelle R 2008 Spectral dynamics of individual bacterial light-harvesting complexes: alternative disorder model Biophysical Journal 94 1348 1358 10.1529/biophysj.107.108431 17921215 Jimenez R van Mourik F Yu JY Fleming GR 1997 Three-pulse photon echo measurements on LH1 and LH2 complexes of Rhodobacter sphaeroides : A nonlinear spectroscopic probe of energy transfer Journal of Physical Chemistry B 101 7350 7359 10.1021/jp970299g Joliot P Verméglio A Joliot A 1990 Electron transfer between primary and secondary donors in Rhodospirillum rubrum : evidence for a dimeric association of reaction centers Biochemistry 29  4355  4361 10.1021/bi00470a014 Joliot P Verméglio A Joliot A 1996 Supramolecular organization of the photosynthetic chain in chromatophores and cells of Rhodobacter sphaeroides Photosynthesis Research 48 291 299 10.1007/BF00041020 24271310 Jordanides XJ Scholes GD Shapley WA Reimers JR Fleming GR 2004 Electronic Couplings and Energy Transfer Dynamics in the Oxidized Primary Electron Donor of the Bacterial Reaction Center Journal of Physical Chemistry B 108 1753 1765 10.1021/jp036516x Jungas C Ranck JL Rigaud JL Joliot P Verméglio A 1999 Supramolecular organization of the photosynthetic apparatus of Rhodobacter sphaeroides EMBO Journal 18 534 542 10.1093/emboj/18.3.534 9927413 Kirchhoff H Mukherjee U Galla HJ 2002 Molecular architecture of the thylakoid membrane: lipid diffusion space for plastoquinone Biochemistry 41 4872 4882 10.1021/bi011650y 11939782 Klamt S Grammel H Straube R Ghosh R Gilles ED 2008 Modeling the electron transport chain of purple non-sulfur bacteria Molecular Systems Biology 4 e09541 10.1038/msb4100191 Koepke J Hu X Muenke C Schulten K Michel H Michel H 1996 The crystal structure of the light-harvesting complex II (B800–850) from Rhodospirillum molischianum Structure 4 581 597 10.1016/S0969-2126(96)00063-9 8736556 Kojadinovic M Armitage JP Tindall MJ Wadhams GH 2013 Response kinetics in the complex chemotaxis signalling pathway of Rhodobacter sphaeroides Journal of the Royal Society Interface 10 e09541 10.1098/rsif.2012.1001 Kosztin I Schulten K 2014 Function, and quantum dynamics of pigment-protein complexes Mohseni M Omar Y Engel G Plenio M. B Quantum Effects in Biology 123 143 Lavergne J Verméglio A Joliot P 2009 Functional coupling between reaction centers and cytochrome bc1 complexes Hunter CN Daldal F Thurnauer MC Beatty JT The Purple Phototrophic Bacteria 509 536 10.1007/978-1-4020-8815-5_26 Lavergne J 2009 Clustering of electron transfer components: Kinetic and thermodynamic consequences Laisk, A Nedbal L Govindjee Photosynthesis in silico: Understanding Complexity from Molecules to Ecosystems Springer 177 205 10.1007/978-1-4020-9237-4_8 Ma Y-Z Cogdell RJ Gillbro T 1997 Energy transfer and exciton annihilation in the B800−850 antenna complex of the photosynthetic purple bacterium Rhodopseudomonas acidophila (Strain 10050). A femtosecond transient absorption study Journal of Physical Chemistry B 101 1087 1095 10.1021/jp962470e Mauzerall D 1986 The optical cross section and absolute size of a photosynthetic unit Photosynthesis Research 10 163 170 10.1007/BF00118279 24435361 McDermott G Prince SM Freer AA Hawthornthwaite-Lawless AM Papiz MZ Cogdell RJ Isaacs NW 1995 Crystal structure of an integral membrane light-harvesting complex from photosynthetic bacteria Nature 374 517 521 10.1038/374517a0 Niederman RA 2013 Membrane development in purple photosynthetic bacteria in response to alterations in light intensity and oxygen tension Photosynthesis Research 116 333 348 10.1007/s11120-013-9851-0 23708977 Novoderezhkin VI Razjivin AP 1996 The theory of Forster-type migration between clusters of strongly interacting molecules: application to light-harvesting complexes of purple bacteria Chemical Physics 211 203 214 10.1016/0301-0104(96)00130-9 Noy D Moser CC Dutton PL 2006 Design and engineering of photosynthetic light-harvesting and electron transfer using length, time, and energy scales Biochimica et Biophysica Acta 1757 90 105 10.1016/j.bbabio.2005.11.010 16457774 Noy D 2008 Natural photosystems from an engineer's perspective: length, time, and energy scales of charge and energy transfer Photosynthesis Research 95 23 35 10.1007/s11120-007-9269-7 17968671 Olaya-Castro A Lee CF Olsen FF Johnson NF 2008 Efficiency of energy transfer in a light-harvesting system under quantum coherence Physical Review B 78 e09541 10.1103/PhysRevB.78.085115 Olsen JD Tucker JD Timney JA Qian P Vassilev C Hunter CN 2008 The organization of LH2 complexes in membranes from Rhodobacter sphaeroides Journal of Biological Chemistry 283 30772 30779 10.1074/jbc.M804824200 18723509 Panitchayangkoon G Hayes D Fransted KA Caram JR Harel E Wen J Blankenship RE Engel GS 2010 Long-lived quantum coherence in photosynthetic complexes at physiological temperature PNAS 107 12766 12770 10.1073/pnas.1005484107 20615985 Papiz MZ Prince SM Howard T Cogdell RJ Isaacs NW 2003 The structure and thermal motion of the B800-850 LH2 complex from Rps.acidophila at 2.0A resolution and 100K: new structural features and functionally relevant motions Journal of Molecular Biology 326  1523  1538 10.1016/S0022-2836(03)00024-X Patole S Vasilev C El-Zubir O Wang L Johnson MP Cadby AJ Leggett GJ Hunter CN 2015 Interference lithographic nanopatterning of plant and bacterial light-harvesting complexes on gold substrates Interface Focus 5 e09541 10.1098/rsfs.2015.0005 Pawlowicz NP van Grondelle R van Stokkum IH Breton J Jones MR Groot ML 2008 Identification of the first steps in charge separation in bacterial photosynthetic reaction centers of Rhodobacter sphaeroides by ultrafast mid-infrared spectroscopy: electron transfer and protein dynamics Biophysical Journal 95 1268 1284 10.1529/biophysj.108.130880 18424493 Peterson DH Perry MJ Bencala KE Talbot MC 1987 Phytoplankton productivity in relation to light intensity: A simple equation Estuarine, Coastal and Shelf Science 24 813 832 10.1016/0272-7714(87)90154-5 Pullerits T Visscher KJ Hess S Sundström V Freiberg A Timpmann K van Grondelle R 1994 Energy transfer in the inhomogeneously broadened core antenna of purple bacteria: a simultaneous fit of low-intensity picosecond absorption and fluorescence kinetics Biophysical Journal 66 236 248 10.1016/S0006-3495(94)80770-2 8130341 Qian P Bullough PA Hunter CN 2008 Three-dimensional reconstruction of a membrane-bending complex: the RC-LH1-PufX core dimer of Rhodobacter sphaeroides Journal of Biological Chemistry 283 14002 14011 10.1074/jbc.M800625200 18326046 Qian P Hunter CN Bullough PA 2005 The 8.5 Å projection structure of the core RC-LH1-PufX dimer of Rhodobacter sphaeroides Journal of Molecular Biology 349 948 960 10.1016/j.jmb.2005.04.032 15907932 Qian P Papiz MZ Jackson PJ Brindley AA Ng IW Olsen JD Dickman MJ Bullough PA Hunter CN 2013 Three-dimensional structure of the Rhodobacter sphaeroides RC-LH1-PufX complex: dimerization and quinone channels promoted by PufX Biochemistry 52 7575 7585 10.1021/bi4011946 24131108 Rebentrost P Mohseni M Aspuru-Guzik A 2009 Role of quantum coherence and environmental fluctuations in chromophoric energy transport Journal of Physical Chemistry B 113 9942 9947 10.1021/jp901724d Remy A Gerwert K 2003 Coupling of light-induced electron transfer to proton uptake in photosynthesis Nature Structural Biology 10 637 644 10.1038/nsb954 12872158 Reynolds NP Janusz S Escalante-Marun M Timney J Ducker RE Olsen JD Otto C Subramaniam V Leggett GJ Hunter CN 2007 Directed formation of micro- and nanoscale patterns of functional light-harvesting LH2 complexes Journal of the American Chemical Society 129 14625 14631 10.1021/ja073658m 17985885 Ritz T Adem S Schulten K 2000a A model for photoreceptor-based magnetoreception in birds Biophysical Journal 78 707 718 10.1016/S0006-3495(00)76629-X 10653784 Ritz T Damjanović A Zhang JP Zhang JP Koyama Y 2000b Efficient light harvesting through carotenoids Photosynthesis Research 66 125 144 10.1023/A:1010750332320 16228415 Ritz T Hu X Damjanović A Schulten K 1998 Excitons and excitation transfer in the photosynthetic unit of purple bacteria Journal of Luminescence 76-77 310 321 10.1016/S0022-2313(97)00286-X Ritz T Park S Schulten K 2001 Kinetics of excitation migration and trapping in the photosynthetic unit of purple bacteria Journal of Physical Chemistry B 105 8259 8267 10.1021/jp011032r Roberts VA Thompson EE Pique ME Perez MS Ten Eyck LF 2013 DOT2: Macromolecular docking with improved biophysical models Journal of Computational Chemistry 34 1743 1758 10.1002/jcc.23304 23695987 Saphon S Jackson JB Witt HT 1975 Electrical potential changes, H+ translocation and phosphorylation induced by short flash excitation in Rhodopseudomonas sphaeroides chromatophores Biochimica et Biophysica Acta 408 67 82 10.1016/0005-2728(75)90159-0 240444 Scheuring S Boudier T Sturgis JN 2007 From high-resolution AFM topographs to atomic models of supramolecular assemblies Journal of Structural Biology 159 268 276 10.1016/j.jsb.2007.01.021 17399998 Scheuring S Gonçalves RP Prima V Sturgis JN 2006 The photosynthetic apparatus of Rhodopseudomonas palustris : structures and organization Journal of Molecular Biology 358 83 96 10.1016/j.jmb.2006.01.085 16500674 Scheuring S Sturgis JN 2009 Atomic force microscopy of the bacterial photosynthetic apparatus: plain pictures of an elaborate machinery Photosynthesis Research 102 197 211 10.1007/s11120-009-9413-7 19266309 Scholes GD Fleming GR Olaya-Castro A van Grondelle R 2011 Lessons from nature about solar light harvesting Nature Chemistry 3 763 774 10.1038/nchem.1145 Scholes GD Jordanides XJ Fleming GR 2001 Adapting the Förster theory of energy transfer for modeling dynamics in aggregated molecular assemblies Journal of Physical Chemistry B 105 1640 1651 10.1021/jp003571m Scholes GD 2010 Quantum-coherent electronic energy transfer: Did nature think of it first? Journal of Physical Chemistry Letters 1 2 8 10.1021/jz900062f Sener M Hsin J Trabuco LG Villa E Qian P Hunter CN Schulten K 2009 Structural model and excitonic properties of the dimeric RC-LH1-PufX complex from Rhodobacter sphaeroides Chemical Physics 357 188 197 10.1016/j.chemphys.2009.01.003 20161332 Sener M Stone JE Barragan A Singharoy A Teo I Vandivort KL Isralewitz B Liu B Goh BC Phillips JC Kourkoutis LF Hunter CN Schulten K 2014 Visualization of energy conversion processes in a light harvesting organelle at atomic detail Proceedings of the International Conference on High Performance Computing, Networking Storage and Analysis IEEE Press Sener M Strümpfer J Hsin J Chandler D Scheuring S Hunter CN Schulten K 2011 Förster energy transfer theory as reflected in the structures of photosynthetic light-harvesting systems ChemPhysChem 12 518 531 10.1002/cphc.201000944 21344591 Sener M Strümpfer J Timney JA Freiberg A Hunter CN Schulten K 2010 Photosynthetic vesicle architecture and constraints on efficient energy harvesting Biophysical Journal 99  67 75 10.1016/j.bpj.2010.04.013 Şener MK Lu D Ritz T Park S Fromme P Schulten K 2002 Robustness and optimality of light harvesting in cyanobacterial photosystem I Journal of Physical Chemistry B 106 7948 7960 10.1021/jp020708v Şener MK Olsen JD Hunter CN Schulten K 2007 Atomic-level structural and functional model of a bacterial photosynthetic membrane vesicle PNAS 104 15723 15728 10.1073/pnas.0706861104 17895378 Şener MK Schulten K 2002 A general random matrix approach to account for the effect of static disorder on the spectral properties of light harvesting systems Physical Review E 65 031916 10.1103/physreve.65.031916 Şener MK Schulten K 2008 From atomic-level structure to supramolecular organization in the photosynthetic unit of purple bacteria Daldal F Thurnauer M. C Beatty J. T The Purple Phototrophic Bacteria 28 Springer 275 294 10.1007/978-1-4020-8815-5_15 Siebert CA Qian P Fotiadis D Engel A Hunter CN Bullough PA 2004 Molecular architecture of photosynthetic membranes in Rhodobacter sphaeroides : the role of PufX EMBO Journal 23 690 700 10.1038/sj.emboj.7600092 14765115 Small GJ 1995 On the validity of the standard model for primary charge separation in the bacterial reaction center Chemical Physics 197 239 257 10.1016/0301-0104(95)00089-7 Strümpfer J Schulten K 2012a Excited state dynamics in photosynthetic reaction center and light harvesting complex 1 Journal of Chemical Physics 137 065101 10.1063/1.4738953 22897312 Strümpfer J Hsin J Sener M Chandler D Schulten K 2011 The light-harvesting apparatus in purple photosynthetic bacteria, introduction to a quantum biological device Roux, B Molecular Machines 19 48 10.1142/9789814343466_0002 Strümpfer J Schulten K 2009 Light harvesting complex II B850 excitation dynamics The Journal of Chemical Physics 131  225101 10.1063/1.3271348 Strümpfer J Schulten K 2011 The effect of correlated bath fluctuations on exciton transfer The Journal of Chemical Physics 134 095102 10.1063/1.3557042 21385000 Strümpfer J Schulten K 2012b Open quantum dynamics calculations with the hierarchy equations of motion on parallel computers Journal of Chemical Theory and Computation 8 2808 2816 10.1021/ct3003833 23105920 Strümpfer J Sener M Schulten K 2012 How quantum coherence assists photosynthetic light harvesting Journal of Physical Chemistry Letters 3 536 542 10.1021/jz201459c 22844553 Sturgis JN Niedermann RA 1996 The effect of different levels of the B800-850 light-harvesting complex on intracytoplasmic membrane development in Rhodobacter sphaeroides Archives of Microbiology 165 235 242 10.1007/s002030050321 8952944 Sumi H 1999 Theory on rates of excitation-energy transfer between molecular aggregates through distributed transition dipoles with application to the antenna system in bacterial photosynthesis Journal of Physical Chemistry B 103 252 260 10.1021/jp983477u Tretiak S Middleton C Chernyak V Mukamel S 2000 Bacteriochlorophyll and carotenoid excitonic couplings in the LH2 system of purple bacteria Journal of Physical Chemistry B 104 9540 9553 10.1021/jp001585m Tucker JD Siebert CA Escalante M Adams PG Olsen JD Otto C Stokes DL Hunter CN 2010 Membrane invagination in Rhodobacter sphaeroides is initiated at curved regions of the cytoplasmic membrane, then forms both budded and fully detached spherical vesicles Molecular Microbiology 76 833 847 10.1111/j.1365-2958.2010.07153.x 20444085 Valkunas L Janusonis J Rutkauskas D van Grondelle R 2007 Protein dynamics revealed in the excitonic spectra of single LH2 complexes Journal of Luminescence 127 269 275 10.1016/j.jlumin.2007.02.032 van Amerongen H Valkunas L van Grondelle R 2000 Photosynthetic Excitons Singapore World Scientific van Grondelle R Dekker JP Gillbro T Sundstrom V 1994 Energy transfer and trapping in photosynthesis Biochimica et Biophysica Acta 1187 1 65 10.1016/0005-2728(94)90166-X van Grondelle R Novoderezhkin VI 2006a Energy transfer in photosynthesis: experimental insights and quantitative models Physical Chemistry Chemical Physics 8 793 807 10.1039/B514032C 16482320 van Grondelle R Novoderezhkin VI 2006b Energy transfer in photosynthesis: experimental insights and quantitative models Phys. Chem. Chem. Phys. 8 793 807 10.1039/B514032C 16482320 Vasilev C Johnson MP Gonzales E Wang L Ruban AV Montano G Cadby AJ Hunter CN 2014 Reversible switching between nonquenched and quenched states in nanoscale linear arrays of plant light-harvesting antenna complexes Langmuir 30 8481 8490 10.1021/la501483s 24988144 Visscher KJ Bergström H Sundström V Hunter CN Van Grondelle R 1989 Temperature dependence of energy transfer from the long wavelength antenna BChl-896 to the reaction center in Rhodospirillum rubrum , Rhodobacter sphaeroides (w.t. and M21 mutant) from 77 to 177K, studied by picosecond absorption spectroscopy Photosynthesis Research 22 211 217 10.1007/BF00048300 24424811 Visser HM Somsen OJG van Mourik F van Grondelle R 1996 Excited-state energy equilibration via subpicosecond energy transfer within the inhomogeneously broadened light-harvesting antenna of LH-1-only Rhodobacter sphaeroides mutant M2192 at room temperature and 4.2 K Journal of Physical Chemistry 100 18859 18867 10.1021/jp960883+ Vredenberg WJ Duysens LN 1963 Transfer of energy from bacteriochlorophyll to a reaction centre during bacterial photosynthesis Nature 197 355 357 10.1038/197355a0 13998015 Woodbury NW Parson WW 1984 Nanosecond fluorescence from isolated photosynthetic reaction centers of Rhodopseudomonas sphaeroides Biochimica et Biophysica Acta 767 345 361 10.1016/0005-2728(84)90205-6 6333897 Woronowicz K Harrold JW Kay JM Niederman RA 2013 Structural and functional proteomics of intracytoplasmic membrane assembly in Rhodobacter sphaeroides Journal of Molecular Microbiology and Biotechnology 23 48 62 10.1159/000346520 23615195 Woronowicz K Niederman RA 2010 Proteomic Analysis of the Developing Intracytoplasmic Membrane in Rhodobacter sphaeroides During Adaptation to Low Light Intensity New York. Springer 10.1007/978-1-4419-1528-3_10 Woronowicz K Sha D Frese RN Niederman RA 2011a The accumulation of the light-harvesting 2 complex during remodeling of the Rhodobacter sphaeroides intracytoplasmic membrane results in a slowing of the electron transfer turnover rate of photochemical reaction centers Biochemistry 50 4819 4829 10.1021/bi101667e 21366273 Woronowicz K Sha D Frese RN Sturgis JN Nanda V Niederman RA 2011b The effects of protein crowding in bacterial photosynthetic membranes on the flow of quinone redox species between the photochemical reaction center and the ubiquinol-cytochrome c2 oxidoreductase Metallomics 3 765 774 10.1039/c1mt00034a 21691621 Xiong J Fischer WM Inoue K Nakahara M Bauer CE 2000 Molecular evidence for the early evolution of photosynthesis Science 289 1724 1730 10.1126/science.289.5485.1724 10976061

PMC005xxxxxx/PMC5001952.txt

==== Front CureusCureus2168-8184Cureus2168-8184Cureus Palo Alto (CA) 10.7759/cureus.711OncologyCorrelation of Antiglobulin Reactivity and Severity of Pancytopenia in a Patient with Hemophagocytic Lymphohistiocytosis: A Case Report and Review of Literature Muacevic Alexander Adler John R Ibrahim Uroosa 1Siddique Muhammad N 1Valecha Gautam 2Asgari Masoud 3Isaac Edhan 4Dhar Meekoo 11 Hematology/Oncology, Staten Island University Hospital 2 Medicine, Staten Island University Hospital 3 Pathology, Staten Island University Hospital 4 Transfusion Medicine, Staten Island University Hospital Uroosa Ibrahim uroosaibrahim@gmail.com26 7 2016 7 2016 8 7 e7119 3 2016 26 7 2016 Copyright © 2016, Ibrahim et al.2016Ibrahim et al.This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.This article is available from http://www.cureus.com/articles/4303-correlation-of-antiglobulin-reactivity-and-severity-of-pancytopenia-in-a-patient-with-hemophagocytic-lymphohistiocytosis-a-case-report-and-review-of-literatureA 46‑year‑old obese male with a medical history of thalassemia minor presented to the emergency room with complaints of severe fatigue and jaundice worsening over two weeks. On further evaluation, the patient was found to have significant hyperbilirubinemia and transaminitis. The hospital course was further complicated by pancytopenia requiring multiple transfusions, worsening hyperbilirubinemia, severe hyperferritinemia, hypofibrinogenemia, and hypertriglyceridemia. He was also found to have splenomegaly and evidence of hemophagocytosis on bone marrow biopsy. On further testing, the patient was also found to have evidence of hemolysis along with a positive direct Coomb's test consistent with autoimmune hemolytic anemia (AIHA), and elevated soluble IL-2 receptor level. The patient was subsequently diagnosed with hemophagocytic lymphohistiocytosis (HLH). He was treated with HLH-94 protocol along with rituximab for AIHA which resulted in improvement of patient's condition. We present a case of HLH with no prior history of autoimmune disease, associated with Coomb's positive AIHA that resolved after therapy for HLH. Our case also delineates how the intensity of antiglobulin reactivity, if present, may correlate with severity of the disease, its progression, and response to treatment. hemophagocytic lymphohistiocytosisautoimmune hemolytic anemiaaggressiverareThe content published in Cureus is the result of clinical experience and/or research by independent individuals or organizations. Cureus is not responsible for the scientific accuracy or reliability of data or conclusions published herein. All content published within Cureus is intended only for educational, research and reference purposes. Additionally, articles published within Cureus should not be deemed a suitable substitute for the advice of a qualified health care professional. Do not disregard or avoid professional medical advice due to content published within Cureus. ==== Body Introduction Hemophagocytic lymphohistiocytosis is a rare hematologic disorder resulting from the disruption of immune homeostasis. It is an aggressive and life-threatening condition characterized by a severe inflammatory state driven by excessive cytokine production. If not treated promptly, HLH can be rapidly fatal. However, due to the rarity of this condition and variable clinical presentation, the diagnosis is often delayed, resulting in poor outcomes. HLH can occur as a sporadic or familial disorder. It can be triggered by a variety of events that disrupt immune regulation. Commonly, infection is one such trigger. A possible role of autoimmune hemolysis in erythrocytopenia encountered in HLH is not formally described in the literature. We present a case of HLH with antiglobulin reactivity correlating with the disease activity and severity of pancytopenia. Case presentation A 46‑year‑old obese male with a medical history of thalassemia minor presented to the emergency room with complaints of severe fatigue and jaundice worsening over two weeks. He denied having any abdominal pain, change in bowel habits, melena, or rectal bleeding. He did not have any of the following: fever, chills, weight loss or night sweats. There was no history of recent travel, toxic exposure, blood transfusions or sick contacts. He was not on any medications or herbal preparations and denied smoking, heavy alcohol use or illicit drug use. His past surgical history included a remote laparoscopic cholecystectomy and Roux-En-Y gastric bypass surgery. The patient worked as an emergency medical services personnel, and his family history was significant for Cooley's anemia. On physical examination, the patient appeared fatigued. Skin and sclerae were significantly icteric. No mucocutaneous lesions were noted. No lymphadenopathy was appreciated. Abdomen was soft with no palpable organomegaly. He was afebrile and hemodynamically stable. The rest of the physical examination was unremarkable. On admission, laboratory studies revealed a total bilirubin of 15.9 mg/dL; direct bilirubin being 1.6 mg/dL. Serum alkaline phosphatase level was within the normal range. Serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST) were both slightly elevated at 63 mg/dL and 44 mg/dL, respectively. Hemoglobin (Hb) was 10.6 g/dL with a mean corpuscular volume (MCV) of 65.5 fL, white blood cell (WBC) count of 7.8 TH/mm3 and platelet count of 309 TH/mm3. Coagulation studies were normal. Blood Urea Nitrogen (BUN) was 27 mg/dL, and serum creatinine was 0.61 mg/dL. Serum iron was 319 ug/dL, total iron binding capacity (TIBC) was 319 ug/dL, percentage saturation was 100%, ferritin was 1267 ng/ml, folate was 4.6 ng/ml and vitamin B12 level was 379 pg/ml. Serum protein electrophoresis (SPEP) was negative for any monoclonal protein. Lactate dehydrogenase (LDH) was high at 545 IU/L, but, reticulocyte count was only 0.79%, and serum haptoglobin was 82 mg/dL. A direct antiglobulin test performed at our facility was negative. A blood sample was sent out to American Red Cross Blood services Southern California region for confirmation. We later learned that when the test was done with standard tube technique, it was indeed negative for anti-IgG, but turned out to be positive with Cold LISS wash technique. A computed tomography (CT) scan of the abdomen revealed an enlarged spleen measuring 20 cm. Notably, the scan did not show any evidence of retroperitoneal hemorrhage, any intrahepatic or extrahepatic biliary duct dilatation, or any abdominal lymphadenopathy. Thoracic lymphadenopathy or any other lung pathology was also ruled out with a CT scan of the chest.    While the diagnostic workup was being done, the patient’s anemia worsened. He was poorly responsive to transfusions and required multiple transfusions daily. His Hb dropped down to 4.1 g/dL at one point. Other blood counts showed a decline too. The lowest WBC count was 3.48 TH/mm3, and platelet count was 80 TH/mm3. Total bilirubin started rising. At one point, it reached 71.9 mg/dL. Alkaline phosphatase remained normal, and a repeat abdominal CT scan ruled out any surgical cause of jaundice. Plasmapheresis was initiated, and the total bilirubin levels decreased to 23.8 mg/dL. Although the reticulocyte count hovered between 0.5 to 1% during this time, serum haptoglobin decreased to less than 3 mg/dL, and serum LDH rose up to 1558 IU/L. At this point, a repeat direct antiglobulin test was sent. We later learned that the test was weakly positive with standard tube technique, but strongly positive with Cold LISS wash technique. A repeat ferritin showed a significant rise to 16,693 mg/dL. Extensive workup ensued to explain the hyperferritinemia, extreme hyperbilirubinemia, and pancytopenia. Viral hepatitis panel, anti-smooth muscle antibody (SMA) and anti-mitochondrial antibody were negative. A liver biopsy revealed cholestasis and suspected hemochromatosis. HFE gene mutation analysis revealed a heterozygous H63D mutation and was negative for the C282Y mutation. Epstein-Barr virus (EBV) serology as well as DNA PCR, parvovirus IgM, IgG antibodies, and HIV expedited assay were all negative. Anti-cytomegalovirus (CMV) IgG antibody was positive, but the PCR study was negative. A peripheral blood flow cytometry ruled out PNH clone. Bone marrow aspiration and biopsy revealed normocellular bone marrow with trilineage hematopoiesis and erythroid hyperplasia without excessive blasts. Flow cytometry was negative for any evidence of lymphoma or leukemia. Notably, the pathologist noted a few foci of erythrophagocytosis. A bone marrow biopsy was repeated a week later, and it confirmed the same findings. At this point, a diagnosis of the hemophagocytic syndrome was contemplated. The suspicion strengthened when serum triglyceride was found to be elevated at 338 mg/dL. Fibrinogen level was 671 mg/dL. At this point, a blood sample for soluble IL-2 receptor level was sent. However, the patient had already fulfilled the minimal requirement of meeting five out of the eight diagnostic criteria for HLH (see Table 1). As we could not identify any triggering condition and the patient was acutely ill, we decided to treat the patient with the HLH-94 protocol. CNS involvement was ruled out by an MRI of the brain and two lumbar punctures.  Table 1 Hemophagocytic lymphohistiocytosis diagnostic criteria Diagnostic Criteria for HLH Used in HLH - 2004 Trial 1. Molecular diagnosis consistent with HLH (e.g. pathologic mutations of PRF1, UNC13D, STX11, SH2D1A, LYST, ITK, SLC7A7, XMEN, HPS or BIRC4) OR 2. Five of the following eight criteria: I). Fever ≥ 38.5⁰C II). Splenomegaly III). Cytopenias (affecting at least 2 of 3 lineages in peripheral blood) a). Hb < 9 g/dL (in infants < 4 weeks: Hb < 10 g/dL) b). Platelets < 100 x 103/mL c). Neutrophils < 1 x 103/mL IV). Hypertriglyceridemia (fasting > 265 mg/dL) and/or hypofibrinogenemia (< 150 mg/dL) V). Hemophagocytosis in bone marrow, spleen, lymph nodes or liver VI). Low or absent NK cell activity VII). Ferritin > 500 ng/mL VIII). Elevated soluble CD25 (that is, soluble IL-2 receptor alpha) 2 standard deviations above age-adjusted laboratory specific norms Subsequently, the results of the first Coombs test sent out to California were reported positive. Thus, a component of his anemia and hyperbilirubinemia could now be attributed to AIHA. We proceeded concurrently with rituximab and HLH‑94 protocol. Altogether, the patient received four doses of rituximab. Dexamethasone was started at 10 mg/m2, and etoposide at 75 mg/m2 with a 50% reduction in view of marked hyperbilirubinemia. The patient was monitored in intensive care unit (ICU). As per the protocol, the patient should have received etoposide twice a week for the first two weeks of treatment, but due to chemotherapy-related cytopenias, he ended up receiving chemotherapy only once a week. During this period, the patient was maintained on supportive care with transfusions as needed. The patient’s soluble IL-2 receptor was elevated at 1517 U/mL. We also sent his peripheral blood sample to a reference laboratory in Cincinnati Children's Hospital, Ohio for further specialized studies. X-linked Inhibitor of Apoptosis (XIAP) level and Natural killer (NK) cell function were both decreased. XIAP levels in CD4, CD8 and CD56 cells were: 75% (normal >= 92%), 78% (normal >= 93%) and 78% (normal >= 94%), respectively. NK Cell function was 54% (normal range 47 - 95%). We later found that he did not have any mutations with known association to the hemophagocytic syndrome, such as STX11, Munc13-4, STXBP2, RAB27A, or PRF1 mutation. SH2D1A mutation and BIRC4 mutation associated with X-linked lymphoproliferative diseases were also negative. A total of six doses of etoposide were administered during the hospital stay which progressively led to decreased transfusion requirements and eventually rendered him transfusion independent, with Hb stabilized at around 9 g/dL and other counts normalized. His aminotransferase levels were improved. Total bilirubin and LDH improved, and stabilized to less than 3 mg/dL and 300 IU/L, respectively.  His ferritin levels had started to show a gradual decline and reached 4666 mg/dL at this point. A repeat direct antiglobulin test for IgG turned negative with the standard tube method as well as with Cold LISS wash technique. The patient was considered stable for discharge. He was to complete a total of ten doses of etoposide along with dexamethasone that was gradually tapered over about six weeks after the last dose of etoposide. During this time he was on antimicrobial prophylaxis with fluconazole, valacyclovir, and co-trimoxazole that was continued for three months after completing his treatment with etoposide. He tolerated the treatment fairly well apart from steroid-induced hyperglycemia which was managed with insulin. A post-treatment serum soluble IL2 receptor, fibrinogen, and serum triglyceride showed complete normalization. His Hb gradually improved further and returned to his baseline about three months after receiving etoposide. At that time, it was 12.2 g/dL along with normal WBC and platelet counts. The patient’s ferritin, however, remained elevated and ranged between 4500 mg/dL to 5000 mg/dL until about six months after finishing his chemotherapy when it almost halved to 2548 mg/dL. This persistent elevation was at least in part attributed to metabolic dysfunction such as him being obese along with steroid-induced hyperglycemia. Indeed, with better glycemic control, it showed a further decline and the latest ferritin was 1746 mg/dL.  Consent was waived. Discussion HLH is a rare hematologic disorder resulting from excessive immune activation. It is an aggressive condition that results from unchecked proliferation and activation of functionally benign lymphocytes and macrophages due to the dysregulation of their function by poorly functioning cytotoxic lymphocytes and NK cells [1]. The resultant cytokine storm and activated macrophages are responsible for most of the clinical manifestations of this disease including hypertriglyceridemia, hypofibrinogenemia, decreased hematopoiesis, hemophagocytosis and multiorgan failure [2]. If not rapidly treated, HLH can result in significant mortality and morbidity. However, the diagnosis is often delayed because of its variable presentation and rarity. It appears that fewer cases are missed now, at least at the tertiary pediatric centers. Significantly higher when compared to a series from the 1970s that reported an incidence of 1.2 children per million per year, a recent review of HLH cases from large pediatric hospitals in Texas, United States revealed a calculated cross-sectional prevalence of one in 100,000 children [3]. This disorder can be both familial and sporadic. The familial form primarily affects the pediatric population. These patients are genetically predisposed due to well-known mutations such as STX11, Munc13‑4, STXBP2, RAB27A, PRF1, LYST and AP3 mutations. On the other hand, most of the adult cases are sporadic and are triggered by the infectious, inflammatory or malignant process. Therefore, diagnosis in the adults is based on a constellation of clinical criteria (Table 1) [4]. One of the most important clinical issues in HLH is anemia, whose mechanism has not been well-elucidated. It has been postulated to be mainly due to the increased destruction of RBC's via hemophagocytosis by the overactive macrophages [5]. However, as many patients with HLH show decreased bone marrow cellularity, bone marrow failure with reduced red cell production has also been hypothesized, and indeed was mechanistically described recently. CD47 is a transmembrane protein located on hematopoietic stem cells that work in collaboration with signal-regulatory protein alpha (SIRPA) to prevent phagocytosis. Its expression can be down-regulated by hypercytokinemia in HLH resulting in increased hemophagocytosis [6]. Besides this, in vivo murine models have clearly demonstrated the role of IFN-γ in macrophage activation that leads to hemophagocytosis [7]. A possible role of autoimmune hemolysis in the pathogenesis of erythrocytopenia encountered in HLH is not formally described in the literature. It has, nevertheless, been seen in settings where HLH occurs in association with an autoimmune disease, and in association with pregnancy [8-9]. Whether its presence in our patient suggests an independent role of AIHA in the mechanism of anemia in certain cases of HLH, or merely exemplifies another manifestation of immune dysregulation seen in HLH, or represents a red flag that warns about the possibility of an imminently developing immunodysregulatory process such as HLH, is obviously difficult to differentiate. The possibility that it merely reflects an incidental concurrent process is unlikely because of the strong correlation observed between the intensity of direct antiglobulin reactivity and disease severity. In our report, not only was the direct antiglobulin test positive before the full spectrum of clinical characteristics appeared, but also the strength of the test positivity increased as the severity of the disease progressed. This observation strongly suggests the possibility that AIHA represented an integral component in the pathophysiology of HLH in our case, which makes it unique. Additionally, this implies that indeed if such underlying pathophysiologic component is identified in a given case, the role of direct antiglobulin testing deserves consideration for disease monitoring. It is well established that assessing the responsiveness promptly is immensely important as the initiation of an alternative treatment plan is needed for patients whose HLH has not completely resolved by the eighth week. Moreover, HLH is prone to reactivation from triggers of the immune response such as new or persistent underlying infection. Disease severity in HLH is highly correlated with soluble IL-2 receptor levels and with NK-cell function assays [4]. However, these are highly specialized tests that are done in only a few reference laboratories in the United States, and the turn out time is at least two weeks. Given the rapidity with which HLH progresses, waiting for these results is not practical. The retrospective analyses show that highly elevated levels of serum ferritin (more than 10,000 ng/ml) are considered 90% sensitive, and 96% specific for the diagnosis of HLH [10]. The treatment resulting in a rapid and deep decline in ferritin levels indicates decreased mortality [11] as well. Yet, the correlation with disease severity is not perfect. Being an acute phase reactant, it may remain elevated due to a concurrent non-HLH process even in the context of an improving HLH, a phenomenon, our case clearly demonstrates. Conclusions The above-described limitations of the current existing treatment monitoring modalities, on one hand, entail the need for more rigorous basic science work. Such work will allow discovery of new disease markers with rapid turnover times. On the other hand, this course will encourage the physicians to search for existing correlative clinical and basic laboratory parameters. Such correlation may provide clues to disease activity, progression, and response. We now share our clinical experience that plainly rolls out how the intensity of antiglobulin reactivity, if present, may correlate with disease severity, progression and response to treatment.  The authors have declared that no competing interests exist. Human Ethics Consent was obtained by all participants in this study ==== Refs References 1 Characteristic immune abnormalities in hemophagocytic lymphohistiocytosis J Pediatr Hematol Oncol Egeler RM Shapiro R Loechelt B 340 345 18 1996 http://repub.eur.nl/pub/62021/00043426-199611000-00002.pdf 8888739 2 Histiocytic disorders: recent insights into pathophysiology and practical guidelines Biol Blood Marrow Transplant Filipovich A McClain K Grom A 82 89 16 2010 3 Hemophagocytic lymphohistiocytosis in Texas: observations on ethnicity and race Pediatr Blood Cancer Niece JA Rogers ZR Ahmad N 424 428 54 2010 19953651 4 How I treat hemophagocytic lymphohistiocytosis Blood Jordan MB Allen CE Weitzman S 4041 4052 118 2011 21828139 5 HLH- 2004: Diagnostic and therapeutic guidelines for hemophagocytic lymphohistiocytosis Pediatr Blood Cancer Henter JI Horne A Aricó M 124 131 48 2007 16937360 6 Engulfment of hematopoietic stem cells caused by down-regulation of CD47 is critical in the pathogenesis of hemophagocytic lymphohistiocytosis Blood Kuriyama T Takenaka K Kohno K 4058 4067 120 2012 22990013 7 Hemophagocytosis causes a consumptive anemia of inflammation J Exp Med Zoller EE Lykens JE Terrell CE 1203 1214 208 2011 21624938 8 Successful rituximab treatment of refractory hemophagocytic lymphohistiocytosis and autoimmune hemolytic anemia associated with systemic lupus erythematosus Mod Rheumatol So MW Koo BS Kim YJ 855 857 24 2014 24517558 9 Pregnancy-induced hemophagocytic lymphohistiocytosis combined with autoimmune hemolytic anemia J Chin Med Assoc Teng CL Hwang GY Lee BJ 156 159 72 2009 19299225 10 Highly elevated ferritin levels and the diagnosis of hemophagocytic lymphohistiocytosis Pediatr Blood Cancer Allen CE Yu X Kozinetz CA 1227 1235 50 2008 18085676 11 Rate of decline of ferritin in patients with hemophagocytic lymphohistiocytosis as a prognostic variable for mortality Pediatr Blood Cancer Lin TF Ferlic-Stark LL Allen CE 154 155 56 2011 20842751

PMC005xxxxxx/PMC5001953.txt

==== Front CureusCureus2168-8184Cureus2168-8184Cureus Palo Alto (CA) 10.7759/cureus.714Pain ManagementNeurosurgeryOrthopedicsMinimally Invasive Lumbar Pedicle Screw Fixation Using Cortical Bone Trajectory – A Prospective Cohort Study on Postoperative Pain Outcomes Muacevic Alexander Adler John R Chen Yi-Ren 1Deb Sayantan 2Pham Lan 1Singh Harminder 11 Department of Neurosurgery, Stanford University Medical Center 2 Medical School, Stanford University School of Medicine Yi-Ren Chen yirenchen@stanford.edu26 7 2016 7 2016 8 7 e71423 5 2016 26 7 2016 Copyright © 2016, Chen et al.2016Chen et al.This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.This article is available from http://www.cureus.com/articles/4660-minimally-invasive-lumbar-pedicle-screw-fixation-using-cortical-bone-trajectory---a-prospective-cohort-study-on-postoperative-pain-outcomesObjective: Our study aims to evaluate the clinical outcomes of cortical screws in regards to postoperative pain. Background: Pedicle screw fixation is the current mainstay technique for posterior spinal fusion. Over the past decade, a new technique called cortical screw fixation has been developed, which allows for medialized screw placement through stronger cortical bone. There have been several studies that showed either biomechanical equivalence or superiority of cortical screws. However, there is currently only a single study in the literature looking at clinical outcomes of cortical screw fixation in patients who have had no prior spine surgery. Methods: We prospectively looked at the senior author’s patients who underwent cortical versus pedicle lumbar screw fixation surgeries between 2013 and 2015 for lumbar degenerative disease. Eighteen patients underwent cortical screw fixation, and 15 patients underwent traditional pedicle screw fixation. We looked at immediate postoperative pain, changes in short-term pain (six to 12 weeks post-surgery), and changes in long-term pain (six to eight months). All pain outcomes were measured using a visual analog scale ranging from 1 to 10. Mann-Whitney or Kruskal-Wallis tests were used to measure continuous data, and the Fisher Exact test was used to measure categorical data as appropriate. Results: Our results showed that the cortical screw cohort showed a trend towards having less peak postoperative pain (p = 0.09). The average postoperative pain was similar between the two cohorts (p = 0.93). There was also no difference in pain six to 12 weeks after surgery (p = 0.8). However, at six to eight months, the cortical screw cohort had worse pain compared to the pedicle screw cohort (p = 0.02). Conclusions: The cortical screw patients showed a trend towards less peak pain in the short-term (one to three days post-surgery) and more pain in the long-term (six to eight months post-surgery) compared to pedicle screw patients. Both cohorts had a statistically significant reduction in pain levels compared to preoperative pain. More studies are needed to further evaluate postoperative pain, long-term functional outcomes, and fusion rates in patients who undergo cortical screw fixation. cortical screwsoutcomesprospective cohort studypostoperative painThe content published in Cureus is the result of clinical experience and/or research by independent individuals or organizations. Cureus is not responsible for the scientific accuracy or reliability of data or conclusions published herein. All content published within Cureus is intended only for educational, research and reference purposes. Additionally, articles published within Cureus should not be deemed a suitable substitute for the advice of a qualified health care professional. Do not disregard or avoid professional medical advice due to content published within Cureus. ==== Body Introduction Every year, over 120,000 lumbar fusions are performed nationwide for degenerative and traumatic spine conditions [1]. The fusions are mostly done with pedicle screw fixation, the current standard technique for accomplishing posterior spinal fusion, due to its reliable fusion rates and construct stability. However, the technique is invasive and requires significant lateral spinal dissection in order to properly place the screws, resulting in large incisions and long operative times. In keeping with the push to perform more minimally invasive spine surgery, a new technique of spinal instrumentation has been developed whereby screws are placed through a starting point at the junction of the superior articular process and pars. This technique is called cortical screw fixation (Figure 1). Figure 1 Cortical Versus Pedicle Screw Fixation Trajectories (A and B) Traditional pedicle screw trajectories in (A) axial and (B) sagittal views. (C and D) Cortical screw trajectories in (C) axial and (D) sagittal views. Over the past decade, there have been several studies that showed either biomechanical equivalence or superiority of cortical screws compared to pedicle screws [2-11]. One study by Baluch, et al. looked at 17 vertebral levels that underwent quantitative computed tomography (CT) [2]. On one side, cortical screws were placed, and on the other, traditional pedicle screws were placed. Cortical screws demonstrated significantly improved resistance to toggle testing, requiring 184 cycles to reach 2 mm of displacement compared to 102 cycles for the traditional pedicle screws (p = 0.002). Due to the promising biochemical studies and the minimally invasive nature of cortical screws, some surgeons are starting to utilize the technique in lumbar fusions. However, there is currently only one study in the literature looking at the clinical efficacy or outcomes of cortical screws in non-redo patients [12]. In that prospective randomized non-inferiority trial, Lee, et al. showed that cortical screw fixation in posterior lumbar interbody fusion (PLIF) provides similar clinical and radiological outcomes compared to pedicle screw fixation. Most past studies have reported on biomechanical strength and not on clinical outcomes [2-10]. In this study, we focus on postoperative and long-term pain in patients who underwent cortical versus pedicle screw fixation in the lumbar spine. Our goal was to evaluate the hypothesis that cortical screw patients should have less postoperative pain due to the smaller incision, less dissection needed to find entry points, and more intraoperative preservation of muscle attachments (Figure 2). Low back pain is often related to muscular stabilization of the “neutral zone” in the back, and the lumbar multifidus muscles are important stabilizers of this neutral zone. Studies have shown that dysfunction of these muscles, such as after surgery, is associated with increased pain [13].   Figure 2 Soft Tissue Exposure Required for One-Level Lumbar Fixation (A) Traditional pedicle fixation. (B) Cortical screw fixation. Materials and methods Cohort selection We prospectively enrolled patients with lumbar degenerative disease and instability who needed lumbar fusion from 2013 to 2015 at the Santa Clara Valley Medical Center (SCVMC) in San Jose, CA. SCVMC is a level one trauma center and the county hospital wing of Stanford University Medical Center. IRB approval for the prospective data registry was obtained through SCVMC (approval #15-026ER). Informed patient consent was obtained at the time of treatment. All patients underwent pedicle or cortical screw placement by the senior author. Patients who underwent surgery because of trauma or infection, such as osteomyelitis, were excluded. In total, 18 patients underwent cortical screw fixation, and 15 underwent pedicle screw fixation. Most of the patients in the lumbar pedicle screw fixation group had their spinal fixation surgeries before the institution of lumbar cortical trajectory screws at SCVMC. The only patients who had pedicle screw fixation done after the institution of cortical trajectory screws were the ones with small pedicles (< 7 mm diameter) on preoperative CT scans. All patients received the same pain regimen postoperatively, with a morphine patient-controlled analgesia pump for 24 hours, followed by long-acting OxyContin and IV morphine and/or Percocet thereafter during the hospital stay. Patients were discharged home on oral Percocet. Surgical technique Traditional pedicle screws were placed using the technique described by Weinstein, et al. [14]. A standard skin incision was made and lateral muscle dissection was performed to expose the transverse processes. An awl was used to breach the cortex at the lateral facet surface, and a pedicle finder was used to extend the trajectory. A tap was then used and screws placed. Overall, a total of 23 levels were fused in the pedicle group. In addition, arthrodesis was also performed in the lateral gutters over the transverse processes. No interbody grafts were used. The cortical screw starting point is at the lateral aspect of the pars interarticularis and, therefore, requires significantly less lateral muscle dissection. The angulation is medial to lateral, rather than lateral to medial as in the traditional pedicle screw technique. Screws are inserted approximately 10 degrees laterally in the axial plane and 25 degrees cranially in the sagittal plane, although actual angulations are determined by intraoperative fluoroscopy (Figure 1) [4]. Pedicles, 7 mm in diameter, were used as the minimum cut-off in order to safely perform cortical screw fixation without lateral vertebral body breach. A total of 26 levels were fused in the cortical group. In addition, arthrodesis was also performed over the facet joints with a high-speed drill. Outcome measurements We looked at peak pain and average immediate postoperative pain from 24 to 72 hours after surgery. We also looked at changes in short-term pain (six to 12 weeks post-surgery) and changes in long-term pain (six to eight months), when compared to preoperative pain in the two cohorts. All pain outcomes were measured using a visual analog scale ranging from 1 to 10. Statistical analysis Mann-Whitney or Kruskal-Wallis tests were used to analyze the difference in pain outcomes between the two groups and over time, respectively. Categorical data was analyzed using Fisher’s exact test. A p - value of less than 0.05 was considered statistically significant. GraphPad Prism v 6.0 was used to conduct all statistical analyses. Results A total of 36 patients were included in the study. Three were excluded because of trauma and infection being the indications for fusion, with a total of 33 patients remaining. Eighteen patients underwent cortical screw fixation and 15 underwent pedicle screw fixation. Overall, the patient characteristics in the cortical versus the pedicle screw groups were similar (Table 1). Table 1 Patient Characteristics ** indicates statistically significant   Cortical Screw (n = 18) Pedicle Screw (n = 15) P-value Age (mean ± SE) 53.39 ± 1.97 59.2 ± 3.12 0.12 Female (%) 38.9 86.7 0.01** Ever Smoked (%) 50 46.7 1.00 Preoperative Pain (mean ± SE) 7.61 ± 0.36 7.43 ± 0.44 0.79 Age, preoperative pain, smoking status, and comorbidities were all similar, with the exception of gender; the pedicle screw group had more females (86.7%) compared to the cortical screw group (38.9%) (p = 0.01). Degenerative disc (DD) disease was the predominant finding on MRI in both cohorts, and presenting pain symptoms were also similar (Table 2). Table 2 Patient Pathology, Presenting Symptoms, and Levels Fused DD - degenerative disease; NFS - neuroforaminal stenosis; b/l - bilateral Pathology Presenting Pain Symptoms Levels Fused Cortical Screw Series     Central and NFS: L5-S1 spondylosis with b/l pars defect Back pain and bilateral radicular leg pain L5-S1 DD and NFS at L4-5 and L5-S1 Back pain and bilateral radicular leg pain L4-S1 Central stenosis L4-5, b/l NFS L5-S1 Back pain and left radicular leg pain L4-S1 Severe central stenosis with Grade 1 anterolisthesis, b/l NFS Back pain and bilateral radicular leg pain L4-L5 Central stenosis, left NFS - facet hypertrophy + synovial cyst Back pain and left radicular leg pain L4-L5 DD and NFS at L4-5 and L5-S1 Back pain and bilateral radicular leg pain L4-S1 Central stenosis and left NFS L5-S1 Back pain and neurogenic claudication L5-S1 DD and right NFS Back pain and bilateral radicular leg pain L4-S1 DD and left NFS Back pain and left radicular leg pain L4-L5 DD and left NFS - L3-4 and L4-5 Back pain and left radicular leg pain L3-S1 DD and left NFS Back pain and left radicular leg pain L4-L5 Central stenosis, left NFS Back pain and left radicular leg pain L4-L5 DD and central stenosis, L4-5 and L5-S1 Back pain and right radicular leg pain L4-S1 Central stenosis, NFS at L3-4 and L4-5 + synovial cyst Back pain and right radicular leg pain L3-L5 DD and right NFS Back pain and right radicular leg pain L4-S1 DD and left NFS Back pain and bilateral radicular leg pain L4-5 Central stenosis, left NFS Back pain and left radicular leg pain L4-5 L5-S1 spondylosis with b/l NFS Back pain and right radicular leg pain L5-S1 Pedicle Screw Series     DD and central and right NFS L5-S1 Back pain and bilateral radicular leg pain L5-S1 Central stenosis and b/l NFS Back pain and bilateral radicular leg pain L3-L4 Central stenosis and left NFS Back pain and left radicular leg pain L4-S1 Central stenosis and b/l NFS L3-5 Back pain and bilateral radicular leg pain L3-S1 L4-S1 central stenosis and right NFS Low back pain and right radicular leg pain L4-S1 DD and central stenosis Low back pain and neurogenic claudication L2-L3 L3-5 central stenosis and b/l NFS Back pain and bilateral radicular leg pain L3-L5 Central stenosis and b/l NFS, anterolisthesis L4 on L5 Back pain and left radicular leg pain L4-L5 Central stenosis and right NFS Low back pain and right radicular leg pain L4-S1 Central stenosis and b/l NFS Back pain and bilateral radicular leg pain L4-5 DD and central stenosis Back pain and bilateral radicular leg pain L4-S1 L5 on S1 anterolisthesis and b/l NFS Back pain and bilateral radicular leg pain L5-S1 Central stenosis with b/l NFS Back pain and Left radicular leg pain L4-S1 Stage II anterolisthesis L3 on L4 Back pain and neurogenic claudication L3-4 Central stenosis and anterolisthesis L4 on L5 Back pain and bilateral radicular leg pain L4-5 Overall, there was no difference in average (p = 0.93) or peak (p = 0.09) immediate postoperative pain between patients who underwent cortical or pedicle screw fixation, as seen in Figure 3. However, there was a trend towards cortical screw patients having less peak postoperative pain, at a pain score of 7.94 versus 9 (p = 0.09). Figure 3 Immediate Postoperative Pain in Cortical Versus Pedicle Screw Patients There was no difference in postoperative pain at the time of short-term follow-up at six to 12 weeks, with an average pain score of 4.97 in the cortical group compared to 4.93 in the pedicle group (p = 0.8) (Figure 3). However, the cortical screw patients did have more pain at the six to eight-month follow-up, with a pain score of 6.14 compared to 3.8 in the pedicle group (p = 0.02) (Figure 4). Figure 4 Long-Term Postoperative Pain in Cortical Versus Pedicle Screw Patients Both groups had a statistically significant reduction in pain levels compared to preoperative pain, as seen in Table 3. Table 3 Numerical Data and Statistical Analysis of Long-term Postoperative Pain in Cortical Versus Pedicle Screw Patients ** indicates levels of significance   Preop 6-12 weeks 6-8 months p-value (Kruskal-Wallis) over time within group Multiple comparisons Mean Pain Cortical Screw (mean ± SE) 7.61 ± 0.36 4.97 ± 0.68 6.14 ± 0.61 0.004 Pre-op vs. 6-12 weeks** Mean Pain Pedicle Screw (mean ± SE) 7.43 ± 0.44 4.93 ± 0.84 3.80 ± 0.75 0.002 Pre-op vs. 6-8 months** p-value Between Groups 0.79 0.80 0.02**   The raw data for all patients in the study are shown in Tables 4-5. Table 4 Cortical Screw Patient Series COPD – chronic obstructive pulmonary disease; DM – diabetes mellitus; HTN – hypertension; HLD – hyperlipidemia; OA – osteoarthritis; CKD – chronic kidney disease; GERD – gastroesophageal reflux disease; NA – Data not available Age Immediate Average Postop Pain Immediate Peak Postop Pain 6-12 Weeks Postop Pain 6-8 Months Postop Pain Ever Smoked Significant Comorbidities 50 5.733 10 7 7 yes Neuropathy 44 6.313 9 6 5 yes COPD 42 3.222 7 4 NA no Chronic neck pain 59 5.5 10 0 5 yes DM, HTN, HLD 59 4.958 10 0 6 yes DM, OA, Sciatica 67 5.375 10 0 6 no HTN, HLD 52 2.875 6 7 8 yes HTN, HLD, DM 57 8.286 10 7 10 no HTN 35 8.5 9 9 8 yes Seizures 50 6.154 9 7 7 no DM, HTN, CKD 56 4.25 8 6 7 no Lumbago 64 2.4 9 4 5 no DM, HTN 58 3.429 6 6 7 yes HTN 44 4 7 5 0 yes HTN 49 4.077 7 4 5 yes DM, HTN 55 3.133 7 4 NA no GERD 59 0 0 8.5 NA no DM, Migraine 61 6 9 NA NA no Fibromyalgia Table 5 Pedicle Screw Patient Series CHF – congestive heart failure; DM – diabetes mellitus; HTN – hypertension; HLD – hyperlipidemia; OA – osteoarthritis; CKD – chronic kidney disease; CAD – coronary artery disease; RA – rheumatoid arthritis; NA – Data not available Age Immediate Average Post-Op Pain Immediate Peak Post-Op Pain 6-12 Weeks Post-Op Pain 6-8 Months Post-Op Pain Ever Smoker Significant Comorbidities 47 7.167 9 9 7 yes HTN, Neuropathy, Lumbago, Seizures 77 5.6 10 7 5 no CHF 56 4.615 10 4 6 yes HTN, HLD, OA 49 6.286 10 9 5 yes HTN, Seizures, Sciatica 57 5.4 10 4 0 no Epilepsy, Sciatica, OA 71 4.364 10 7 9 no OA, DM, HTN, CKD 78 5.222 10 6 3 no DM, HTN, HLD 63 1.7 4 NA 2 no HTN, HLD, OA 63 4.1538 8 0 0 yes CAD 65 4.9444 9 5 4 yes HLD, HTN, DM 38 4.0714 9 5 0 no Lumbar stenosis 62 5.375 10 0 0 yes DM, HTN, HLD CHF 39 5 10 8 7 no DM 65 0.81 7 0 4 no RA 58 NA NA 5 5 yes HTN, OA Discussion Multiple biomechanical studies have shown equivalence or superiority of the cortical bone trajectory compared to the standard technique for pedicle screw fixation [2-3, 5-10]. Most studies show that cortical screws traverse denser cortical bone and, thus, result in increased pullout strength and improved rigidity. This has theoretical advantages for patients with osteoporosis, failed fusion requiring reoperation, and even first-time fusions for degenerative disease and spinal instability. However, there is only one study looking at the efficacy and clinical outcomes of cortical screw fixation in previously non-instrumented patients. In their prospective randomized trial, Lee, et al. showed that cortical screw fixation with interbody fusion provides comparable pain reduction and fusion rates to that of pedicle screw fixation with interbody fusion. Another study in the literature that is non-cadaveric is by Rodriguez, et al. [15]. The authors retrospectively reviewed five patients who underwent cortical screw fixation and posterior interbody grafting for adjacent segment lumbar disease. All cases were reoperations after prior lumbar instrumentation. The average age of the patients was 69.4, and all five patients reported improved low back pain compared with preoperative pain at 10- to 15-month follow-up. The authors concluded that cortical screw fixation was a good technique in patients requiring a reoperation because it obviates the need for previous hardware removal. Our paper is the first prospective cohort study looking at cortical versus pedicle screw fixation in patients with lumbar degenerative disease and spinal instability, with or without an interbody fusion. None of the cases had prior instrumentation, as we sought to evaluate the pain outcomes in patients who underwent cortical screw fixation as the first-line technique. Overall, there was no statistically significant difference in either average or peak immediate postoperative pain. However, there was a trend towards the cortical screw patients having less peak immediate postoperative pain (average pain score of 9 versus 7.94 in the pedicle group) (p = 0.09). This is consistent with the hypothesis that a smaller incision with less muscle detachment and soft-tissue dissection in cortical screws may lead to less postoperative pain. This is also consistent with the findings in the Lee, et al. study where cortical screws were associated with lower immediate postoperative pain (within one week of surgery) compared to pedicle screws [12]. Although there was no difference in pain between cortical and pedicle screw patients at the six to 12-week follow-up, pedicle screw patients appeared to have less pain at the six to eight-month follow-up (pain score of 6.14 in cortical patients versus 3.8 in pedicle patients) (p = 0.02). This is an interesting finding, as most of the biomechanical studies have shown superior pullout strength and stability in cortical screws compared to pedicle screws [6]. One explanation for these phenomena might be that, even though cortical screws are biomechanically stronger and prevent spinal micromotion and, therefore, pain generation in the short-term, long-term stability of the construct depends on the formation of a stable fusion mass. This might be happening more effectively in the pedicle screw cohort with posterolateral fusion in the lateral gutters, compared to the cortical screw cohort where arthrodesis was performed over the facet joints, but not in the lateral gutters. Although plain radiographs did not show any hardware failure in either group at the six to eight-month follow-up, long-term follow-up is needed to see if cortical screw patients have higher levels of pseudoarthrosis compared to pedicle screw patients and whether the difference in pain outcomes between the two cohorts remains divergent into the future. Notably, Lee, et al. did not find any difference in the pain scores or fusion rates between their two cohorts at one-year follow-up [12]. However, it has to be noted that they performed an interbody fusion (PLIF) in all of their patients, so their construct stability was boosted by the interbody fusion and was not solely reliant on the lateral gutter formation of a fusion mass in the long-term. There are some limitations to our study, including the small sample size and the lack of randomization. Although patient characteristics overall were similar between the two groups, the pedicle screw group did have a higher percentage of female patients compared to the cortical group (Table 1). This may be the result of females having smaller diameter pedicles, thus, limiting the feasibility of the cortical trajectory technique. A 7-mm diameter was used as the minimum cut-off for performing cortical screw fixation in our study. Placing cortical trajectory lumbar screws in narrow pedicles is extremely challenging because of the medial to lateral course of the screw within the pedicle. The incidence of lateral vertebral breach with cortical screws seen in cadaveric specimens with narrow pedicles was found to be very high in the lab. This difference in the sex ratio between the two cohorts may confound the data, as women may have a different pain threshold compared to men due to mood, sex-role beliefs, or hormonal effects [16]. The study also only included patients from a single center with surgery performed by a single surgeon. Conclusions Our paper is the first prospective cohort study comparing pain outcomes for 33 patients who underwent cortical versus pedicle screw fixation in lumbar degenerative disease, with or without interbody fusion. Overall, cortical screw fixation results in similar to improved immediate postoperative pain but showed a trend towards worsening low back pain at the six to eight-month follow-up compared to pedicle screw patients. Both cohorts had statistically significant reduction in pain levels compared to preoperative pain after surgery (cortical at six to 12 weeks and pedicle at six to eight months). Our paper is a pilot study, and more prospective randomized clinical studies are needed to further evaluate postoperative pain, long-term functional outcomes, and fusion rates in patients who undergo cortical screw fixation. The authors have declared that no competing interests exist. Human Ethics Santa Clara Valley Medical Center (SCVMC) Institutional Review Board issued approval #15-026ER Animal Ethics Animal subjects: This study did not involve animal subjects or tissue. Yi-Ren Chen and Sayantan Deb contributed equally to this article. We thank Cheryl J. Christensen and Cindy H Samos for assistance with the manuscript and Paul Schiffmacher for the artwork in Figures 1 and 2. ==== Refs References 1 United States trends in lumbar fusion surgery for degenerative conditions Spine (Phila Pa 1976) Deyo RA Gray DT Kreuter W Mirza S Martin BI 1441 1445 30 2005 15959375 2 Effect of physiological loads on cortical and traditional pedicle screw fixation Spine (Phila Pa 1976) Baluch DA Patel AA Lullo B Havey RM Voronov LI Nguyen NL Carandang G Ghanayem AJ Patwardhan AG 0 302 39 2014 3 Cortical screws used to rescue failed lumbar pedicle screw construct: a biomechanical analysis J Neurosurg Spine Calvert GC Lawrence BD Abtahi AM Bachus KN Brodke DS 166 172 22 2015 25478820 4 Biomechanical evaluation of the fixation strength of lumbar pedicle screws using cortical bone trajectory: a finite element study J Neurosurg Spine Matsukawa K Yato Y Imabayashi H Hosogane N Asazuma T Nemoto K 471 478 23 2015 26161515 5 The "medio-latero-superior trajectory technique": an alternative cortical trajectory for pedicle fixation Orthop Surg Mobbs RJ 56 59 5 2013 23420749 6 Pedicle screw insertion angle and pullout strength: comparison of 2 proposed strategies J Neurosurg Spine Inceoğlu S Montgomery WH Jr St Clair S McLain RF 670 676 14 2011 21388287 7 Cortical bone trajectory for lumbar pedicle screws Spine J Santoni BG Hynes RA McGilvray KC Rodriguez-Canessa G Lyons AS Henson MA Womack WJ Puttlitz CM 366 373 9 2009 18790684 8 Pedicle screw placement in the lumbar spine: effect of trajectory and screw design on acute biomechanical purchase J Neurosurg Spine Wray S Mimran R Vadapalli S Shetye SS McGilvray KC Puttlitz CM 503 510 22 2015 25679236 9 Should we use cortical bone screws for cortical bone trajectory? J Neurosurg Spine Ueno M Sakai R Tanaka K Inoue G Uchida K Imura T Saito W Nakazawa T Takahira N Mabuchi K Takaso M 416 421 22 2015 25594731 10 In vivo analysis of insertional torque during pedicle screwing using cortical bone trajectory technique Spine (Phila Pa 1976) Matsukawa K Yato Y Kato T Imabayashi H Asazuma T Nemoto K 0 45 39 2014 11 Biomechanical fixation properties of cortical versus transpedicular screws in the osteoporotic lumbar spine: an in vitro human cadaveric model J Neurosurg Spine Sansur CA Caffes NM Ibrahimi DM Pratt NL Lewis EM Murgatroyd AA Cunningham BW 1 10 (Epub ahead of print) 2016 27176113 12 The comparison of pedicle screw and cortical screw in posterior lumbar interbody fusion: a prospective randomized noninferiority trial Spine J Lee GW Son JH Ahn MW Kim HJ Yeom JS 1519 1526 15 2015 25728553 13 The role of the lumbar multifidus in chronic low back pain: a review PM R Freeman MD Woodham MA Woodham AW 142 146 2 2010 20193941 14 Spinal pedicle fixation: reliability and validity of roentgenogram-based assessment and surgical factors on successful screw placement Spine (Phila Pa 1976) Weinstein JN Spratt KF Spengler D Brick C Reid S 1012 1018 13 1988 http://journals.lww.com/spinejournal/Abstract/1988/09000/Spinal_Pedicle_Fixation__Reliability_and_Validity.8.aspx7 3206294 15 Novel placement of cortical bone trajectory screws in previously instrumented pedicles for adjacent-segment lumbar disease using CT image-guided navigation. Technical note Neurosurg Focus Rodriguez A Neal MT Liu A Somasundaram A Hsu W Branch CL Jr 0 36 2014 16 Sex, gender, and pain: women and men really are different Curr Rev Pain Fillingim RB 24 30 4 2000 10998712

PMC005xxxxxx/PMC5001954.txt

==== Front CureusCureus2168-8184Cureus2168-8184Cureus Palo Alto (CA) 10.7759/cureus.716Infectious DiseaseGeneral SurgeryPrimary Hydatid Cyst of the Small Intestine: A Rare Case Report and Brief Review of the Literature Muacevic Alexander Adler John R Ertekin Suleyman Caglar 1Ozmen Tolga 11 General Surgery, Marmara University Hospital Suleyman Caglar Ertekin ozmens@yahoo.com28 7 2016 7 2016 8 7 e71617 6 2016 28 7 2016 Copyright © 2016, Ertekin et al.2016Ertekin et al.This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.This article is available from http://www.cureus.com/articles/4650-primary-hydatid-cyst-of-the-small-intestine-a-rare-case-report-and-brief-review-of-the-literatureHydatid disease is an endemic disease especially in underdeveloped and developing countries affecting mostly the liver and lungs. The hydatid cysts located in other sites are mostly due to rupture of primary liver or splenic cysts. We present a primary small intestine hydatid cyst resected laparoscopically with the affected intestinal segment. As far as we know, this is the first report of a primary small intestine hydatid disease in the literature. hydatid diseasesmall intestinesurgerylaparoscopyThe content published in Cureus is the result of clinical experience and/or research by independent individuals or organizations. Cureus is not responsible for the scientific accuracy or reliability of data or conclusions published herein. All content published within Cureus is intended only for educational, research and reference purposes. Additionally, articles published within Cureus should not be deemed a suitable substitute for the advice of a qualified health care professional. Do not disregard or avoid professional medical advice due to content published within Cureus. ==== Body Introduction Hydatid cysts may be found in almost any site of the body, either from primary inoculation or via secondary spread. The liver (66%) and lung (25%) are the most commonly affected sites, but other organs (e.g. spleen, brain, muscle, kidneys, adrenal glands, bone, heart, pancreas) can also be affected [1-2]. Mesenteric hydatid disease can occur due to iatrogenic rupture of visceral (i.e., liver, spleen) hydatid cysts. Primary intestinal hydatid disease is exceptional. Patients who have an extrahepatic hydatid cyst present mostly with abdominal pain and discomfort. Diagnosis can be challenging [3-4]. We present here an unusual case of primary intestinal hydatid cyst and a review of the literature. Case presentation A 31-year-old woman presented with a palpable mass on the left upper quadrant of the abdomen. She had a 14-month history of intermittent abdominal pain attacks, discomfort, and early satiety. The patient's medical history and family history were unrevealing. Abdominal examination revealed a palpable, partially mobile mass at the left hypochondriac region. Laboratory findings were unremarkable. Hydatid serology (IHA) was negative. Abdominal computed tomography (CT) scans revealed a heterogeneous mass originated from the stomach wall, measuring 4.1 x 4.3 cm (Figures 1-2). No pathological appearance was detected in the chest x-ray. Upper GI endoscopy was also unrevealing. Figure 1 The Lesion (arrow) Appearing on Computed Tomography (CT) Figure 2 The Mass (arrow) in Coronal CT Plane A preliminary diagnosis of a gastrointestinal stromal tumor (GIST) originating from the stomach was made. During diagnostic laparoscopy, the stomach was intact, but we found a 5 cm mass located at the jejunum, 30 cm distally to the Treitz ligament. The affected jejunum segment was resected 5 cm proximally and 5 cm distally with a stapling device. The mass was taken out of the abdomen inside a specimen retrieval bag. The two jejunum ends were anastomosed. The postoperative period was uneventful and the patient was externalized on day three. The pathology report revealed a 5.5 x 5 x 4 cm mass on the serosal aspect of the small intestine. The mass was rounded, had regular borders, and a white color. There was no association between the mucosa of the intestine and the mass. The inside cavity of the mass was filled with cheese-like necrotic tissue. There was no immune expression on CD117, DOG1, SMA, S100, and Ki-67 dying of the fibrotic capsule of the cystic lesion. Around the cystic lesion, periodic acid-Schiff (PAS) staining positive for cuticular membrane was detected. In the fibrotic tissue surrounding the cystic lesion, collagenation was detected with Masson's trichromatic dye. These histological and morphological findings supported a diagnosis of a hydatid cyst. After the pathological report of this rare clinical entity became absolute, informed consent was gained from the patient to publish these findings. A three-cycle albendazole treatment was ordered to the patient. In each cycle, the patient took 400 mg albendazole twice a day for 28 days and stopped for 14 days. No recurrence was noted during the 12-month follow-up period. Discussion Hydatid disease, caused by the Echinococcus granulosus is an endemic disease especially in regions like Eurasia and South American countries [5]. The small intestine is an unusual site for a hydatid cyst, and there are a few primary peritoneal hydatid cyst case reports in the literature [5]. The mechanism of infestation is not clear. For intestinal mesenteric hydatid cysts, dissemination via lymphatic or systemic circulation has been implicated as a possible route, and we think this might also be the mechanism in our case  [6]. Extrahepatic hydatid disease usually remains asymptomatic for years. Patients mostly present after the cyst becomes large enough to palpate or to cause non-specific symptoms as abdominal discomfort. The combination of clinical, laboratory, and radiological findings help for a preliminary diagnosis. Among all imaging modalities, ultrasonography is superior to both MRI and CT in visualization and evaluation of the morphology of liver cysts and hydatid disease [7]. In the literature, very high sensitivity (88% to 98%) and specificity (93% to 100%) rates are given for ultrasonography in the diagnosis of hydatid disease [7]. Abdominal cystic lesions (i.e., mesenteric cysts, ovarian cysts, lymphangioma) must be considered in the differential diagnosis [6]. In our routine practice, the workup of a patient with a mass located in the abdomen starts with a CT scan. Since the CT scan reported a non-cystic mass originating from the stomach wall, hydatid disease was not considered in the differential diagnosis, and we did not proceed with any other imaging modality. Commonly used serological methods in the diagnosis of hydatid disease are the enzyme-linked immunosorbent assay (ELISA), the indirect hemagglutination test (IHA), the latex agglutination test and immunoblots [7]. Nevertheless, the usage of these tests still remains controversial due to inadequate sensitivity and specificity rates. In the literature, a wide range of sensitivity rates (50% to 100%) is given for the IHA test. The specificity rate is also reported to be 83% to 88% [7]. Studies suggest that combining the ELISA test and the IHA test increases the sensitivity up to 94.7% [8]. In our case, the IHA test was done in another clinic before admission to our hospital. Since it was negative and the imaging study was also not supporting a diagnosis of hydatid disease, no further serological test was done for this patient. There are some arguments given to explain false-negative IHA results. It is argued that only 60% to 80% of hydatid disease patients become seropositive. It is also argued that patients with a cystic lesion less than 9 cm diameter or a cystic lesion that is solitary, extrahepatic, unilocular, or degenerative are more prone to a false-negative IHA result [7]. Since in our case the lesion was 5 cm in diameter, solitary, and extrahepatic; these might be the reasons of serological false negativity. The gold standard treatment for hydatid disease is complete surgical excision though according to the site of origin, partial or subtotal cystectomy can be performed to avoid adjacent organ injuries [9]. In our case we had a preliminary diagnosis of GIST, so we resected the affected intestinal segment and took the specimen out in a specimen retrieval bag. Mebendazole or albendazole is given to the patient adjuvantly to prevent recurrence [4]. In our case, we preferred albendazole treatment, and there was no recurrence in the 12-month follow-up period. As far as our knowledge, this is the first case report of a primary small intestine hydatid cyst resected laparoscopically.  Conclusions Hydatid disease is a significant public health problem in underdeveloped and developing countries affecting mostly the liver and lungs. Primary small intestine hydatid disease is a very rare clinical entity, which should be kept in mind for patients with an intra-abdominal mass. The authors have declared that no competing interests exist. Human Ethics Consent was obtained by all participants in this study ==== Refs References 1 Long-term outcome of asymptomatic liver hydatidosis J Hepatol Frider B Larrieu E Odriozola M 228 231 30 1999 10068100 2 Complications of extrahepatic echinococcosis: fistulization of an adrenal hydatid cyst into the intestine World J Gastroenterol Ruiz-Rabelo JF Gomez-Alvarez M Sanchez-Rodriguez J Rufian Peña S 1467 1469 7 14 2008 http://www.wjgnet.com/1007-9327/full/v14/i9/1467.htm 18322969 3 Biological, epidemiological, and clinical aspects of echinococcosis, a zoonosis of increasing concern Clin Microbiol Rev Eckert J Deplazes P 107 135 17 2004 14726458 4 Isolated primary hydatid cyst of small intestinal mesentery: an exceptional location of hydatid disease Pan Afr Med J Najih M Chabni A Attoulou G 17 13 2012 http://www.panafrican-med-journal.com/content/article/13/17/full/ 23308322 5 Diagnosing and staging of cystic echinococcosis: how do CT and MRI perform in comparison to ultrasound? PLoS Negl Trop Dis Stojkovic M Rosenberger K Kauczor HU Junghanss T Hosch W 0 6 2012 6 Primary peritoneal echinococcosis masquerading as an ovarian cyst Indian J Surg Khare DK Bansal R Chaturvedi J Dhasmana JP Gupta S 173 68 2006 http://www.bioline.org.br/pdf?is06052 7 Diagnostics in cystic echinococcosis: serology versus ultrasonography Turk J Gastroenterol Wuestenberg J Gruener B Oeztuerk S 398 404 25 2014 25254522 8 Evaluation of two ELISA and two indirect hemagglutination tests for serodiagnosis of pulmonary hydatid disease Korean J Parasitol Eris FN Akisu C Aksoy U 427 429 47 2009 19967097 9 Isolated pelvic hydatid cyst: one pediatric case Arch Pediatr Tajdine MT Daali M 1367 1368 Nov 14 2007 17931841 10 Liver cystic echinococcosis: Which cysts are correlated with false negative indirect passive hemagglutination (IHA)? Tunis Med Mzali R Ben Amar M Kallel W Kolsi K Beyrouti MI Ayadi A 367 370 85 2007 17657919

PMC005xxxxxx/PMC5001955.txt

==== Front CureusCureus2168-8184Cureus2168-8184Cureus Palo Alto (CA) 10.7759/cureus.715RadiologyNeurosurgeryMobile Schwannoma of the Lumbar Spine: A Case Report and Review of the Literature Muacevic Alexander Adler John R Toscano Daniel T 1Felbaum Daniel R 1Ryan Joshua E 1Sayah Anousheh 2Nair Mani N 11 Neurosurgery, Medstar Georgetown University Hospital 2 Neuroradiology, Medstar Georgetown University Hospital Daniel R. Felbaum rocky.felbaum@gmail.com27 7 2016 7 2016 8 7 e71511 6 2016 27 7 2016 Copyright © 2016, Toscano et al.2016Toscano et al.This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.This article is available from http://www.cureus.com/articles/4735-mobile-schwannoma-of-the-lumbar-spine-a-case-report-and-review-of-the-literatureMobile schwannomas of the spine have been sparsely documented in the literature. In cases referred to in existing literature, the migratory schwannoma was documented to occur in the lumbar spine. We added another case to the small available literature. In our case report, the patient had a previously known lumbar schwannoma that was being managed conservatively. Due to an acute change in clinical symptoms, repeat imaging was performed. A magnetic resonance imaging (MRI) of his spine revealed migration of the schwannoma two levels rostral to his recent imaging from six weeks earlier. The patient underwent surgical resection of his lesion. During the operation, the ultrasound was utilized to confirm the lesion prior to dural opening. In this report, we attempt to provide further evidence of the utility of an intraoperative ultrasound for intradural lesions and intend to add to the published literature of mobile schwannomas of the spine lumbar radiculopathyschwannomaultrasoundneurosurgical complicationsThe content published in Cureus is the result of clinical experience and/or research by independent individuals or organizations. Cureus is not responsible for the scientific accuracy or reliability of data or conclusions published herein. All content published within Cureus is intended only for educational, research and reference purposes. Additionally, articles published within Cureus should not be deemed a suitable substitute for the advice of a qualified health care professional. Do not disregard or avoid professional medical advice due to content published within Cureus. ==== Body Introduction Although rare, mobile schwannomas have been observed at various vertebral levels. Most of them have been reported within the lumbar spine. As described in the available literature, preoperative imaging confirmed the level of the operative intradural lesion. In several instances, imaging was performed just several hours prior to the operation. In a surgical nightmare, operative findings revealed normal spinal anatomy at the previously determined position [1-4]. Regarding these cases, subsequent exploratory laminectomy and extension of dural opening were not always successful in locating the lesion. When used for an intraoperative imaging, the ultrasound may help prevent an unsuccessful operation by providing real-time confirmation of the intradural location just prior to dural opening [3,5]. We present the case of a migratory lumbar schwannoma and the utility of an intraoperative ultrasound in confirming its location in aiding surgical resection. Case presentation A 40-year-old gentleman presented with a history of progressive lower back pain and circumferential right leg pain. Over a period of one week, the pain developed into higher low back pain along with bilateral lower extremity weakness and increasing difficulty with balance. Previously, a magnetic resonance imaging (MRI) of his thoracolumbar spine demonstrated an intradural, extramedullary mass in the T11 to T12 intradural region. A repeat MRI was performed due to the new onset bladder incontinence. Upon further review of his most recent imaging, the lesion was located at T12-L1, no longer as high as T11 (Figure 1). Figure 1 Initial and preoperative MRI imaging Initial sagittal MRI with gadolinium demonstrated an intradural extramedullary mass in the T11-T12 region. Immediate preoperative imaging obtained after new onset bladder incontinence demonstrated tumor movement caudally, down to the T12-L1 region. Informed consent was acquired for the study. The patient was then taken to the operating room for surgical excision of the lesion. An intraoperative ultrasound confirmed the intradural location and the resection went without untoward events (Figure 2). Figure 2 Intraoperative ultrasound An intraoperative ultrasound confirmed the location of the intradural lesion just prior to the opening of the dura. Pathology was consistent with a schwannoma. Ultimately, the patient had an unremarkable hospital stay and was discharged to an acute rehabilitation facility. Discussion Mobile schwannomas of the spine are rare with only 18 computer tomography (CT)/MRI confirmed cases reported to date, this one included. Table 1 displays all reported cases of mobile spinal schwannomas diagnosed using CT and MRI [1-10]. Table 1 Summary of Published Cases of Mobile Spinal Schwannomas Summary of availalbe literature regarding mobile schwannomas of the spine Author/Year Age/Sex Initial Location Final Location Discrepancy (Vertebrae) Migration Imaging Modality Additional durotomy and/or lengthening of original incision? Isu et al [6] 1989 51F T11-T12 T12-L1 1 Caudal MRI Unknown 42M T12 L1 1 Caudal MRI Unknown 52M L1 L1-L2 1/2 Caudal MRI Unknown Namura et al [7] 1993 51M T4-T5 T9-T10 5 Caudal Myelography, MRI No Varughese et al [2] 1997 65M L5 L4 1 Rostral Myelography, CT Unknown 78M L3 L2 1 Rostral Myelography, CT Unknown Iizuka et al [8] 1998 48M C7-T1 T1-T2 1 Caudal MRI, intraop-Myelography Unknown Friedman et al [9] 2003 28M L4 L3 1 Rostral MRI, intraop-US No 33M L5 L4-L5 1/2 Rostral MRI, intraop-US No 41F L2 L1-L2 1/2 Rostral MRI, intraop-US No Marin-Sanabria et al [3] 2007 27M L1 L1-L2 1/2 Caudal MRI, intraop-MRI Yes, L1 \begin{document}\rightarrow\end{document} T12-L2; L2* 41M L2-L3 L1-L2 1 Rostral MRI, intraop-MRI Yes, L2-L3 \begin{document}\rightarrow\end{document} L2-L4; L1-L2* Kim et al [4] 2009 45M L3-L4 L2-L3 1 Rostral MRI No 32M T10-T11 T11 1/2 Rostral MRI Yes, T10 \begin{document}\rightarrow\end{document} T10-T11 27M L3-L4 L2-L3 1 Rostral MRI Yes, L3-L4; L2* Khan et al [1] 2013 52M T10-T11 T7-T8 3 Rostral MRI Yes, T9-T11 \begin{document}\rightarrow\end{document} T9-T12; T7-T12* Terada et al [10] 2016 68M C5-C7 C6-T1 1 Caudal MRI, intraop-US No Toscano et al, 2016 40M T11-T12 T12-L1 1 Caudal MRI, intraop-US No *Second Surgery When comparing preoperative tumor location and final tumor position upon excision, each case demonstrates a significant change in the vertebral level of the lesion. Primary factors driving tumor movement include postural adjustments and arrangement on the operating table as well as positive pressure respiration and other actions that influence intrathecal, intrathoracic, or intra-abdominal pressures such as the Valsalva maneuver [2, 3, 4, 7, 9-10]. Surgical and imaging procedures such as laminectomy and contrast injection also affect tumor position [3]. Also, tumors originating in the cauda equina may exhibit heightened positional freedom due to an increase in lumbar nerve root length and the absence of a solid cord structure which is present in cervical and thoracic schwannomas [4, 10]. In our case, this may have contributed to the lesion’s mobility observed on imaging. The enhanced risk of surgically removing a mobile schwannoma comes not from dural attachment as almost all cases reported no dural adhesion. Instead, they arose from an unpredictable and often substantial movement of these lesions up until the moment of excision [4, 9]. With their patient post L1-L3 laminectomy and just prior to dural opening, Varughese et al. reported observing a mobile schwannoma oscillating between L1 and L3 in step with each respiration [2]. As dural incision is necessary to remove intradural mobile schwannomas, it is important to confirm tumor location intraoperatively before entering the subdural space as the tumor may have shifted since preoperative imaging. Several cases reported finding no lesion after initial dural incision at the preoperatively determined location thus necessitating further exploratory laminectomy and durotomy [3, 4]. An intraoperative myelography has in the past, and can be used to verify tumor-positon before durotomy. However, the injection of a contrast material may itself induce tumor mobility [4]. In two separate cases, Marin-Sanabria et al. reported the use of an intraoperative MR imaging to determine tumor location after initial laminectomy through the dural opening was unsuccessful. It is also worth noting that access to intraoperative MR may not be feasible in many operating suites [3]. An intraoperative ultrasound has proven very effective in tumor localization before entering the dural space and does not significantly increase surgical risk or procedure length [9]. Additionally, an intraoperative ultrasound has inherent benefits over an intraoperative myelography and an MR imaging. The reason being, the ultrasound procedure does not introduce variations of intrathecal pressure, and anatomical changes can be observed in real time. In the event of tumor mobility, the ultrasound may help to determine precisely the levels of lamina that need to be removed. When considering the cases where the ultrasound was used, this technique is successful in confirming or appropriately modifying the site of the dural incision. As observed in our particular case, schwannomas involving the cauda equina may have more variability and potential to migrate. An intraoperative ultrasound was essential in minimizing morbidity by confirming the location of the lesion prior to dural opening. In further operations involving intradural lesions, schwannomas of the cauda equina in particular, an intraoperative ultrasound should be utilized. Conclusions Migratory schwannomas of the spine are rarely reported in the literature. We add to the available literature another case of a mobile lumbar schwannoma. Due to its location at the cauda equina, a schwannoma in this region may be more prone to migration. An intraoperative ultrasound is essential in confirming operative location prior to dural opening. Also, we present a brief review of the available literature regarding this unusual phenomenon. The authors have declared that no competing interests exist. Human Ethics Consent was obtained by all participants in this study ==== Refs References 1 Double migration of a schwannoma of thoracic spine BMJ Case Reports Khan RA Rahman A Bhandari PB 2013 2013 2 Mobile tumours in the lumbar spinal canal: a diagnostic problem Canadian Journal of Surgery Varughese G Mazagri R 59 62 40 1997 http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3949883/ 3 Mobile cauda equina schwannomas Singapore Med J Marin-Sanabria EA Sih IM Tan KK 53 56 48 2007 http://smj.sma.org.sg/sma_web/smj/4802/4802cr7.pdf 4 Mobility of intradural extramedullary schwannoma at spine : report of three Cases with literature review Journal of Korean Neurosurgical Society Kim S-B Kim H-S Jang J-S 64 67 47 2010 20157382 5 Migration of an intraspinal schwannoma documented by intraoperative ultrasound: case report Surgical Neurology Friedman JA Atkinson JLD Lane JI 455 457 54 2000 http://www.researchgate.net/publication/12092241_Migration_of_an_intraspinal_schwannoma_documented_by_intraoperative_ultrasound_Case_report 11240178 6 Mobile schwannoma of the cauda equina diagnosed by magnetic resonance imaging Neurosurgery Isu T Iwasaki Y Akino M 968 971 25 1989 http://journals.lww.com/neurosurgery/Abstract/1989/12000/Mobile_schwannoma_of_the_cauda_equina_diagnosed_by.20.aspx 2601828 7 Thoracic mobile neurinoma Journal of Neurosurgery Namura S Hanakita J Suwa Hideyuki 277 279 79 1993 http://thejns.org/doi/abs/10.3171/jns.1993.79.2.0277 8331413 8 Mobile neurinoma of the cervicothoracic junction Surgical Neurology Iizuka H Iida T Akai T 492 493 50 1998 9842879 9 Utility of intraoperative ultrasound for tumors of the cauda equina Spine Friedman JA Wetjen NM Atkinson D 288 290 28 2003 12567033 10 A mobile schwannoma of the cervical spinal cord: case report and review of the literature Neurosurgery Terada Y Toda H Yokote A 156 159 78 2016

PMC005xxxxxx/PMC5001956.txt

==== Front CureusCureus2168-8184Cureus2168-8184Cureus Palo Alto (CA) 10.7759/cureus.712Radiation OncologyCardiologyAnalysis of Pulmonary Vein Antrums Motion with Cardiac Contraction Using Dual-Source Computed Tomography Muacevic Alexander Adler John R Bahig Houda 1de Guise Jacques 2Vu Toni 1Chartrand-Lefebvre Carl 3Blais Danis 4Lebeau Martin 1Nguyen Nhu-Tram 5Roberge David 67161 Department of Radiation Oncology, Centre hospitalier de l'université de Montréal (CHUM) 2 Medical Imaging, Centre hospitalier de l'université de Montréal (CHUM) 3 Department of Radiology, Centre Hospitalier de l'Université de Montréal (CHUM) 4 Department of Radiation Oncology, Centre hospitalier de l'université de Montréal (CHUM) - Hôpital Notre-Dame 5 Department of Radiation Oncology, McMaster University-Juravinski Cancer Centre, Hamilton, ON 6 Department of Oncology, Division of Radiation Oncology, McGill University Health Center 7 Department of Radiology, Radiation Oncology and Nuclear Medicine, University of Montreal Houda Bahig houdabahig@gmail.com26 7 2016 7 2016 8 7 e71215 6 2016 26 7 2016 Copyright © 2016, Bahig et al.2016Bahig et al.This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.This article is available from http://www.cureus.com/articles/4781-analysis-of-pulmonary-vein-antrums-motion-with-cardiac-contraction-using-dual-source-computed-tomographyPurpose: The purpose of the study was to determine the extent of displacement of the pulmonary vein antrums resulting from the intrinsic motion of the heart using 4D cardiac dual-source computed tomography (DSCT). Methods: Ten consecutive female patients were enrolled in this prospective planning study. In breath-hold, a contrast-injected cardiac 4-dimensional (4D) computed tomography (CT) synchronized to the electrocardiogram was obtained using a prospective sequential acquisition method including the extreme phases of systole and diastole. Right and left atrial fibrillation target volumes (CTVR and CTVL) were defined, with each target volume containing the antral regions of the superior and inferior pulmonary veins. Four points of interest were used as surrogates for the right superior and inferior pulmonary vein antrum (RSPVA and RIPVA) and the left superior and inferior pulmonary vein antrum (LSPVA and LIPVA). On our 4D post-processing workstation (MIM Maestro™, MIM Software Inc.), maximum displacement of each point of interest from diastole to systole was measured in the mediolateral (ML), anteroposterior (AP), and superoinferior (SI) directions. Results: Median age of the enrolled patients was 60 years (range, 56-71 years). Within the CTVR, the mean displacements of the superior and inferior surrogates were 3 mm vs. 1 mm (p=0.002), 2 mm vs. 0 mm (p= 0.001), and 3 mm vs. 0 mm (p=0.00001), in the ML, AP, and SI directions, respectively. On the left, mean absolute displacements of the LSPVA vs. LIPVA were similar at 4 mm vs. 1 mm (p=0.0008), 2 mm vs. 0 mm (p= 0.001), and 3 mm vs. 1 mm (p=0.00001) in the ML, AP, and SI directions. Conclusion: When isolated from breathing, cardiac contraction is associated with minimal inferior pulmonary veins motion and modest (1-6 mm) motion of the superior veins. Target deformation was thus of a magnitude similar or greater than target motion, limiting the potential gains of cardiac tracking. Optimal strategies for cardiac radiosurgery should thus either incorporate the generation of an internal target or cardiac gating. In either case, cardiac 4D DSCT would allow for personalized margin definition. radio-surgerycardiacatrial fibrillationdual source ctpulmonary vein isolationGrant from the Quebec Bio-Imaging NetworkThe content published in Cureus is the result of clinical experience and/or research by independent individuals or organizations. Cureus is not responsible for the scientific accuracy or reliability of data or conclusions published herein. All content published within Cureus is intended only for educational, research and reference purposes. Additionally, articles published within Cureus should not be deemed a suitable substitute for the advice of a qualified health care professional. Do not disregard or avoid professional medical advice due to content published within Cureus. ==== Body Introduction Atrial fibrillation (AF) is the most common type of cardiac arrhythmia, with an estimated prevalence of four million in the United States [1], and up to 70% of the cases occurring in patients aged between 65 and 85 years old [2]. Its consequences involve increased risk of death, increased thromboembolic events as well as decreased quality of life [1]. AF is caused by aberrant electrical impulses from the pulmonary veins entering the left atrium, causing ineffective rapid contraction of the left atrium leading to an irregular heart rhythm and potential thrombus formation [3]. Electrical isolation of the pulmonary veins through catheter ablation is a well-established treatment approach for atrial fibrillation that has been associated with up to 60% control of AF at five years [4]. Pulmonary vein isolation by catheter ablation is an invasive procedure that is associated with up to 6% complications rate, including thromboembolic events, myocardial infarction, cardiac tamponade, oesophageal injury, or even death [5-6]. A significant proportion of the elderly AF population (with frequent co-morbidities) are thus ineligible for the procedure. Stereotactic radiosurgery (SRS) as a treatment for AF has been investigated in a limited number of studies. In fact, the ability to create a fibrotic cardiac lesion [7-10] and the potential to induce an electrophysiological effect have been suggested in animal studies [7, 11]. SRS potential resides in offering a non-invasive alternative to older patients with significant comorbidities. When considering cardiac SRS, respiratory motion and cardiac contraction both contribute to target motion. Whereas several techniques such as breath hold, respiratory gating, or near real-time tracking are used to take into account respiratory motion, cardiac contractions remain challenging. Respiratory tracking has been investigated using the Cyberknife system (Accuray Inc. Sunnyvale, USA) with placement of fiducial markers near the target volume [7-9] as well as, more recently, using real-time cardiac magnetic resonance imaging [12]. Ipsen et al., [12] conducted a dosimetric study in one human subject targeting all four pulmonary veins antrums. They showed that with the use of larger safety margins, heart and esophagus dose constraints were exceeded, and target volume coverage was compromised, therefore emphasizing the importance of minimizing unnecessary margins. Whether patients are treated under respiratory tracking method or using a breath-hold technique, a better understanding of the dynamics of cardiac displacement across the four pulmonary vein antrums is needed. In this study, we used a dual-source computed tomography (DSCT) synchronized to the patient’s electrocardiogram in order to assess diastole-to-systole associated cardiac displacement [13]. The purpose of the study was to determine the extent of displacement of each pulmonary vein antrum resulting from cardiac contraction in order to generate individualized internal target volume margin accounting for cardiac deformation. Materials and methods Study population Ten consecutive female patients with left breast cancer aged 18 years and older and no known allergy to iodine contrast were prospectively enrolled in a study of cardiac DSCT radiotherapy simulation between July 2015 and March 2016. All patients had a baseline electrocardiogram and biochemistry including creatinine level. The protocol and patient consent form were reviewed and approved by our institutional ethics committee. Cardiac 4D DSCT A contrast-injected (Isovue 370), cardiac 4D DSCT synchronized to the patients' electrocardiogram (ECG) scan was acquired on a Somatotom Flash Definition (Siemens Healthcare, Erlangen, Germany). Contrast injection was at a rate of 4ml/sec and had three phases: a) 60 ml of contrast over the first fifteen seconds, b) 10 ml of contrast and 10 ml of Sodium Chloride (NaCl) 0.9% over the following five seconds, and c) 40 ml of NaCl 0.9% over the final ten seconds. With a rotation time of 0.28 second using a detector with 64 × 0.6 mm beam collimation, the acquisition time for the entire cardiac volume was <0.3 seconds at a temporal resolution of 75 milliseconds [13]. The cardiac 4D DSCT was obtained in the supine position with arms up, using a breast board and vaclock as immobilization devices.  A prospective sequential acquisition method including eight phases of the cardiac cycle (20%, 30%, 40%, 50%, 60%, 70%, 80% and 90%) including the extreme phases of systole and diastole was used. Cardiac 4D DSCT was acquired in breath-hold (either deep inspiration breath-hold or natural inspiration breath-hold reproduced using an Abches system (APEX Medical, Tokyo, Japan)). ​ Pulmonary vein motion assessment Two clinical target volumes (CTV) were defined for isolation of all four pulmonary veins [12]: a right CTV (CTVR) which included the right superior and inferior pulmonary vein antrums (RSPVA and RIPVA) and a left CTV (CTVL) which included the left superior and inferior pulmonary vein antrums (LSPVA and LIPVA). Four regions of interest (ROI) were defined as a 5 mm surface on axial view at the most antero-superior intersection of each pulmonary vein with the left atrium. These ROIs representing the RSPVA, RIPVA, LSPVA, and LIPVA were contoured on each of the eight cardiac phases by a same observer with particular care to ensure ROI was segmented at the same location in all phases. Maximal displacement of the centroid of each ROI was derived automatically in a MIM Maestro workstation (MIM Software Inc., Cleveland, OH) in the mediolateral (ML), anteroposterior (AP), and superoinferior (SI) directions. Data were collected in an encrypted electronic database. Student T-test was used to compare maximal displacements. Analyses were completed using the SPSS statistics package (IBM Corp, 2013. IBM SPSS Statistics for Windows, Version 22.0. Armonk, NY).  Results Patients characteristics Ten female patients with a median age of 60 years (range, 56-71 years) were included in this study. Baseline ECG showed sinus rhythm in all patients, and none of the patients had a past medical history of AF, congestive heart failure, or valvular disease. One patient (patient #8), was on an angiotensin II receptor antagonist for hypertension. No patient presented anatomical variants such as a common pulmonary trunk or an accessory vein. All patients presented a total of four pulmonary veins. Table 1 summarizes basic scan parameters. Table 1 Basic DSCT-Related Parameters CTDIvol= volume CT dose index, DLP= dose length product, mSv= milisievert, N= number of patients, DIBH= deep inspiration breath-hold. Effective dose was estimated by the product of the DLP and a conversion coefficient for the chest (k= 0.014 mSv*mGy-1*cm-1). Scan-Related Parameters Mean Range CTDIvol-32 cm (mGy) 60.7 42.3-73.0 DLP (mGy-cm) 1072 729-1554 Effective dose (mSv) 15 (10-22)   N % DIBH 4 40% Motion analysis Plots of maximum cardiac displacement of right superior vs. right inferior pulmonary antrums as well as left superior vs. left inferior pulmonary vein antrums are shown in Figures 1A-1B, respectively. Figure 1 Plots of Maximum Cardiac Displacement of Right Superior vs. Right Inferior Pulmonary Antrums (A), as well as Left Superior vs. Left Inferior Pulmonary Vein Antrums (B) RSPVA= right superior pulmonary vein antrum, RIPVA= right inferior pulmonary vein antrum, LSPVA= left superior pulmonary vein antrum, LIPVA= left inferior pulmonary vein antrum. Displacements shown represent the motion of each structure in the ML, AP, and SI directions, from diastole to systole, secondary to deformation of the CTVR and CTVL. These plots show the that magnitude as well as the direction of the displacements are different for the superior and inferior veins. Figure 2 shows a 5mm ROI in the axial plane at the most anterior intersection of the right superior pulmonary vein with the left atrium, representative of the RSPVA.  Figure 2 ROI Representing the RSPVA Defined as a 5 mm Axial Surface at the Anterosuperior Intersection of the Superior Pulmonary Vein with the Left Atrium ROI= region of interest, RSPVA= right superior pulmonary vein antrum Mean displacements of RSPVA, RIPVA, LSPVA, and LIPVA are shown in Table 2.  For the CTVR, mean absolute displacement of RSPVA vs. RIPVA was 3 ±2 mm vs. 1 ±1 mm (p=0.002), 2 ±1 mm vs. 0 ±0 mm (p= 0.001), and 3 ±3 mm vs. 0 ±1 mm (p=0.00001), in the ML, AP, and SI directions, respectively. For the CTVL, mean absolute displacement of LSPVA vs. LIPVA was 4 ±2 mm vs. 1 ±1 mm (p=0.0008), 2 ±1 mm vs. 0 ±0 mm (p= 0.001), and 3 ±2 mm vs. 1 ±1 mm (p=0.00001) in the ML, AP, and SI directions respectively. Table 2 Mean Displacements of RSPVA, RIPVA, LSPVA, and LIPVA RSPVA= right superior pulmonary vein antrum, RIPVA= right inferior pulmonary vein antrum, LSPVA= left superior pulmonary vein antrum, LIPVA= left superior pulmonary vein antrum. SD= standard deviation, ML= mediolateral, AP= anteroposterior, SI= superoinferior.   ML (mm) AP (mm) SI (mm)   Mean SD p Mean SD p Mean SD p RSPVA 3 ± 2   0.002 2 ± 1   0.001 3 ± 3   0.00001 RIPVA 1 ± 1 0 ± 0 0 ± 1 LSPVA 4 ± 2   0.0008 2 ± 1   0.001 3 ± 2   0.00001 LIPVA 1 ± 1 0 ± 0 1 ± 1 Discussion In this study, we determined the extent of displacement of the pulmonary veins resulting from the contractile motion of the heart. Using a cardiac 4D DSCT, we have shown that margins accounting for cardiac deformation can be individualized and that inferior pulmonary veins were significantly less mobile than their superior counterpart. We found that mean displacement of the superior pulmonary veins was typically 4 mm in the ML direction (reaching up to 8 mm in one patient), and 3 mm in the SI direction (reaching up to 7 mm in one patient). Mean displacement of the inferior veins was 1 mm or less in all directions. Our study was conducted in healthy individuals with no known AF. Cardiac anatomy remodeling in patients with AF has been described [14], and therefore, it is possible that our reported pulmonary veins motion dynamics be different from that expected in patients with known AF. In addition, our results may be limited by the uncertainty related to intra-observer variability in the measure of pulmonary veins displacements. However, displacements reported in our study are similar to the results from a study by Rettman et al. [15]. In that study, clips were placed within the pulmonary vein ostia and left atrial appendage of canine hearts; average displacement of eleven clips placed in three canine hearts was 2 mm, 2 mm, and 1 mm in the ML, AP, and SI directions, respectively. Importantly, the definition of a precise target volume for cardiac radiosurgery is poorly defined in the literature. In addition, no patient included in our study had a known diagnosis AF, which may limit the applicability of our findings to healthy individuals. AF is particularly challenging when it comes to cardiac imaging, as the rapid rates are associated with significant motion artifacts and impaired image quality. A distinctive feature of our study is the use of DSCT which provides high temporal resolution even in patients with AF rhythm [16]. Literature on the role of radiosurgery for the treatment of AF by pulmonary vein isolation remains preliminary. A study on mini swine using the Cyberheart system (Portola Valley, CA), where fiducials were implanted next to the target volume, demonstrated the feasibility of using stereotactic robotic radiosurgery to create cardiac fibrotic lesions as well as to induce a significant decrease in voltage at the pulmonary vein–left atrial junction at a dose of 25 Gy [7]. In another animal study targeting the right pulmonary vein ostia, the use of in-vivo thermoluminescent dosimeter near the right pulmonary vein showed that the accuracy of the CyberKnife radiosurgery system within 5% of the predicted dose [9]. Using an internal target volume method accounting for both respiratory and cardiac motion, Bode et al., [11] investigated the feasibility of lesion formation in a porcine model. Radiation doses between 23 and 40 Gy were delivered to the right superior pulmonary vein, to which additional margins of 2-3 mm and 10-15 mm for cardiac and respiratory motion were added, respectively. At six months, right pulmonary vein voltage was reduced, and pathological analysis revealed transmural scarring with doses beyond 30 Gy [11]. However, reported toxicities included broncho-mediastinal fistula and AV node block [11]. Importantly, treatment of all four pulmonary veins would have been associated with a significantly higher treatment volume and likely increased toxicity, highlighting the importance of minimizing internal target volume margins when possible. Using a target volume similar to our study, Ipsen et al., [12] conducted a planning study on one patient with AF where all four pulmonary vein antrums were targeted. Using an incremental safety margin from 0 to 8 mm, the authors reported that increasing the margins was associated with exceeding of heart and esophagus normal tissue tolerance and significant compromise on the target volume coverage. The heart is subject to two sources of motion during treatment: respiratory motion as well as intrinsic cardiac contraction. If respiratory motion management is now commonplace through breath hold, tracking, or gating; cardiac contraction remains a challenge. Delivering a dose that can create fibrosis in the pulmonary vein antrums without complications to the surrounding healthy tissues requires maintaining high spatial dose gradients. In our study, we showed that cardiac contraction is associated with deformation of target structures, leading to different displacements of the superior and inferior pulmonary veins, that cannot be taken into account by simple tracking. Whether the patient respiratory motion is taken into by the Cyberknife System, real-time MRI, or with breath-hold technique, individualizing assessment of cardiac motion would allow for selection of optimal margins for deformation. An alternative strategy would reside in the development of a cardiac-gated treatment synchronized to ECG signal [17], in which case the multiphase DSCT would help in choosing an optimal gating window. Conclusions We found cardiac contraction to be associated with negligible inferior pulmonary vein displacement and a mean superior pulmonary vein displacement of 4 mm. This difference is the result of cardiac deformation and highlights the potential limitations of cardiac tracking. In an internal target volume strategy accounting for cardiac deformation, cardiac 4D DSCT would allow for personalized assessment of the displacements of the superior and inferior pulmonary veins and selection of optimal margins allowing optimal target coverage while avoiding unnecessary irradiation of healthy tissues. The authors have declared financial relationships, which are detailed in the next section. David Roberge declare(s) a grant and personal fees from Accuray. David Roberge declare(s) a grant and personal fees from Siemens. David Roberge declare(s) a grant and personal fees from Varian. Houda Bahig declare(s) a grant from Varian. Human Ethics Centre Hospitalier de l'Université de Montréal issued approval 14.192 Animal Ethics Animal subjects: This study did not involve animal subjects or tissue. ==== Refs References 1 The impact of age on the epidemiology of atrial fibrillation hospitalizations Am J Med Naderi S Wang Y Miller AL 158 127 2014 24332722 2 Electrophysiologic and electroanatomic changes in the human atrium associated with age J Am Coll Cardiol Kistler PM Sanders P Fynn SP 109 116 44 2004 15234418 3 Spontaneous initiation of atrial fibrillation by ectopic beats originating in the pulmonary veins N Engl J Med Haïssaguerre M Jaïs P Shah DC 659 666 339 1998 9725923 4 Long-term follow-up after catheter ablation of paroxysmal atrial fibrillation: the incidence of recurrence and progression of atrial fibrillation Circ Arrhythm Electrophysiol Takigawa M Takahashi A Kuwahara T 267 273 7 2014 24610740 5 2012 HRS/EHRA/ECAS expert consensus statement on catheter and surgical ablation of atrial fibrillation: recommendations for patient selection, procedural techniques, patient management and follow-up, definitions, endpoints, and research trial design: a report of the Heart Rhythm Society (HRS) Task Force on Catheter and Surgical Ablation of Atrial Fibrillation. Developed in partnership with the European Heart Rhythm Association (EHRA), a registered branch of the European Society of Cardiology (ESC) and the European Cardiac Arrhythmia Society (ECAS); and in collaboration with the American College of Cardiology (ACC), American Heart Association (AHA), the Asia Pacific Heart Rhythm Society (APHRS), and the Society of Thoracic Surgeons (STS). Endorsed by the governing bodies of the American College of Cardiology Foundation, the American Heart Association, the European Cardiac Arrhythmia Society, the European Heart Rhythm Association, the Society of Thoracic Surgeons, the Asia Pacific Heart Rhythm Society, and the Heart Rhythm Society Heart Rhythm Calkins H Kuck KH Cappato R 632 696 Society 9 2012 22386883 6 In-hospital complications associated with catheter ablation of atrial fibrillation in the United States between 2000 and 2010: analysis of 93,801 procedures Circulation Deshmukh A Patel NJ Pant S 2104 2112 128 2013 24061087 7 Noninvasive stereotactic radiosurgery (CyberHeart) for creation of ablation lesions in the atrium Heart Rhythm Sharma A Wong D Weidlich G 802 810 7 2010 20156591 8 Cardiac radiosurgery (CyberHeart™) for treatment of arrhythmia: physiologic and histopathologic correlation in the porcine model Cureus 5 2016 Maguire PJ Gardner E Jack AB 0 3 2011 9 In vivo dose measurement using TLDs and MOSFET dosimeters for cardiac radiosurgery J Appl Clin Med Phys Gardner EA Sumanaweera TS Blanck O 3745 13 2012 http://www.jacmp.org/index.php/jacmp/article/view/3745/2524 22584173 10 Dose-escalation study for cardiac radiosurgery in a porcine model Int J Radiat Oncol Biol Phys Blanck O Bode F Gebhard M 590 598 89 2014 24751407 11 Pulmonary vein isolation by radiosurgery: implications for non-invasive treatment of atrial fibrillation Europace Bode F Blanck O Gebhard M 1868 1874 17 2015 25736725 12 Radiotherapy beyond cancer: target localization in real-time MRI and treatment planning for cardiac radiosurgery Med Phys Ipsen S Blanck O Oborn B 120702 41 2014 25471947 13 Flash imaging in dual source CT (DSCT) Proc SPIE 7258 Medical Imaging Bruder H Petersilka M Mehldau H 7258OD 2009 14 Cardiac remodeling as a consequence of atrial fibrillation: An anatomical study of perfusion-fixed human heart specimens J Geriatr Cardiol Rolfes CD Howard SA Goff RP Iaizzo PA 141 146 8 2011 22783300 15 Analysis of left atrial respiratory and cardiac motion for cardiac ablation therapy Proc SPIE Int Soc Opt Eng Rettmann ME Holmes DR 3rd Johnson SB Lehmann HI Robb RA Packer DL 0 9415 2015 16 Effectiveness of dual-source CT coronary angiography for the evaluation of coronary artery disease in patients with atrial fibrillation: initial experience Radiology Oncel D Oncel G Tastan A 703 711 245 2007 18024451 17 Reduction of organ motion by combined cardiac gating and respiratory gating Int J Radiat Oncol Biol Phys Wang Z Willett CG Yin FF 259 266 68 2007 17321071

PMC005xxxxxx/PMC5001957.txt

==== Front 3 Biotech3 Biotech3 Biotech2190-5738Springer Berlin Heidelberg Berlin/Heidelberg 49710.1007/s13205-016-0497-4Short ReportsParameters influencing Agrobacterium-mediated transformation system in safflower genotypes AKS-207 and PKV Pink Dhumale Dipti Raghunath 1Shingote Prashant Raghunath prashantshingote2008@gmail.com 12Dudhare Mahendra Shankarrao 1Jadhav Pravin Vishwanath 1Kale Prashant Bhaskar kalepbbtl@gmail.com 11 Department of Agricultural Botany, Biotechnology Centre, Dr Panjabrao Deshmukh Agricultural University, Akola, MS 444104 India 2 National Research Centre on Plant Biotechnology, Pusa Campus, New Delhi, 110012 India 26 8 2016 26 8 2016 12 2016 6 2 18113 2 2016 16 8 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Shoot regeneration in safflower (Carthamus tinctorius ‘AKS 207’ and ‘PKV Pink’) genetically transformed using Agrobacterium was used for assessing various constraints to the efficiency of transformation including infection period, virulence induction medium, co-cultivation period, bacterial titre, selection regime, and the natural phenolic compound acetosyringone. Transformation frequency was promising with 8–10-day-old cotyledonary leaf explants. Therefore, explants of that age cultured on Agrobacterium minimal medium (AB) containing 100 µM acetosyringone were infected with Agrobacterium (cell titre 0.5 OD600nm) for 15 min followed by 48 h of co-cultivation on kanamycin-enriched medium (50 mg/L). Transformation of the shoots was confirmed using β-glucuronidase (GUS) histochemical assay and polymerase chain reaction (PCR). With the transformation protocol thus optimized, the transformation frequency as determined using GUS assays was 54.0 % for AKS 207 and 47.6 % for PKV Pink. The corresponding figures using PCR were 27.0 and 33.3 %. The transformed shoots required 10–14 weeks of culture initiation but produced very few roots. Electronic supplementary material The online version of this article (doi:10.1007/s13205-016-0497-4) contains supplementary material, which is available to authorized users. Keywords Indirect regenerationAgrobacterium tumefaciensGUS histochemical assayCarthamus tinctorius L.Oilseedissue-copyright-statement© King Abdulaziz City for Science and Technology 2016 ==== Body Introduction Safflower (Carthamus tinctorius), family Asteraceae (Compositae), is annually cultivated in the tropics and subtropics of the world. It is cultivated mainly for its seeds and flowers which have commercial values (Vijaya Kumar et al. 2008). The flowers are useful in medicine, as a source of a colouring agent and of fibres, and for food flavouring. Safflower oil is rich in linoleic acid (75–90 %) and is believed to lower cholesterol levels in blood. Safflower is, therefore, important in food, pharmaceutical, paint, and lubricant industries (Lijiao and Meili 2013). Safflower oil commands a higher price than other edible oils, but the higher price is offset by the potential health benefits of the oil (Li and Mündel 1996). Safflower is particularly suited for molecular farming and is used in the production of human insulin, lipoproteins, growth hormones, and specialty oils of high nutritional value (Shilpa et al. 2010). A major drawback of safflower is that it is susceptible to insect pests. The safflower aphid (Uroleucon compositae) is the most serious pest, which is estimated to reduce yields by 30–80 % (Hanumantharaya et al. 2008); at the same time, the existing measures for pest control are expensive, and various management practices and breeding strategies are not particularly effective. Genetic engineering offers some advantages over traditional methods of breeding, and Agrobacterium-meditated transformation of genes has become a gold standard (Shingote et al. 2015; Kharte et al. 2016), a method that integrates fewer copies of trans genes into plants compared to the biolistic method and is, therefore, the preferred method for obtaining stable expression of trans genes and avoiding trans gene silencing (Joyce et al. 2010; Ramesh et al. 2004). In this process, only the T-DNA region of the vector is transferred, a region engineered to encompass a selectable marker, a reporter gene, and the genes of interest, which are then transferred from the bacterium to the host plant’s nuclear genome. These functions are facilitated by a set of Virulence (Vir) genes in the presence of acetosyringone (AS), a phenolic inducer released by wounded plant cells (Ali et al. 2007). Safflower, however, has not proved amenable to genetic manipulation: in vitro regeneration has proved difficult; and Agrobacterium-mediated transformation has been tried in only a few varieties of safflower (Ying et al. 1992; Nikam and Shitole 1999; Rao and Rohini 1999; Rohini and Rao 2000; Belide et al. 2011; Motamedi et al. 2011). The present work, therefore, examines some of the transformation parameters in detail to find out ways to deploy genetic transformation in two safflower genotypes, namely AKS 207 and PKV Pink. Materials and methods Tissue culture Certified seeds of safflower genotypes AKS 207 and PKV Pink were obtained from the Oilseeds Research Unit, Dr Punjabrao Deshmukh Agricultural University, Akola, Maharashtra, India, and germinated in vitro. Cotyledonary leaf explants (about 0.5–1 cm2) were excised from 8- to 10-day-old seedlings and placed in the callus induction medium (CIM), which was nothing but Murashige and Skoog (MS) Medium (Murashige and Skoog 1962) supplemented with 2,4-dichlorophenoxyacetic acid and kinetin. The calli were induced and maintained on a mix containing one part of MS medium mixed with one, two, or three parts of CIM. Shoot induction from embryogenic calli was carried out on different combinations of the shoot induction medium (SIM), which contained MS medium supplemented with BAP (1, 2, 3, 4 or 5 mg/L), alone or in combination with naphthalene acetic acid (0.5 mg/L). Rooting of the induced multiple shoots was attempted on half-strength MS medium supplemented with different hormonal combinations reported in the literature and with a few more combinations devised for the present experiment. Bacterial strain, vector construct, and culture conditions Agrobacterium strain EHA-105 harbouring the recombinant plasmid pCAMBIA2301 (CAMBIA, Canberra, Australia) was used in the safflower transformation system (Fig. 1). The bacteria were grown aseptically in Luria–Bertani (LB) medium containing two antibiotics, namely kanamycin (50 mg/L) and rifampicin (10 mg/L), 28 °C.Fig. 1 Recombinant pCambia2301::Lentil-lectin gene construct for safflower transformation Sensitivity to kanamycin Because the vector construct harboured nptII gene as the plant selection marker, kanamycin was used as the selection agent. To determine the effective dose of kanamycin, the explants were cultured in optimized CIM supplemented with different concentrations of kanamycin (0, 25, 50, 75 and 100 mg/L) in Petri dishes, and kanamycin-free CIM was used as a control (Sujatha et al. 2012). The resulting calli were subcultured routinely every 2 weeks and observed closely every 3–4 days for necrosis. The percentage of explants showing necrosis was recorded after 4 weeks of incubation. Agrobacterium infection and co-cultivation Cotyledonary leaf explants prepared as above were inoculated by exposing them for 20 min to Agrobacterium cultured on bacterial growth medium (BGM) enriched with AS. The explants were blotted dry and transferred to the optimized callusing medium supplemented with appropriate AS concentration and co-cultivated in dark at 26 ± 2 °C. After co-cultivation, the explants were washed with sterile distilled water alone and with 500 mg/L cefotaxime and transferred to MS medium supplemented with the optimized callusing and selection medium. After 10–14 weeks, the successfully regenerated and putatively transformed multiple shoots from calli were selected for further confirmation of gene integration. For optimizing a genotype-independent transformation method for safflower, the cotyledonary leaf explants were cultured under different conditions. The deliberate variables included bacterial titre (0.25, 0.50, 0.75, and 1.0 OD at 600 nm) in combination with different lengths of the infection period (5, 10, 15, and 20 min), variations of the BGM (LB, AB, and MS), AS concentration (0, 50, 100, and 200 µM), and duration of assisted transformation (0, 24, 48 and 72 h). To enhance the penetration of the target tissues by Agrobacterium, the explants were pre-treated in individual experiments. Transformation efficiency was recorded as the number of putatively transformed multiple shoots obtained after 5–7 cycles of selection (10–14 weeks) (Sujatha et al. 2012). Histochemical analysis of gene expression The antibiotic-resistant putatively transformed shoots were subjected to histochemical analysis (Jefferson 1987) of gene expression using GUS assays 10–14 weeks after Agrobacterium infection. Untransformed explants cultured under identical conditions served as controls (Shilpa et al. 2010). The cultures were checked visually for regenerating shoots and photographed after colour development and then examined for the presence of blue spots. Extraction of genomic DNA and confirmation using polymerase chain reaction The integration of the Lentil-lectin gene with safflower genome of the putatively transformed GUS-positive multiple shoots was further confirmed through polymerase chain reaction (PCR). A modification of the Doyle and Doyle method that uses cetyltrimethyl ammonium bromide (CTAB) was used for isolating the DNA from putatively transformed and non-transformed (control) multiple shoots (Doyle and Doyle 1990). For PCR confirmation, a pair of Lentil-lectin-gene-specific primers was used. The recombinant plasmid served as a positive control, and DNA from the non-transformed plants served as a negative control. The amplified products were separated on 1.0 % (w/v) agarose gel using gel electrophoresis (Mini-Sub® Bio-Rad, USA). Statistical analysis Observations on the explants producing putatively transformed calli were recorded after 4 weeks of incubation for each selection cycle. Data were analysed using one-way analysis of variance (ANOVA). The mean values of treatments were subjected to Duncan’s multiple range test (DMRT) at 0.05 % level of significance and determined using SPSS ver. 11.09. Results Tissue culture Within 3–4 weeks, extremely fragile and pale yellow (embryogenic) calli appeared on MS medium supplemented with 1 mg/L each of 2–4, D and kinetin (Fig. S1). The callusing percentage was found to be maximum (100 % in AKS 207 and 99 % in PKV Pink) in that version of MS medium. When the ratio was changed to 2 or 3 parts of kinetin to 1 part of 2, 4-D, callusing in AKS 207 was 99 and 98 %, respectively, whereas in PKV Pink the corresponding figures were 98 and 97 %. When the calli were about 7–10 weeks old, they were transferred to SIM for inducing multiple shoots. In MS medium supplemented with BAP at 1, 2, 3, 4 and 5 mg/L, shoot induction percentages were, respectively, 40, 54, 67, 66 and 66 % in AKS 207 and 34, 40, 55, 65 and 61 % in PKV Pink (Fig. S2). Root induction was attempted in elongated multiple shoots employing various hormonal combinations as reported by Nikam and Shitole (1999) and by Mandal and Gupta (2001) but rooting 8 % in both the genotypes was observed only in MS supplemented with 2 mg/L NAA: all the other combinations failed to induce rooting. The regeneration of whole plant was complete within 12–13 weeks of culture initiation (Fig. 2).Fig. 2 Optimized regeneration protocol for safflower genotypes AKS-207 and PKV pink. a Callus formation. b Shoot induction. c Multiple shoot formation. d Rooting. e Hardening in pro tray Sensitivity test for selecting transformed tissues The transformed tissues proved sensitive to kanamycin: the greater the concentration of kanamycin, the higher the frequency of necrosis in the explants and the lower the percentage of callus induction and survival (Table 1). On the other hand, the explants grown on media without kanamycin (the control) showed higher survival and better growth. After 4 weeks of culture initiation, kanamycin at 25 mg/L led to 30.8 % explants showing necrosis in AKS 207 and 22.8 % in PKV Pink; when the concentration was increased to 50 mg/L, the LD50 of necrosis was 52 % in AKS 207 and 54 % in PKV Pink; even higher concentrations proved more damaging still, with 83.3 % of the explants in AKS 207 and 97.3 % of the explants in PKV Pink being necrosed at 75 mg/L. The corresponding figures were 80.7 and 98.7 % at 100 mg/L. At higher concentrations, the explants showed necrosis within 4–8 days of culture (Table 1).Table 1 Sensitivity of safflower to kanamycin during somatic organogenesis through callus induction from cotyledonary leaf explant Kanamycin mg/L Necrosis in  % AKS-207 PKV PINK 0 9.33 ± 2.9e 6.00 ± 2.3e 25 30.67 ± 2.9d 22.67 ± 2.9d 50 52.00 ± 3.5c 54.00 ± 5.8c 75 83.33 ± 4.4b 80.67 ± 5.8b 100 97.33 ± 1.8a 98.67 ± 1.3a For each treatment, 20 cotyledonary leaf explants were used and maintained in three replicates; mean percentage of callus necrosis calculated after 4 weeks of incubation and the mean number of callus produced by each inoculated explant counted after 4 weeks of incubation in the dark. Values are mean ± SE. Means followed by the same letter are not significantly different at 0.05 % level based on Duncan’s Test Factors affecting transformation efficiency Infection period After 2–3 selection cycles at different lengths of the infection period (10, 15, and 20 min) but at the same concentration of bacterial cells (OD of 0.5), explant survival in AKS 207 was the highest (48.3 %) at 15 min, 33.3 % at 10 min, and 23.3 % at 20 min (Table 2). In PKV Pink too, the maximum survival (53.3 %) was recorded when the length of the infection period was 15 min (Table 2).Table 2 Effect of different transformation parameters on explant survival and callus formation under selection pressure Bacterial cell density Infection period (min) No of explants inoculated No of callus survived AKS-207 No of callus survived PKV PINK Callus formation of AKS-207(%) Callus formation of PKV PINK (%) Transformation conditions  0.25 5 60 3 4 5.00 ± 2.9fg 6.67 ± 1.7hi 10 60 5 5 8.33 ± 1.7efg 8.33 ± 4.4ghi 15 60 11 9 18.33 ± 3.3cdefg 15.00 ± 2.9fgh 20 60 6 7 10.00 ± 5.0defg 11.67 ± 1.7ghi  0.5 5 60 10 13 16.67 ± 4.4defg 21.67 ± 4.4def 10 60 20 22 33.33 ± 3.3bc 36.67 ± 4.4b 15 60 29 32 48.33 ± 6.0a 53.33 ± 7.3a 20 60 14 11 23.33 ± 6.0cde 18.33 ± 4.4efg  0.75 5 60 13 14 21.67 ± 3.3cde 23.33 ± 9.3efg 10 60 23 19 38.33 ± 8.8bcd 31.67 ± 6.0bc 15 60 15 18 25.00 ± 5.8ab 30.00 ± 7.6cde 20 60 11 10 18.33 ± 4.4bcd 16.67 ± 6.0efg  1 5 60 12 12 20.00 ± 5.8cdef 20.00 ± 2.9efg 10 60 11 17 18.33 ± 4.4cdefg 28.33 ± 6.7bcd 15 60 7 6 11.67 ± 4.4defg 10.00 ± 5.0hi 20 60 2 1 3.33 ± 1.7g 1.67 ± 1.7i BGM containing AS in µM   LB 0 150 1 3 0.67 ± 0.7e 2.00 ± 1.2c 50 150 15 10 10.00 ± 3.5de 6.67 ± 2.9c 100 150 41 32 27.33 ± 6.6bc 21.33 ± 1.8b 200 150 35 34 23.33 ± 4.8bcd 22.67 ± 2.9b   AB 0 150 21 3 14.00 ± 3.5cde 2.00 ± 1.2c 50 150 29 27 19.33 ± 5.8bcd 18.00 ± 3.1b 100 150 59 63 39.33 ± 2.9a 42.00 ± 4.0a 200 150 36 42 24.00 ± 2.3bcd 28.00 ± 3.5b   MS 0 150 13 4 8.67 ± 2.9de 2.67 ± 1.8c 50 150 35 36 23.33 ± 5.8bcd 24.00 ± 5.3b 100 150 50 52 33.33 ± 3.5b 34.67 ± 5.5a 200 150 40 36 26.67 ± 8.2bc 24.00 ± 6.4b  Co-cultivation duration hours 0 60 1 0 1.67 ± 1.7c 0.00 ± 0.0c 24 60 26 23 43.33 ± 4.4b 38.33 ± 7.3b 48 60 41 47 68.33 ± 10.9a 78.33 ± 6.0a 72 60 13 13 21.67 ± 4.4c 21.67 ± 4.4b Mean percentage of callus necrosis calculated after 4 weeks of incubation and the mean number of callus produced by each infected explant counted after 4 weeks of incubation in the dark. Values are mean ± SE. Means followed by the same letter are not significantly different at 0.05 % level based on Duncan’s test Acetosyringone and bacterial growth medium Explant survival under selection pressure was significantly affected by the BGM (Table 2), the survival percentage being significantly higher in calli infected with bacteria grown on the AB medium than that on MS or LB medium. After 4 weeks, the transformation frequency was significantly higher on AB medium containing 100 µM AS than that on AB medium without AS (Table 2). The transformation frequencies for AKS 207 and PKV Pink were 39.3 and 42.0 %, respectively, on AB medium, followed by 33.3 and 34.7 % on MS medium and 27.3 and 21.3 % on LB medium (Table S1). Duration of co-cultivation Transformation efficiency using Lentil-lectin gene construct was directly correlated with the duration of co-cultivation, increasing significantly with duration up to 48 h but declining thereafter. A longer co-cultivation period (72 h) led to bacterial contamination, and no healthy shoots could be recovered. Genotype Deploying the optimized transformation parameters mentioned above, a genotype-independent protocol was used for both genotypes, namely cotyledonary leaf explants infected with a diluted culture of Agrobacterium (0.5 OD) grown on AB medium supplemented with 100 µM AS. The explants thus raised were washed and transferred to the selection medium (kanamycin 50 mg/L and cefotaxime 250 mg/L). Callus formation was observed 4 weeks after transformation (Fig. 3a): the putatively transformed shoots developed on MS medium containing BAP—3 mg/L for AKS 207 and 4 mg/L for PKV Pink—in 4–5 weeks after callus formation (Fig. 3b), and further growth of putatively transformed multiple shoots was observed 2–3 weeks after shoot induction (Fig. 3c). The multiple shoots grew 1.5–2.5 cm in length after 1–2 weeks of their formation (Fig. 3d).Fig. 3 Different stages of putative transformed callus and in vitro shoot regeneration on the selection medium. a Callus induction from cotyledonary leaf explants after 3–4 weeks, b shoot induction after 4–5 weeks of Callus formation, c multiple shoot after 2–3 week of Shoot induction and d elongated shoots seen after 1–2 weeks of multiple shoot formation Histochemical assay Primary transformants obtained after 5–7 cycles of selection from all the above experiments were used for GUS assay and non-transformed shoots were used as the negative control: GUS expression was ascertained by looking for blue spots (Fig. 4). A total 37 shoots of AKS 207 and 21 shoots of PKV Pink, each 1.5–2.5 cm long, were excised and used, of which 20 and 10, respectively, showed GUS expression, thus giving a transformation frequency of kanamycin-resistant shoots of 54.0 and 47.6 %, respectively (Table S1).Fig. 4 Screening of Putative transformed shoots of safflower through GUS assay observed under microscope by horizontal thin section. a putative transformed multiple shoot tissue of AKS-207 showing GUS expression, b non-transformed multiple shoot tissue of AKS-207 used as negative Control, c putative transformed multiple shoot tissue of PKV Pink showing GUS expression and d non-transformed multiple shoot tissue of PKV Pink Confirmation of GUS-positive shoots using polymerase chain reaction Genomic DNA of GUS-positive tissues of AKS 207 and PKV Pink was isolated and amplified using primers specific to the Lentil-lectin gene: ten multiple shoots of AKS 207 and seven of PKV Pink showed positive integration of the Lentil-lectin gene (Fig. 5a, b). Inline results were reported by Ying et al. (1992) for transformation of safflower calli. The frequency of transformation in GUS-positive plants through PCR analysis was 27.0 % for AKS 207 and 33.3 % for PKV Pink (Table S1). After 3–4 weeks, a total of 12 plants (eight putatively transformed and four controls) of AKS 207 and a total of eight plants (two and six) of PKV Pink showed very few roots on half-strength MS medium supplemented with NAA (2 mg/L). For hardening, these plants were transferred to pots filled with soil but, with such sparse rooting, none could survive (Fig. S3).Fig. 5 a PCR amplification of gDNA of GUS-positive AKS-207 putative transformed shoots. M-1 Kb DNA ladder, 1–20-GUS-positive plantlets, 21-positive control plasmid DNA and 22-negative control gDNA of untransformed PKV pink. b PCR amplification from gDNA of GUS-positive PKV Pink putative transformed shoots. M-1 Kb DNA ladder, 1-10-GUS-positive plantlets, 11-negative control gDNA of un-transformed PKV Pink and 12-positive control plasmid DNA Discussion The safflower aphid (Uroleucon compositae) is the most serious pest of safflower; depending on the environmental conditions, yield losses can be 30–80 % (Hanumantharaya et al. 2008). Developing safflower cultivars resistant to the aphid through transgenic means is a more environment-friendly option than using non-biodegradable and persistent chemical insecticides. Agrobacterium-mediated transformation in safflower was limited only to selectable marker and reporter gene constructs (Ying et al. 1992; Rohini and Rao 2000; Belide et al. 2011; Motamedi et al. 2011; Sujatha et al. 2012). Insect-resistant transgenic crops have been produced using overexpression of Bt (Bacillus thuringiensis) toxins, but these toxins exhibit little toxicity against such homopteran insects as aphids, mealybugs, and whiteflies (Chougule and Bonning 2012; Shingote et al. 2013). We, therefore, opted for the more effective Lentil-lectin gene construct for transformation of safflower. Selection and optimization of the selective medium with specific antibiotics are necessary for the growth of transformed tissue. We found that 50 mg/L of kanamycin (LD50) optimal for both the genotypes (50.0 % survival even after 4 weeks of the kanamycin selection cycle; Table 1). Motamedi et al. (2011) and Belide et al. (2011) used the same concentration of kanamycin for selecting transformed calli in safflower, and so did Venkatachalam et al. (1998) in groundnut and Yadav et al. (2010) in sesame. The bacteriostatic antibiotic cefotaxime at 250 mg/L was found effective for a safflower transformation protocol similar to that reported by Rao and Rohini (1999). Both titre (strength of the suspension of Agrobacterium cells) and the length of the infection period affected the success of transformation significantly (Table 2). At low titres, the extent of colonization and the frequency of transformed shoots were also low; at the same time, very high titres (OD of 1.0) and prolonged exposure (20 min) to them also proved detrimental to transformation efficiency and led to greater bacterial contamination, probably because the explants were severely injured. The ideal combination comprised a titre of 0.5 (OD600) and 15 min of exposure. Similar results were reported by Shilpa et al. (2010) although Belide et al. (2011) found 0.4 OD to be optimal and Sujatha et al. (2012) found 0.6 OD to be optimal for transformation. The influence of AS on T-DNA transfer was tested in terms of the survival percentage of calli after co-cultivation on kanamycin (50 mg/L) (Fig. 3b). A combination of 100 µM of AS and AB medium was found optimum, leading to as many as 39–42 % of the explants of both genotypes forming calli (Table 2). Further increase in AS concentration above 100–200 µM resulted in reduced transformation efficiency for both genotypes. Transferring the explants directly onto the selection medium drastically lowered the transformation efficiency whereas extending the period of co-cultivation up to 48 h increased the transformation efficiency significantly. Co-cultivation beyond 48 h, however, resulted in excessive bacterial growth and lowered the efficiency. Similar results have been reported by several researchers (Muthukumar et al. 1996; Sujatha et al. 2012). The frequency of transformation was 27.0 % in AKS 207 and 33.3 % in PKV Pink (Table S1). Earlier researchers reported very low regeneration frequency, which is fully dependent on the genotype, source and age of explants, and composition of the growth medium (Mandal and Gupta 2001; Radhika et al. 2006). We also studied the effect of different concentrations of hormones, but were unable to harden the regenerated and putatively transformed shoots sufficiently for them to survive the transfer to soil. Shoot regeneration frequency is very high but needs to be converted to genetic transformation frequencies for successful development of transgenic plants. In safflower, rooting frequency was very low in our earlier study (Dhumale et al. 2015) and in other studies involving other varieties of safflower (Nikam and Shitole 1999; Sujatha 2007, Belide et al. 2011). Poor rooting is a critical setback and must be overcome for successful application of transgenic methods in safflower. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary material 1 (DOCX 22136 kb) ==== Refs References Ali S Xianyin Z Xue Q Hassan MJ Qian H Investigation for improved genetic transformation mediated by Agrobacterium tumefaciens in two rice cultivars Biotechnology 2007 6 138 147 10.3923/biotech.2007.138.147 Belide S Luch H Singh SP Allan GG Wood CC Agrobacterium mediated transformation of safflower and the efficient recovery of transgenic plant via grafting Plant Methods 2011 7 12 10.1186/1746-4811-7-12 21595986 Chougule NP Bonning BC Toxins for transgenic resistance to hemipteran pests Toxins 2012 4 405 429 10.3390/toxins4060405 22822455 Dhumale DR Dudhare Mohite NR Shingote PR Jadhav PV Moharil MP Refinement of in vitro regeneration system in elite safflower (Carthamus tinctorius L.) genotypes J Plant Cell Tissue Res 2015 15 1 4849 4854 Doyle J Doyle J Isolation of plant DNA from fresh tissue Focus 1990 12 13 15 Hanumantharaya L, Balikai RA, Mallapur CP, Venkateshalu CP, Kumar CJ (2008) Integrated Pest Management Strategies against safflower aphid, Uroleucon compositae (Theobald). Proceedings of the 7th International Safflower Conference, Wagga Wagga, New South Wales, Australia Jefferson RA Assaying chimeric genes in plants, the GUS gene fusion system Plant Mol Bio Rep 1987 5 387 405 10.1007/BF02667740 Joyce P Kuwahata M Turner N Lakshmanan P Selection system and co-cultivation medium are important determinants of Agrobacterium -mediated transformation of sugarcane Plant Cell Rep 2010 29 2 173 183 10.1007/s00299-009-0810-3 20041254 Kharte SB Watharkar AS Shingote PR Sharanya C Kawar PG Pagariya MC Govindwar SP Functional characterization and expression study of sugarcane MYB transcription factor gene PEaMYBAS1 promoter from Erianthus arundinaceus that confers abiotic stress tolerance in tobacco RSC advances 2016 Li D, Mündel HH (1996) Safflower (Carthamus tinctorius L.) promoting the conservation and use of underutilized and neglected crops. 7. Institute of Plant Genetics and Crop Plant Research, Gatersleben/International Plant Genetic Resources Institute, Rome, Italy Lijiao F, Meili G (2013) Progress of safflower (Carthamus tinctorius L.) regeneration through tissue culture. Department of Pharmacognosy, School of Pharmacy, Second Military Medical University, Shanghai 200433, China, pp 289–301 Mandal AK Gupta SD Direct shoot organogenesis and plant regeneration in safflower Vitro Cell and Devl Bio Plant 2001 37 1 50 54 10.1007/s11627-001-0010-5 Motamedi J Zebarjadi A Kahrizi D Salmanian AH In vitro propagation and Agrobacterium mediated transformation of safflower (Carthamus tinctorius L .) using a bacterial mutated aroA gene Aust J Crop Sci 2011 5 4 479 486 Murashige ET Skoog F A revised medium for rapid growth and bioassays with tobacco tissue cultures Physiol Plant 1962 15 3 473 497 10.1111/j.1399-3054.1962.tb08052.x Muthukumar MK Mariamma V Gnanam A Genetic transformation of cotyledon explants of cowpea (Vigna unguiculata L.) using Agrobacterium tumefaciens Plant Cell Rep 1996 15 980 985 24178287 Nikam TD Shitole MG In vitro culture of safflower L.cv. Bhima: initiation, growth optimization and organogenesis Plant Cell Tissue Organ Cult 1999 5 15 22 Radhika K Sujatha M Rao NT Thidiazuron stimulates adventitious shoot regeneration in different safflower explants Biol Plant 2006 50 2 174 179 10.1007/s10535-006-0003-7 Ramesh S Nagadhra D Reddy VD Rao KV Production of transgenic indica rice resistance to yellow stem borer and sap-sucking insects, using super-binary vectors of Agrobacterium tumefaciens Plant Sci 2004 166 1077 1087 10.1016/j.plantsci.2003.12.028 Rao SK Rohini VK Gene transfer into Indian cultivars of safflower (Carthamus tinctorius L.) using Agrobacterium tumefaciens Plant Biotechnol 1999 16 201 206 10.5511/plantbiotechnology.16.201 Rohini VK Rao SK Embryo transformation, a practical approach for realizing transgenic plants of safflower (Carthamus tinctorius L .) Ann Bot 2000 86 1043 1049 10.1006/anbo.2000.1278 Shilpa SK Kumar VD Sujatha M Agrobacterium mediated genetic transformation of safflower (Carthamus tinctorius L.) Plant Cell Tissue Organ Cult 2010 103 387 401 10.1007/s11240-010-9792-7 Shingote PR Moharil MP Dhumale DR Jadhav PV Satpute NS Dudhare MS Screening of vip1/vip2 binary toxin gene and its isolation and cloning from local Bacillus thuringiensis isolates Sci Asia 2013 39 620 624 10.2306/scienceasia1513-1874.2013.39.620 Shingote PR Kawar PG Pagariya MC Kuhikar RS Thorat AS Babu KH (2015) SoMYB18, a sugarcane MYB transcription factor improves salt and dehydration tolerance in tobacco Acta Physiol Plant 2015 37 217 10.1007/s11738-015-1961-1 Sujatha M Advances in safflower biotechnology Funct Plant Sci Biotechnol 2007 1 160 170 Sujatha M Vijay S Vasavi S Reddy PV Rao CS Agrobacterium mediated transformation of cotyledons of mature seeds of multiple genotypes of sunflower (Helianthus annuus L .) Plant Cell Tissue Organ Cult 2012 110 275 287 10.1007/s11240-012-0149-2 Venkatachalam P Natesan G Narayana J Sita L Agrobacterium mediated genetic transformation of groundnut (Arachis hypogeal L .): an assessment of factors affecting regeneration of transgenic plants J Plant Res 1998 4 565 572 10.1007/BF02507792 Vijaya Kumar J Ranjitha Kumari BD Enrique C Cyclic somatic embryogenesis and efficient plant regeneration from callus of safflower Biol Plant 2008 52 429 436 10.1007/s10535-008-0087-3 Yadav M Chaudhary D Sainger M Jaiwal PK Agrobacterium tumefaciens mediated genetic transformation of sesame (Sesamum indicum L .) Plant Cell, Tissue Organ Cult 2010 103 3 377 386 10.1007/s11240-010-9791-8 Ying M Dyer WE Bergman JW Agrobacterium tumefaciens mediated transformation of Safflower (Carthamus -tinctorius L .) cv. centennial Plant Cell Rep 1992 11 581 585 10.1007/BF00233097 24213291

PMC005xxxxxx/PMC5001958.txt

==== Front AMB ExpressAMB ExpressAMB Express2191-0855Springer Berlin Heidelberg Berlin/Heidelberg 23610.1186/s13568-016-0236-6Original ArticleLipid production through simultaneous utilization of glucose, xylose, and l-arabinose by Pseudozyma hubeiensis: a comparative screening study Tanimura Ayumi ayumit@kais.kyoto-u.ac.jp 1Takashima Masako masako@jcm.riken.jp 2Sugita Takashi sugita@my-pharm.ac.jp 3Endoh Rikiya rikiyasu@jcm.riken.jp 2Ohkuma Moriya mohkuma@riken.jp 2Kishino Shigenobu kishino@kais.kyoto-u.ac.jp 1Ogawa Jun ogawa@kais.kyoto-u.ac.jp 1Shima Jun +81-77-599-5718shima@agr.ryukoku.ac.jp 41 Division of Applied Life Sciences, Graduate School of Agriculture, Kyoto University, Kitashirakawa Oiwake-cho, Sakyo-ku, Kyoto, 606-8502 Japan 2 Japan Collection of Microorganisms, RIKEN BioResource Center, Koyadai, Tsukuba, Ibaraki 305-0074 Japan 3 Department of Microbiology, Meiji Pharmaceutical University, Noshio, Kiyose, Tokyo 204-8588 Japan 4 Faculty of Agriculture, Ryukoku University, Seta Oe-cho, Otsu, Shiga 520-2194 Japan 26 8 2016 26 8 2016 2016 6 1 5817 6 2016 23 8 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Co-fermentation of glucose, xylose and l-arabinose from lignocellulosic biomass by an oleaginous yeast is anticipated as a method for biodiesel production. However, most yeasts ferment glucose first before consuming pentoses, due to glucose repression. This preferential utilization results in delayed fermentation time and lower productivity. Therefore, co-fermentation of lignocellulosic sugars could achieve cost-effective conversion of lignocellulosic biomass to microbial lipid. Comprehensive screening of oleaginous yeasts capable of simultaneously utilizing glucose, xylose, and l-arabinose was performed by measuring the concentration of sugars remaining in the medium and of lipids accumulated in the cells. We found that of 1189 strains tested, 12 had the ability to co-ferment the sugars. The basidiomycete yeast Pseudozyma hubeiensis IPM1-10, which had the highest sugars consumption rate of 94.1 %, was selected by culturing in a batch culture with the mixed-sugar medium. The strain showed (1) simultaneous utilization of all three sugars, and (2) high lipid-accumulating ability. This study suggests that P. hubeiensis IPM1-10 is a promising candidate for second-generation biodiesel production from hydrolysate of lignocellulosic biomass. Keywords Oleaginous yeastFatty acidsXylosel-arabinosePseudozyma hubeiensishttp://dx.doi.org/10.13039/501100002241Japan Science and Technology Agencyissue-copyright-statement© The Author(s) 2016 ==== Body Introduction The lipid produced by microorganisms is considered to have powerful potential for the development of a new kind of energy, and has received significant interest from sustainable energy researchers. Lipid accumulated by oleaginous yeast is viewed as a promising alternative to second-generation biodiesel, since the composition of the fatty acids produced by yeast is suitable for biodiesel production. That is, it contains palmitic (16:0), stearic (18:0), oleic (18:1), and linoleic (18:2) acids at a high ratio, mainly in the form of triacylglycerol (TAG) (Beopoulos et al. 2011; Knothe 2009; Meng et al. 2009; Sitepu et al. 2014). Compared to other oleaginous microorganisms, oleaginous yeasts are advantageous due to their rapid growth rate (Li et al. 2008), and they are deemed to have the potential to convert various carbon sources, such as cellobiose, xylose and starch, to lipid (Gong et al. 2012; Hu et al. 2011; Huang et al. 2014; Tanimura et al. 2014a). Second-generation biodiesel is made from non-food sources such as rice straw, wood residue, corncob, and sugarcane bagasse. Lignocellulosic hydrolysates from these feedstocks are composed mainly of glucose, xylose, and l-arabinose (hereafter referred to simply as arabinose) (Huang et al. 2009; Kumar et al. 2009; Madhavan et al. 2012; Roberto et al. 1995; Tsigie et al. 2011). The ratio of the sugars and their concentration in the hydrolysates vary depending on the feedstock used and pretreatment conditions (Behera et al. 2014; Kumar et al. 2009). A previous study investigated lipid accumulation using a medium containing 3 % glucose by Vanrija musci JCM 24512 (formally Cryptococcus musci) (Tanimura et al. 2014b). The strain showed higher lipid-producing ability from glucose compared to typical oleaginous yeasts such as Lipomyces starkeyi and Rhodosporidium toruloides. Strains like this that can convert glucose to lipid with high productivity are well suited for the production of glucose-rich hydrolysate such as the hydrolysate of starchy biomass. However, because pentoses content ranged from 20 to 40 % of the total released sugars (Sumphanwanich et al. 2008; Tanimura et al. 2012), glucose utilization alone is insufficient for the conversion of lignocellulosic biomass. In other words, sequential utilization of the sugars extends fermentation times. Therefore, economically feasible production of lipid will require a yeast strain with the ability to co-ferment the lignocellulosic sugars. Research has shown that engineered yeast can be valuable in expanding the substrate range. For example, Tai engineered Yarrowia lipolytica to make it utilize xylose (Tai 2012). In that case, the xylose reductase encoding gene (XYL1) and xylitol dehydrogenase encoding gene (XYL2) were transferred from the xylose-fermenting yeast Scheffersomyces stipitis into the strain. The uptake of arabinose has not yet been reported, and therefore, research in this area is expected. In addition, to avoid the problem caused by glucose repression, the quest for novel oleaginous yeasts able to co-ferment glucose, xylose, and arabinose would seem to be an efficient strategy. To the best of our knowledge, there has not yet been a screening study of oleaginous yeasts able to ferment the three sugars. The application of the following new oleaginous yeasts to the conversion of lignocellulosic sugars to lipids has been carried out: Trichosporon fermentans (Huang et al. 2009, 2014), L. starkeyi (Anschau et al. 2014), Cryptococcus curvatus (Liang et al. 2014), R. toruloides (Wiebe et al. 2012) and Y. lipolytica (Tsigie et al. 2011). However, in these sugar-consumption profiles, sequential utilization of arabinose was not observed. In this study, exhaustive screening of 1189 isolates was undertaken to identify an oleaginous yeast strain that was able to convert the glucose, xylose, and arabinose in artificial hydrolysate to lipid. We here report the discovery of Pseudozyma hubeiensis IPM1-10, which shows a significant utilization of a mixture of the sugars. Materials and methods Strains and media Yeast strains collected and taxonomically identified by Takashima et al. (2012) were our primary resources. Yeast strains isolated by Dr. Ando, Kyoto University, from the Kushiro and Kyoto area (Japan) were also assessed. YM agar medium (Difco, Detroit, MI, USA) was used for pre-culture and maintenance of yeast strains. The artificial hydrolysate of lignocellulosic biomass (mixed-sugar medium) was based on the medium used by Gong et al. (2012), which contained ammonium sulfate 1 g/L, yeast extract 0.5 g/L, potassium dihydrogenphosphate 1 g/L, magnesium sulfate 1 g/L, glucose 20 g/L, xylose 10 g/L and arabinose 5 g/L. The single sugar medium contained ammonium sulfate 1 g/L, yeast extract 0.5 g/L, potassium dihydrogenphosphate 1 g/L, magnesium sulfate 1 g/L, and glucose 35 g/L or xylose 35 g/L or arabinose 35 g/L. Screening The screening procedure is depicted in Fig. 1. For the first round of screening, one loop of 3-day-old yeast culture was suspended in 3 mL of mixed-sugar medium in a test tube, and incubated for 3 days at 28 °C, with reciprocal shaking at 300 rpm. Sugar concentrations of the culture supernatants were determined by HPLC, as described below.Fig. 1 Experimental flow scheme In the secondary screening, the yeast strains selected by the first screening were used. These strains have been deposited in the Japan Collection of Microorganisms (JCM). One loop of 3-day-old yeast culture was suspended in 25 mL of mixed-sugar medium in an Erlenmeyer flask and incubated at 28 °C, with rotary shaking at 150 rpm. The process was performed in batch culture. Culture broth was withdrawn after 4 days. Sugars concentrations of the supernatants were determined by HPLC. Cells from culture broth were harvested by centrifugation (15,000 rpm for 10 min), and washed with distilled water. Cell mass was determined by dry weight after lyophilization. Intracellular total lipids were determined by gas chromatography, as described below. Kinetic analysis of selected strains The yeast strains screened by the secondary screening were used. One loop of 3-day-old yeast culture was suspended in 100 mL of mixed-sugar medium and single sugar medium in Erlenmeyer flasks and incubated at 28 °C, with rotary shaking at 150 rpm for 10 days. All of the experiments were performed in batch culture. Fermentation broth was withdrawn at specific time intervals, and intracellular total lipids and sugar concentrations were determined. All experiments were performed in triplicate. Measurement of fatty acids Total intracellular lipid was estimated as total fatty acids. The accumulated lipid of the yeast strain was extracted from the lyophilized cells by a hydrochloric acid-catalyzed direct methylation method (Ichihara and Fukubayashi 2010). In brief, after cultivation, the centrifuged cells were lyophilized and weighed. The cells were suspended in toluene and methanol, then directly transmethylated with 8 % methanolic HCl at 100 °C for 1 h. The resultant fatty acid methyl esters were extracted with n-hexane and analyzed using a gas chromatograph (GC-2010 Plus; Shimadzu, Kyoto, Japan) equipped with a flame ionization detector (FID) and an autosampler (AOC20; Shimadzu). A TC-17 capillary column (GL Science, Tokyo, Japan) was used. Heptadecanoic acid (C17: 0) was used as an internal standard for the determination of fatty acid concentrations. Measurements of sugars Glucose, xylose, and arabinose concentrations were determined using an HPLC (Shimadzu, Kyoto, Japan) equipped with an Aminex Fermentation Monitoring Column (Bio-Rad Laboratories, Hercules, CA, USA) and Micro-Guard Cation H Refill Cartridges with a Standard Cartridge Holder (Bio-Rad Laboratories). The detector was an RID 10A refractive index detector (Shimadzu). The column was kept at 60 °C using a CTO 20A column oven (Shimadzu). Sulfuric acid solution (5 mM) was used as the mobile phase at a constant flow rate of 0.6 mL/min. Results Screening As mentioned above, the experimental flow scheme is shown in Fig. 1. A total of 1189 yeast strains were tested in test tubes containing 3 mL of mixed-sugar medium during the first screening step. The consumed glucose, xylose, and arabinose concentration ranged from 0–20 g/L (0–100 %), 0–5.8 g/L (0–58 %) and 0–5 g/L (0–100 %), respectively. Twelve oleaginous yeast strains with relatively high sugar-consuming ability were obtained through the process (Table 2). Among the 12 yeast strains selected, seven strains belonged to P. hubeiensis. In the secondary screening grown in 25 mL of mixed-sugar medium in batch culture, sugars and lipid concentration were measured after 4 days of fermentation (Fig. 2). All the tested yeast strains showed pentose-assimilating ability. P. hubeiensis IPM1-10 consumed 94.1 % of total sugar (Fig. 2a). The strain produced approximately 1.56 g/L, which was higher than the lipid concentrations (Fig. 2b) of the other selected strains. Therefore, this strain was selected for further studies.Fig. 2 Residual sugars (a) and lipid concentration (b) of 12 selected oleaginous yeast strains after a 4-day culture. Data are mean ± standard deviation (error bars) of three assays Sugar consumption and lipid production by P. hubeiensis IPM1-10 To investigate the selected strain, P. hubeiensis IPM1-10, the lipid-accumulating ability and sugar-consumption profile were determined using three kinds of single-sugar medium in batch culture. The medium contained 35 g/L of glucose, xylose, or arabinose as a sole carbon source. The strain was able to utilize xylose and arabinose for lipid fermentation, and assimilated those sugars in the same manner as glucose; approximately 83 % of each sugar was consumed in 10 days (Fig. 3). When P. hubeiensis IPM1-10 was cultivated in a medium containing 35 g/L glucose, 35 g/L xylose, or 35 g/L arabinose for 10 days, the lipid content per dry weight of cells was 21.61, 24.59 and 17.26 %, respectively (Fig. 4). The mass of accumulated lipid per cell was highest using xylose.Fig. 3 Time course analyses of P. hubeiensis IPM1-10: consumption of glucose (a), xylose (b), and arabinose (c) in single-sugar medium. Data are mean ± standard deviation (error bars) of three assays Fig. 4 Time-course analyses of P. hubeiensis IPM1-10: lipid concentration (solid line) and in cell mass (broken line) in the medium containing glucose (a), xylose (b), and arabinose (c) as a carbon source. Data are mean ± standard deviation (error bars) of three assays The fatty acid compositions of P. hubeiensis IPM1-10 after the 10-day culture are shown in Table 3. Although slight differences can be seen among the fatty acid compositions, the predominant fatty acids found in all cultures were palmitic (C16:0), stearic (C18:0), oleic (C18:1) and linoleic (C18:2) acids. Mixed sugar consumption and lipid production by P. hubeiensis IPM1-10 To demonstrate the sugar-assimilating ability of P. hubeiensis IPM1-10 using sugar mixtures as the carbon source, a time-course analysis of sugar consumption, lipid concentration, and cell mass was carried out (Fig. 5). The initial concentrations of glucose, xylose, and arabinose were 20, 10 and 5 g/L, respectively. The sugars decreased simultaneously rather than preferentially, although glucose was used at a higher rate (Fig. 5a). After 10 days of cultivation, 93.06 % of sugars had been consumed, which was comparable to the results of the secondary screening.Fig. 5 Time-course analyses of P. hubeiensis IPM1-10: consumption of sugars (a); lipid concentration (solid line) and cell mass (broken line) (b) in mixed-sugar medium. Data are mean ± standard deviation (error bars) of three assays Discussion As shown in Table 1, the hydrolysate of lignocellulosic biomass mainly contains 3 kinds of sugars: glucose, xylose, and arabinose. The ratio of the released sugars varies, depending on the raw material types and pretreatment conditions; the ratio of hexoses to pentoses generally ranged from 1.5:1 to 3:1 (Huang et al. 2013). The final product contains a not negligible amount of arabinose. In the present study, glucose, xylose, and arabinose concentration were set at 20, 10 and 5 g/L, respectively. These concentrations are within the measured values, and were appropriate as the screening medium.Table 1 Sugar composition of lignocellulosic hydrolysates Material Glucose [g/L] Xylose [g/L] Arabinose [g/L] Total [g/L] References Rice straw 15.5 84.3 17.1 116.9 Huang et al. (2009) Rice straw 22.6 79.3 13.4 115.3 Roberto et al. (1995) Rice straw 55 10 3 68 Oberoi et al. (2012) Rice bran 43 5 2 50 Tsigie et al. (2012) Sugarcane bagasse 4 14 3 21 Tsigie et al. (2011) Wheat straw 30 25 5 60 Zhang et al. (2014) Bagasse 16.8 92.9 11.4 121.1 Huang et al. (2012) Interestingly, all selected strains belonged to the Ustilaginales species (Table 2). Incidentally, IP056 is assumed to be a new species in clade 7 of Wang et al. (2015), as the sequence of the D1/D2 region of LSU rRNA gene showed a 5-nucleotide difference from that of Macalpinomyces viridans (HQ013125) and a 6-nucleotide difference from that of Macalpinomyces spermophorus (HQ013130), respectively. According to Wang et al., the species in clade 7 was not reclassified due to the taxonomic confusion of teleomorphic genera; consequently we treat this strain as an unidentified yeast strain (Wang et al. 2015). Furthermore, the selected strains (except for Moesziomyces aphidis RS041) were isolated from plants (leaf surface) collected on Iriomote Island. As previously reported, Ustilaginales species are generally distributed on the surface of leaves (Wang et al. 2006; Yoshida et al. 2014). This suggests that the inhabitants of the phyllosphere are associated with the fermentation ability of lignocellulosic sugars. Although this phenomenon is not presently understood, it is likely that the strains assimilate lignocellulose degradation products supplied by themselves or another microorganism. The xylanases-producing ability of the species has actually been reported (Adsul et al. 2009). Another feature of Ustilaginales species is their biosurfactant-producing ability (Jaseetha and Das 2013; Morita et al. 2010); namely, the strain can accumulate lipid intracellularly and/or produce biosurfactant extracellularly. This is the first report of mixed-sugar fermentation and of lipids accumulation using Ustilaginales species.Table 2 Yeast species, source, and JCM number of 12 selected oleaginous yeasts Strain Species Source JCM number IPM1-7 Pseudozyma hubeiensis Plant, Iriomote Island 24583 IPM1-9 Pseudozyma hubeiensis Plant, Iriomote Island 24584 IPM1-10 Pseudozyma hubeiensis Plant, Iriomote Island 24585 RS041 Moesziomyces aphidis Soil, Rishiri Island 24586 IP068 Pseudozyma hubeiensis Plant, Iriomote Island 24587 IP045 Pseudozyma hubeiensis Plant, Iriomote Island 24588 IP037 Ustilago siamensis Plant, Iriomote Island 24589 IP040 Moesziomyces antarctica Plant, Iriomote Island 24590 IP026 Pseudozyma hubeiensis Plant, Iriomote Island 24591 IP004 Pseudozyma hubeiensis Plant, Iriomote Island 24592 IP056 Unidentified Ustilaginales species Plant, Iriomote Island 24593 IPM46-16 Anthracocystis elionuri Plant, Iriomote Island 24544 As shown in Fig. 2, all 12 candidates showed favorable results in terms of the assimilation of pentoses. The lipid concentration of M. aphidis RS041, U. siamensis IP037, M. antarctica IP040, and A. elionuri IPM46-16 were relatively higher from the viewpoint of sugar yield (g of lipid produced per g of sugar consumed). However, their sugar consumption was not comparable to that of P. hubeiensis IPM1-10, which led to the lower lipid productivity (duration of time needed for lipid concentration), because the slow sugar uptake increased cultivation time. Lipid productivity is considered to be the most important parameter. Higher lipid productivity decreases production cost. In the selected strain, P. hubeiensis IPM1-10, the highest lipid concentration and cell mass were achieved with almost complete utilization of the sugars. Similar to the other Ustilaginales species, P. hubeiensis has been recognized as a biosurfactant producer (Konishi et al. 2008). P. hubeiensis produces lipases, assimilates oil (soy oil or bovine fat), and secrets biosurfactant (Bussamara et al. 2010, 2012). Since P. hubeiensis can also convert lignocellulosic sugars to lipid, it has great potential for utilization of unused biomass and low-cost raw materials. As shown in Fig. 4, the lipid-producing ability using arabinose was 30 % lower than those using glucose and xylose, even though the sugar consumption rates were similar (Fig. 3). The data suggested that arabinose was a less effective carbon source for P. hubeiensis IPM1-10 in terms of lipid concentration. It seems that the assimilated arabinose converted to lipid and supported cell growth at the same time, because no significant difference was observed in the cell mass between carbon sources (Fig. 4). The fatty acid composition of the lipid accumulated in P. hubeiensis IPM1-10 (Table 3) was similar to that of plant oil, which consists mainly of C16 and C18. These fatty acids are widely applicable, e.g., for biodiesel, chemicals, and toiletries. Compared to plant oil, lipid from oleaginous yeast is advantageous in terms of elements of economical production, such as reductions in the lifecycle, the amount of land required, and the effects of climate.Table 3 Fatty acid composition of P. hubeiensis IPM1-10 after a 10-day culture Carbon source C12:0 C14:0 C16:0 C16:1 C18:0 C18:1 C18:2 C22:0 C24:0 Glucose 2.9 1.5 18.1 0.2 21.4 25.1 18.2 3.5 8.7 Xylose 0.2 1.3 22.8 0.5 16.4 26.7 21.9 3.4 6.9 Arabinose 0.9 1.3 20.4 0.3 19.8 33.6 11.3 3.3 9.1 Glucose, xylose and arabinose 2.7 1.4 19.5 0.3 20.0 26.6 17.6 3.4 8.2 Data are mean of three independent assays When grown in the mixed-sugar medium, P. hubeiensis IPM1-10 required a 10-day culture. There have been several previous reports on lipid production by oleaginous yeast from mixtures of glucose, xylose, and arabinose. Sugar exhaustion was achieved at 11 days from rice straw hydrolysate by T. fermentans (Huang et al. 2009), 10 days from a semi-defined medium by T. fermentans (Huang et al. 2014), and 7 days from sugarcane bagasse hydrolysate by Y. lipolytica (Tsigie et al. 2011). Further consideration is needed to determine how best to improve fermentation conditions. On the other hand, to increase lipid accumulation, continuous or fed-batch culture might be effective (Gong et al. 2012; Zhao et al. 2008). When the sugar mixtures were used as the carbon source, the lipid concentration was higher than with glucose alone. Increasing the proportion of pentoses in the carbon source increased lipid accumulation. Papanikolaou and Aggelis indicated that xylose affected lipid yield rather than glucose, because oleaginous microorganisms exclusively utilize the phosphoketolase pathway for xylose (Papanikolaou and Aggelis 2011). Therefore, P. hubeiensis IPM1-10 provides an efficient process for converting lignocellulosic biomass, such as the glucose, xylose, and arabinose present in hydrolysates, into lipid. Comprehensive screening of oleaginous yeasts capable of simultaneously utilizing glucose, xylose, and l-arabinose was performed. Among the strains tested here, P. hubeiensis IPM1-10 had the best lipid productivity grown on lignocellulosic sugars. The strain may also be useful as a genetic resource for engineering pentoses metabolism in oleaginous microorganisms in order to improve their ability to convert sugar mixtures to lipid. More importantly, the absence of glucose repression could facilitate further study to unravel the unique sugar-assimilation mechanism. Authors’ contributions AT performed experiments and drafted the manuscript. MT, TS, RE and MO isolated and identified the tested yeast strains and revised the manuscript. SK and JO assisted with the data analysis. JS managed the overall project and revised the manuscript. All authors read and approved the final manuscript. Acknowledgements This work was supported partly by the Research Institute for Food and Agriculture of Ryukoku University, and partly by the Advanced Low Carbon Technology Research and Development Program (ALCA) of the Japan Science and Technology Agency (JST). The authors would like to thank Dr. Akinori Ando, Kyoto University, for supplying a portion of the yeast strains used in this study. Competing interests The authors declare that they have no competing interests. Ethics approval and consent to participate This article does not contain any studies with human participants or animals performed by any of the authors. ==== Refs References Adsul MG Bastawde KB Gokhale DV Biochemical characterization of two xylanases from yeast Pseudozyma hubeiensis producing only xylooligosaccharides Bioresour Technol 2009 100 6488 6495 10.1016/j.biortech.2009.07.064 19692229 Anschau A Xavier MCA Hernalsteens S Franco TT Effect of feeding strategies on lipid production by Lipomyces starkeyi Bioresour Technol 2014 157 214 222 10.1016/j.biortech.2014.01.104 24556374 Behera S Arora R Nandhagopal N Kumar S Importance of chemical pretreatment for bioconversion of lignocellulosic biomass Renew Sust Energ Rev 2014 36 91 106 10.1016/j.rser.2014.04.047 Beopoulos A Nicaud JM Gaillardin C An overview of lipid metabolism in yeasts and its impact on biotechnological processes Appl Microbiol Biotechnol 2011 90 1193 1206 10.1007/s00253-011-3212-8 21452033 Bussamara R Dall’agnol L Schrank A Fernandes KF Vainstein MH Optimal conditions for continuous immobilization of Pseudozyma hubeiensis (strain HB85A) lipase by adsorption in a packed-bed reactor by response surface methodology Enzyme Res 2012 2012 329178 10.1155/2012/329178 22315670 Bussamara R Fuentefria AM Oliveira ESd Broetto L Simcikova M Valente P Schrank A Vainstein MH Isolation of a lipase-secreting yeast for enzyme production in a pilot-plant scale batch fermentation Bioresour Technol 2010 101 268 275 10.1016/j.biortech.2008.10.063 19700311 Gong Z Wang Q Shen H Hu C Jin G Zhao ZK Co-fermentation of cellobiose and xylose by Lipomyces starkeyi for lipid production Bioresour Technol 2012 117 20 24 10.1016/j.biortech.2012.04.063 22609709 Hu C Wu S Wang Q Jin G Shen H Zhao ZK Simultaneous utilization of glucose and xylose for lipid production by Trichosporon cutaneum Biotechnol Biofuels 2011 4 25 10.1186/1754-6834-4-25 21864398 Huang C Xf Chen Xiong L XD Chen LL Ma Chen Y Single cell oil production from low-cost substrates: the possibility and potential of its industrialization Biotechnol Adv 2013 31 129 139 10.1016/j.biotechadv.2012.08.010 22960618 Huang C Li YY Liu LP Wu H Zong MH Lou WY Kinetics and Mechanism analysis on microbial oil production by Trichosporon fermentans in Rice straw hydrolysate Ind Eng Chem Res 2014 53 19034 19043 10.1021/ie502185u Huang C Zong MH WuH QP Liu Microbial oil production from rice straw hydrolysate by Trichosporon fermentans Bioresour Technol 2009 100 4535 4538 10.1016/j.biortech.2009.04.022 19433350 Huang C Wu H Li RF Zong MH Improving lipid production from bagasse hydrolysate with Trichosporon fermentans by response surface methodology N Biotechnol. 2012 29 3 372 378 10.1016/j.nbt.2011.03.008 21458601 Ichihara K Fukubayashi Y Preparation of fatty acid methyl esters for gas-liquid chromatography J Lipid Res 2010 51 635 640 10.1194/jlr.D001065 19759389 Jaseetha AS Das N Enhanced biodegradation of lindane using oil-in-water bio-microemulsion stabilized by biosurfactant produced by a new yeast strain, Pseudozyma VITJzN01 J Microbiol Biotechnol 2013 23 1598 1609 10.4014/jmb.1307.07016 23928846 Knothe G Improving biodiesel fuel properties by modifying fatty ester composition Energ Environ Sci 2009 2 759 766 10.1039/b903941d Konishi M Morita T Fukuoka T Imura T Kakugawa K Kitamoto D Efficient production of mannosylerythritol lipids with high hydrophilicity by Pseudozyma hubeiensis KM-59 Appl Microbiol Biotechnol 2008 78 37 46 10.1007/s00253-007-1292-2 18071643 Kumar P Barrett DM Delwiche MJ Stroeve P Methods for pretreatment of lignocellulosic biomass for efficient hydrolysis and biofuel production Ind Eng Chem Res 2009 48 3713 3729 10.1021/ie801542g Li Q Du W Liu D Perspectives of microbial oils for biodiesel production Appl Microbiol Biotechnol 2008 80 749 756 10.1007/s00253-008-1625-9 18690426 Liang Y Jarosz K Wardlow AT Zhang J Cui Y Lipid production by Cryptococcus curvatus on hydrolysates derived from corn fiber and sweet sorghum bagasse following dilute acid pretreatment Appl Biochem Biotechnol 2014 173 2086 2098 10.1007/s12010-014-1007-y 24928546 Madhavan A Srivastava A Kondo A Bisaria VS Bioconversion of lignocellulose-derived sugars to ethanol by engineered Saccharomyces cerevisiae Crit Rev Biotechnol 2012 32 22 48 10.3109/07388551.2010.539551 21204601 Meng X Yang JM Xu X Zhang L Nie QJ Xian M Biodiesel production from oleaginous microorganisms Renew Energ 2009 34 1 5 10.1016/j.renene.2008.04.014 Morita T Takashima M Fukuoka T Konishi M Imura T Kitamoto D Isolation of basidiomycetous yeast Pseudozyma tsukubaensis and production of glycolipid biosurfactant, a diastereomer type of mannosylerythritol lipid-B Appl Microbiol Biotechnol 2010 88 679 688 10.1007/s00253-010-2762-5 20652239 Papanikolaou S Aggelis G Lipids of oleaginous yeasts. Part I: biochemistry of single cell oil production Eur J Lipid Sci Technol 2011 113 1031 1051 10.1002/ejlt.201100014 Oberoi HS Babbar N Sandhu SK Dhaliwal SS Kaur U Chadha BS Bhargav VK Ethanol production from alkali-treated rice straw via simultaneous saccharification and fermentation using newly isolated thermotolerant Pichia kudriavzevii HOP-1 J Ind Microbiol Biotechnol. 2012 39 4 557 66 10.1007/s10295-011-1060-2 22131104 Roberto IC Sato S Demancilha IM Taqueda MES Influence of media composition on xylitol fermentation by Candida guiliiermondii using response surface methodology Biotechnol Lett 1995 17 1223 1228 10.1007/BF00128390 Sitepu IR Garay LA Sestric R Levin D Block DE Bruce German J Boundy-Mills KL Oleaginous yeasts for biodiesel: current and future trends in biology and production Biotechnol Adv 2014 32 1336 1360 10.1016/j.biotechadv.2014.08.003 25172033 Sumphanwanich J Leepipatpiboon N Srinorakutara T Akaracharanya A Evaluation of dilute-acid pretreated bagasse, corn cob and rice straw for ethanol fermentation by Saccharomyces cerevisiae Ann Microbiol 2008 58 219 225 10.1007/BF03175320 Tai M Metabolic engineering of oleaginous yeast for the production of biofuels 2012 Cambridge Massachusetts Institute of Technology Takashima M Sugita T Van BH Nakamura M Endoh R Ohkuma M Taxonomic richness of yeasts in Japan within subtropical and cool temperate areas PLoS One 2012 7 e50784 10.1371/journal.pone.0050784 23226383 Tanimura A Nakamura T Watanabe I Ogawa J Shima J Isolation of a novel strain of Candida shehatae for ethanol production at elevated temperature SpringerPlus 2012 1 27 10.1186/2193-1801-1-27 23961357 Tanimura A Takashima M Sugita T Endoh R Kikukawa M Yamaguchi S Sakuradani E Ogawa J Ohkuma M Shima J Cryptococcus terricola is a promising oleaginous yeast for biodiesel production from starch through consolidated bioprocessing Sci Rep 2014 4 4776 24758986 Tanimura A Takashima M Sugita T Endoh R Kikukawa M Yamaguchi S Sakuradani E Ogawa J Shima J Selection of oleaginous yeasts with high lipid productivity for practical biodiesel production Bioresour Technol 2014 153 230 235 10.1016/j.biortech.2013.11.086 24368271 Tsigie YA Wang C-Y Chi-Thanh T Ju Y-H Lipid production from Yarrowia lipolytica Po1g grown in sugarcane bagasse hydrolysate Bioresour Technol 2011 102 9216 9222 10.1016/j.biortech.2011.06.047 21757339 Tsigie YA Wang CY Kasim NS Diem QD Huynh LH Ho QP Truong CT Ju YH Oil Production from Po1g Using Rice Bran Hydrolysate J Biomed Biotechnol. 2012 2012 1 10 10.1155/2012/378384 21836813 Wang QM Begerow D Groenewald M Liu XZ Theelen B Bai FY Boekhout T Multigene phylogeny and taxonomic revision of yeasts and related fungi in the Ustilaginomycotina Stud Mycol 2015 81 55 83 10.1016/j.simyco.2015.10.004 26955198 Wang QM Jia JH Bai FY Pseudozyma hubeiensis sp nov and Pseudozyma shanxiensis sp nov., novel ustilaginomycetous anamorphic yeast species from plant leaves Int J Syst Evol Microbiol 2006 56 289 293 10.1099/ijs.0.63827-0 16403900 Wiebe MG Koivuranta K Penttila M Ruohonen L Lipid production in batch and fed-batch cultures of Rhodosporidium toruloides from 5 and 6 carbon carbohydrates BMC Biotechnol 2012 12 26 10.1186/1472-6750-12-26 22646156 Yoshida S Morita T Shinozaki Y Watanabe T Sameshima-Yamashita Y Koitabashi M Kitamoto D Kitamoto H Mannosylerythritol lipids secreted by phyllosphere yeast Pseudozyma antarctica is associated with its filamentous growth and propagation on plant surfaces Appl Microbiol Biotechnol 2014 98 6419 6429 10.1007/s00253-014-5675-x 24706213 Zhao X Kong X Hua Y Feng B Zhao Z Medium optimization for lipid production through co-fermentation of glucose and xylose by the oleaginous yeast Lipomyces starkeyi Eur J Lipid Sci Technol 2008 110 405 412 10.1002/ejlt.200700224 Zhang Y Chen X Qi B Luo J Shen F Su Y Khan R Wan Y Improving lactic acid productivity from wheat straw hydrolysates by membrane integrated repeated batch fermentation under non-sterilized conditions Bioresour Technol. 2014 163 160 166 10.1016/j.biortech.2014.04.038 24811443

PMC005xxxxxx/PMC5001959.txt

==== Front SpringerplusSpringerplusSpringerPlus2193-1801Springer International Publishing Cham 309110.1186/s40064-016-3091-7Short ReportDetection of mesenchymal stem cells senescence by prelamin A accumulation at the nuclear level http://orcid.org/0000-0001-9900-7357Bellotti Chiara chiara.bellotti@ior.it 12Capanni Cristina ccapanni@area.bo.cnr.it 34Lattanzi Giovanna lattanzi@area.bo.cnr.it 34Donati Davide davide.donati@ior.it 12Lucarelli Enrico enrico.lucarelli@ior.it 1Duchi Serena serena.duchi@ior.it 121 Osteoarticular Regeneration Laboratory, 3rd Orthopaedic and Traumatologic Clinic Prevalently Oncologic, Rizzoli Orthopaedic Institute, via di Barbiano 1/10, Bologna, 40036 Italy 2 Department of Biomedical and Neuromotor Sciences, University of Bologna, Bologna, Italy 3 Institute of Molecular Genetics - Unit of Bologna, CNR-National Research Council of Italy, Bologna, Italy 4 Laboratory of Musculoskeletal Cell Biology, Rizzoli Orthopaedic Institute, Bologna, Italy 26 8 2016 26 8 2016 2016 5 1 142711 2 2016 17 8 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Background Human mesenchymal stem cells (MSC), during in vitro expansion, undergo a progressive loss of proliferative potential that leads to the senescent state, associated with a reduction of their “medicinal” properties. This may hampers their efficacy in the treatment of injured tissues. Quality controls on MSC-based cell therapy products should include an assessment of the senescent state. However, a reliable and specific marker is still missing. From studies on lamin-associated disorders, has emerged the correlation between defective lamin A maturation and cellular senescence. Findings Primary cultured hMSC lines (n = 3), were analyzed by immunostaining at different life-span stages for the accumulation of prelamin A, along with other markers of cellular senescence. During culture, cells at the last stage of their life span displayed evident signs of senescence consistent with the positivity of SA-β-gal staining. We also observed a significant increase of prelamin A positive cells. Furthermore, we verified that the cells marked by prelamin A were also positive for p21Waf1 while negative for Ki67. Conclusions Overall data support that the detection of prelamin A identifies senescent MSC, providing an easy and reliable tool to be use alone or in combination with known senescence markers to screen MSC before their use in clinical applications. Electronic supplementary material The online version of this article (doi:10.1186/s40064-016-3091-7) contains supplementary material, which is available to authorized users. Keywords Mesenchymal stem cellsLamin APrelamin ASenescenceCell- and tissue-based therapyhttp://dx.doi.org/10.13039/501100003407Ministero dellIstruzione, dellUniversità e della RicercaRBPA10447Donati Davide http://dx.doi.org/10.13039/501100003196Ministero della Salute5 per mille 2011/2012Lattanzi Giovanna issue-copyright-statement© The Author(s) 2016 ==== Body Background Human mesenchymal stem cells (MSC) have raised high hopes in various therapeutic applications and their use is currently tested in about 500 clinical trials (www.clinicaltrials.gov). For many clinical approaches these cells are usually expanded in vitro prior to their utilization. However, MSC can undergo only a limited number of cell divisions under standard culture conditions, and it has been demonstrated that during in vitro proliferation they suffer a progressive and continuous process of aging (Wagner et al. 2008) that affects the proliferation and multilineage differentiation potential (Banfi et al. 2000; Kim et al. 2010), the immunomodulatory properties (Sepúlveda et al. 2014), the secretory profile (Coppe et al. 2010), the gene expression profile, and the epigenetic signature (Schellenberg et al. 2011; Yoo et al. 2013). These changes may impact the efficacy of MSC to treat injured tissues. For this reason, it is generally recommended in clinical applications to use MSC that have been cultured only for a restricted amount of time, fixing a threshold that limits their number of passages or population doublings (de Girolamo et al. 2013; Wuchter et al. 2014). Cellular senescence is a complex process, which manifests with different, multifaceted phenotypes depending on the species, the cell type, and the senescence-inducing stimulus (Kosar et al. 2011). This lack of a unique signature implies the necessity for a mindful selection of the proper marker or panel of markers when assessing the senescence status of in vitro cultured MSC. The detection of SA-β-gal activity is the most common assay for the evaluation of cell senescence due to the availability of well-established protocols and commercial kits, and the abundant bibliography reporting its use. However the limitations of this marker are universally recognized (Crowe et al. 2014; Debacq-Chainiaux et al. 2008; Wagner et al. 2009). Other markers of senescence, opportunely reviewed by de Jesus (de Jesus and Blasco 2012), such as telomere shortening, Senescence-Associated Heterochromatin Foci (SAHF), Promyelocytic Leukemia Protein Nuclear Bodies, and Senescence-associated secretory phenotype (SASP) have been proposed over years. Recently, new approaches specifically focused on MSC have been suggested. Shibata et al. explored the expression of the p16INK4A gene, suggesting its methylation state could be monitored as a surveillance against the transformation of MSC during culture (Shibata et al. 2007). Koch et al. proposed instead the use of the DNA-methylation changes observed at specific CpG sites in MSC and fibroblast to track the state of cellular senescence (Koch et al. 2012). We moved from the observation of the phenomena that influence or lead to aging in humans, in order to identify a reliable and specific marker of the senescence phenotype of MSC. In this regard, alteration of the nuclear lamina is known as one of the cellular changes observed in physiological aging (Vlcek and Foisner 2007; Lattanzi et al. 2014). Of particular interest is the LMNA gene that encodes two components of the nuclear envelope: lamin A and C. The maturation of lamin A is an elaborate process which involves several consecutive steps including: farnesylation, the proteolytic cleavage of three N-terminal amino acids, the carboxymethylation and the final removal of additional fifteen N-terminal amino acids including the farnesyl group. The final step is exclusively catalyzed by the zinc-metallopeptidase ZMPSTE24 encoded by the FACE-1 gene. Mutations affecting different steps or actors of the maturation process, which elicits the accumulation of wild-type or mutated prelamin A, are associated with progeroid laminopathies or lipodystrophy (Broers and Ramaekers 2006; Davies et al. 2011). These diseases, including the Hutchinson-Gilford progeria syndrome that is characterized by premature aging, mainly affect tissues of mesenchymal origin, suggesting a link between prelamin A and MSC senescence. The existence of this correlation was supported by the work of Scaffidi and Misteli. Their results demonstrated that the accumulation of wild type or mutant lamin A by means of expression vectors or drugs leads to an accelerated aging of human fibroblast and immortalized MSC (Scaffidi and Misteli 2008). Our goal was to verify this correlation the other way around, and so where replicative senescence of primary MSC culture leads to prelamin A accumulation. The presence of lamin A precursors in cells after their prolonged in vitro culture or in tissue specimens from aged donors was already observed by other investigators, but their analysis was focused on Vascular Smooth Muscle Cells (Ragnauth et al. 2010). As far as we know, a general and robust detection analysis of lamin A precursor in MSC that have naturally exited the replicative cycle in normal culture conditions has never been reported. Therefore, in our work we used primary cultures of human MSC isolated from the bone marrow of healthy donors to investigate the presence of unprocessed lamin A precursor during early and late stages of in vitro cultures, with the ultimate scope of proposing a proper marker to detect senescent MSC. Methods Primary human MSC were obtained from 3 non-oncologic patients (aged 20, 26, 6) during routine orthopedic surgical procedures. Cell isolation and expansion is described in the Additional file. Definition of early and late stages of in vitro MSC culture MSC were maintained in culture until they reached their maximal life span as evidenced by growth arrest (i.e. the cells failed to become confluent within 4 weeks of culture). The number of population doublings (PD) for each passage was calculated using the formula: log2(N1/N0), where N0 is the number of cells seeded and N1 is the number of cells harvested at the end of the passage. Cumulative population doublings (CPD) were calculated as the sum of PDs over passages. CPD curves were normalized with GraphPad Prism 6 Software to set the maximum CPD value as the 100 % of the cell line life-span. Early and late life-span stages were then identified by graphical interception on the CPD curve tracing horizontal lines at y coordinates equal to 50 and 80 % (Stenderup et al. 2003). Experimental observations were performed on cell samples at passages comprised in the “early stage” (life-span <50 %) or “late stage” (life-span >80 %). Senescence associated β-galactosidase assay SA-β-gal activity was detected with a senescent cell staining kit (Sigma Aldrich, St. Louis, MO, USA) according to the manufacturer’s instructions. Briefly, the 40 mg/ml stock solution of 5-bromo-4-chloro-3-indolyl β-d-galactopyranoside (X-gal) was prepared in the laboratory by dissolving the X-gal powder (Sigma Aldrich, USA) in N,N-dimethylformamide. Once prepared, it was stored at −20 °C and used within a month, to ensure the accuracy of the assay. Cells were seeded in a 24-well plate and cultivated until 60 % confluence. Plates were washed with PBS, fixed for 5 min at room temperature, and incubated at 37 °C overnight in a dry incubator with freshly prepared 1 mg/ml X-gal buffered solution. Cell nuclei were counterstained with 5 µg/ml Hoechst 33342 (Life Technologies, Eugene, OR, USA). Microphotographs of a minimum of six random fields for each sample were taken using an epifluorescence microscope (Nikon Eclipse TE2000-U, Amsterdam, Netherland) equipped with a Nikon DS-Vi1-U3 CCD color digital camera. Brightfield and fluorescent images were merged with NIS-D software (Nikon, Amsterdam, Netherlands) to count total and β-gal positive cells. Prelamin A detection and immunostaining MSC grown on coverslips were fixed in cold methanol at −20 °C for 7 min. Samples were incubated with PBS containing 4 % BSA to saturate non-specific binding and incubated overnight at 4 °C with anti-prelamin A antibody (Santa Cruz Sc-6214) diluted at 1:100. Coverslips were then washed several times in PBS and incubated 1 h at RT with donkey anti-goat secondary antibody (Santa Cruz Sc-3853) diluted 1:100. After washes with PBS, the nuclei were counterstained with 4,6-diamino-2-phenylindole (DAPI). The slides were mounted with an anti-fade reagent in glycerol and observed. To measure the percentage of prelamin A positive cells, microphotographs of a minimum of 6 random fields were taken for each sample and prelamin A positive cells were manually counted. Imaging was performed using a laser-scanning motorized confocal system (Nikon A1R, Nikon, Amsterdam, Netherlands) equipped with an Eclipse Ti-E inverted microscope and four laser lines (405, 488, 561, and 638 nm). A Plan Apo VC 60x/1.4NA Oil DIC N2 objective lens was used. Images were processed using NIS-Elements AR 4.10.01 software (Nikon, Amsterdam, Netherlands). As a positive control, accumulation of prelamin A was obtained using 25 μM mevinolin (M2147, Sigma) in complete growth medium for 18 h. Mevinolin is an isoprenoid synthesis inhibitor that causes inhibition of hydroxymethyl-glutaryl-synthase implicated in the farnesyl pathway. This compound elicits an accumulation of non-farnesylated (unprocessed) prelamin A, as reported in several studies (Mattioli et al. 2008). To observe prelamin A nuclear distribution, confocal imaging was performed using the 488 nm laser line, and laser power was adjusted to minimize photobleaching. A Plan Apo 100x/1.4NA Oil DIC H objective lens was used. Zoom and field-of-view dimensions were adjusted to give a resolution of 1024 × 1024. Z-slices were acquired every 0.175 μm for a total of 31 steps. NIS elements software permits 3D rendering of the Z-slices. Two double immunostainings of prelamin A with p21 or Ki67 were performed. The rabbit polyclonal anti-Ki67 antibody (Santa Cruz Sc-15402) was diluted 1:100 and incubated overnight at 4 °C while the rabbit monoclonal anti-p21antibody [Pierce p21 Waf1/Cip1 Antibody (R.229.6)] was diluted 1:300 and incubated overnight at 4 °C. Immunofluorescence microscopy was performed using a Nikon E600 epifluorescence microscope and a Nikon oil-immersion objective [100× magnification, 1,3 NA (numerical aperture)]. Photographs were taken using a Nikon digital camera (DXm) and NIS-Element AR software. Results and discussion In the production of MSC-based cell therapy products, safety and potency evaluations are mandatory, and among other aspects, the senescent state of culture should be assessed. Though alternatives to the classical SA-β-gal staining have been proposed (Shibata et al. 2007; Righolt et al. 2011; Koch et al. 2012) we believe an optimal marker for the identification of senescent MSC is not yet available. A correlation between accumulation of prelamin A and cellular aging has been already postulated (Scaffidi and Misteli 2008; Yu and Kang 2013; Yu et al. 2013). However, the main findings of the previous studies were mainly obtained through an induced accumulation of prelamin A with expression vectors or drugs. Our goal was to demonstrate that the spontaneous accumulation of prelamin A occurs in human MSC under a condition of replicative senescence induced by the prolonged in vitro culture. MSC samples were isolated from the bone marrow of three healthy donors and maintained under standard culture condition until they reached replicative senescence in culture. As expected, when expanded in vitro each of them presented a distinctive growth curve and progression toward the Hayflick limit (Shay and Wright 2000) (Fig. 1a). Even if cells are isolated and cultured under the same conditions, the number of passages is evidently an unreliable indicator of cellular aging. Observations were therefore planned after the analysis of the growth curves of each cell line both at early and late stages (Fig. 1b), as described in the Materials and Methods section. The comparison of cells belonging to the same life-span stage attenuates the inter-donor variability and allows the observation of a coherent range of modifications such as a gradual change from a fibroblastic-like spindle shape to a large widespread morphology (Fig. 1c) and a reduced proliferation potential in the late stage (Fig. 1d). These morphological changes are associated with cytoskeleton alteration. In fact, β-tubulin and Vimentin immunostaining analyses (Additional file 1: Fig. S1A, B), performed at early and late stages demonstrated the altered polarization and disruption of cytoskeletal microtubules filaments of senescent cells (Geißler et al. 2012).Fig. 1 MSC undergo replicative senescence during in vitro expansion. a Cumulative population doublings (CPD) of cell cultures from three different donors. b Passages and PDT intervals corresponding to the early and late stages of each MSC cell line. The end of the late stage matches with replicative senescence of the culture. c Representative brightfield images of MSC at early and late stages. White boxes indicate areas of magnification highlighting the altered shape of senescent cells compared to the well defined spindle-shape of MSC at early stages. Scale bar 200 µm. d Doubling times (DT) calculated from proliferation assay confirm the remarked reduction of proliferation potential of cells according to the progression of life-span stages The senescent status of the cell cultures at late stages was confirmed by the SA-β-gal assay that revealed an increased activity of the lysosomal β-D-galactosidase at the suboptimal pH of 6.0, compared to the cells at early stage (Fig. 2a). Similarly, the specific immunostaining for the full-length form of lamin A precursor (with an antibody directed to the C-terminal residue of the unprocessed protein) showed a higher number of cells positive for prelamin A at late stages (Fig. 2b). Confocal microscopy performed at high magnification confirmed a nucleoplasmic and rim localization of prelamin A together with an abnormal nuclear morphology (Fig. 2c, c’). From the single z-stacks representative of different planes along the nucleus (Fig. 2c’), multiple invaginations of the nuclear membrane are evident. Interestingly, the observation of an increased incidence of wrinkled nuclei in senescent cells was exploited by Righolt et al. to elaborate an imaging method that quantifies the intensity and curvature of the nuclear lamina to identify abnormal cells during aging, in vitro proliferation, and in lamina disorders (Righolt et al. 2011).Fig. 2 MSC under replicative senescence accumulate prelamin A. a Representative microphotograph of SA-β-gal assay performed on early and late stage cells. Blue staining indicates the presence of β-galactosidase activity in senescent cells. Scale bar 200 μm. b Cells at early and late stages were fixed and immunostained with a specific antibody against prelamin A and counterstained with DAPI. Scale bar 50 µm. c 3D digital rendering of Z-slices confocal images of a representative prelamin A positive cell at late stage. c’ Single slice images of different Z-stacks illustrating the distribution of prelamin A, from surface (left panel) and from middle plane (right panel) point of views. Scale bar 10 µm. d Quantification of β-gal positive and prelamin A positive cells at early and late stages. Blue stained cells and Hoechst stained nuclei were counted in a minimum of six random fields to report the percentage β-gal positive cells at early and late stages. To count prelamin A positive and total cell numbers a minimum of six random fields was checked at early and late stages. Data are expressed as percentage of positive cells respect to the total cell count The manual count of the percentage of SA-β-gal or prelamin A positive cells over the total performed on the 3 MSC lines confirmed the qualitative results. The percentage of prelamin A-positive cells increases dramatically from early to late stages (from 6 to 59 %), with a trend similar to the one observed for the SA-β-gal staining (Fig. 2d). Though the trends were similar we can observe that the results from prelamin A staining indicate a slightly higher percentage of positive cells than for SA-β-gal assay both at early and late stage. The number of observations is insufficient to operate a statistical comparison, but the observed difference could indicate that the accumulation of prelamin A is a more sensitive marker to reveal senescent cells even at earlier stages. The detection of prelamin A positive cells at early stage is not entirely unexpected, as primary MSC culture are known to be composed by an heterogeneous population that might harbor among the active proliferating cells, resting, terminally differentiated or senescent cells (Sherley 2002; Whitfield et al. 2013). In addition, the variability among MSC lines observed for SA-β-gal counting despite we performed our analyses at stages selected to reduce the inter-donor variability (Fig. 1), is reduced in the prelamin A scoring, making the latest a more steady marker. The discordant results from the two markers might also be a consequence of the limitations and technical pitfalls inherent to the SA-β-gal assay caused by the limited specificity and instability of β-gal substrate, as already outlined by other investigators (Yang and Hu 2005; Lee et al. 2006; Debacq-Chainiaux et al. 2009). In this regard, the assessment of prelamin A accumulation can be performed with an easy available positive control, in order to check the technical quality of the immunostaining. The positive control for prelamin A consists in cells treated with mevinolin, a drug known to inhibit farnesyl production and to cause accumulation of unprocessed prelamin A (Lattanzi et al. 2014) (Additional file 1: Fig. S2). As expected from previous studies (Caron et al. 2007), treatment with mevinolin did not induce SA-β-gal expression (Additional file 1: Fig. S2). This is in agreement with our data showing that cells from centenarian individuals spontaneously accumulate prelamin A and display a more efficient response to oxidative stress-induced DNA damage (Lattanzi et al. 2014). These data support the view that prelamin A accumulation does not induce senescence per se, but it is triggered in response to stress stimuli in the attempt to counteract geroconversion. In particular, we observed in human fibroblast that the oxidative and replicative stress affect prelamin A processing, through the reduction of ZMPSTE24 expression (Lattanzi et al. 2014) and it could be speculated that a similar mechanism is involved in prelamin A accumulation observed in senescent MSC. Biochemical evaluation of prelamin A was also performed by Western blotting analysis. Prelamin A, in accordance with results obtained by prelamin A immunofluorescence staining, is detectable at the predicted molecular weight in both senescent and mevinolin-treated MSCs (Additional file 1: Fig. S3). We further checked the expression of p21Waf1 (Cdkn1n gene), a nuclear protein that indicates senescence-associated cell-cycle arrest (Kong et al. 2011), and Ki67, a marker of proliferating cells expressed in all active phases of the cell cycle (G1, S, G2). As hypothesized, cells positive for prelamin A showed p21Waf1 staining in nuclei and were negative for Ki67 (Fig. 3), confirming that prelamin A-labeled MSC underwent replicative senescence during in vitro culture.Fig. 3 Cell-cycle arrested cells are positive for prelamin A. Representative images of MSC immunostained for prelamin A (green), Ki67, and p21 (both red in the panels). Cell nuclei were counterstained with DAPI (blue) and merged images are shown in the last column. Prelamin A positive cells (arrows) are negative for Ki67 and positive for p21. Scale bar 10 µm Based on our observations, the nuclear accumulation of prelamin A identifies senescent cells in human MSC cultures. It is worth noting that previous reports indicate that the accumulation of mutated forms or precursors of lamin A triggers a series of molecular changes that diminish the specific properties of the MSC such as the multilineage differentiation potential (Yu et al. 2013; Malashicheva et al. 2015) and the capacity to promote the repair of injured tissues (Infante et al. 2014). Therefore, prelamin A accumulation can be therefore considered a candidate marker for the detection of senescent cells during MSC expansion and can be exploited to discard aged cells, characterized by low differentiation and regeneration capacity, and thus prevent their release for clinical purposes. Conclusions Despite the underlying molecular mechanisms are still unraveled, replicative senescence has evident consequences for cellular therapy. In this paper we have demonstrated that in human MSC cultured in vitro under standard growth condition, the detection of prelamin A identifies senescent MSC. Since the onset of senescence in a MSC culture is currently difficult to predict by the use of available markers, prelamin A staining could be successfully used to screen MSC populations before they are used for clinical applications alone or in combination with know and currently used senescence marker. Additional file 10.1186/s40064-016-3091-7 Supplementary data. Abbreviations SA-β-galsenescence associated β-galactosidase MSCmesenchymal stem cells Authors’ contributions CB designed and performed the experiments, analysed the data and wrote the manuscript. CC formulated the initial hypothesis, gave conceptual contribution to the study, performed the experiments and revised the manuscript. GL provided financial support, gave conceptual contribution to the study and revised the manuscript. DD provided financial support and performed the harvest of bone-marrow samples. EL gave conceptual contribution to the study and revised the manuscript. SD designed and coordinated the study, performed the experiments, analysed the data and wrote the manuscript. All authors read and approved the final manuscript. Acknowledgements The authors would like to thank the staff of the Third Orthopaedics and Traumatology Clinic (IOR, Bologna), for providing the cells used in this work, and Miss Lauren DeMaria BSc (Columbia University NY) for editorial assistance and English revision of the manuscript. The research was supported by Progetto FIRB-Accordi di programma 2010 (cod.RBAP10447J) of Italian Ministry of Education Universities and Research and by “5 per mille” 2011/2012 funding from Italian Ministry of Health. Competing interests The authors declare that they have no competing interests. Research involving human participants The study was reviewed and approved by the local Ethical Committee (Approval Record No. 21621). Patients signed an informed consent before the donation of bone marrow. ==== Refs References Banfi A Muraglia A Dozin B Proliferation kinetics and differentiation potential of ex vivo expanded human bone marrow stromal cells: implications for their use in cell therapy Exp Hematol 2000 28 707 715 10.1016/S0301-472X(00)00160-0 10880757 Broers J Ramaekers F Nuclear lamins: laminopathies and their role in premature ageing Physiol Rev 2006 86 967 1008 10.1152/physrev.00047.2005 16816143 Caron M Auclair M Donadille B Human lipodystrophies linked to mutations in A-type lamins and to HIV protease inhibitor therapy are both associated with prelamin A accumulation, oxidative stress and premature cellular senescence Cell Death Differ 2007 14 1759 1767 10.1038/sj.cdd.4402197 17612587 Coppe J-P Desprez P-Y Krtolica A Campisi J The senescence-associated secretory phenotype: the dark side of tumor suppression Annu Rev Pathol 2010 5 99 118 10.1146/annurev-pathol-121808-102144 20078217 Crowe EP Nacarelli T Bitto A Noguchi E Gadaleta MC Detecting senescence: methods and approaches Cell cycle control 2014 New York, NY Springer 425 445 Davies BS Coffinier C Yang SH Investigating the purpose of prelamin A processing Nucleus 2011 2 4 9 10.4161/nucl.13723 21647293 de Girolamo L Lucarelli E Alessandri G Mesenchymal stem/stromal cells: a new cells as drugs’’ paradigm. Efficacy and critical aspects in cell therapy Curr Pharm Des 2013 19 2459 2473 10.2174/1381612811319130015 23278600 de Jesus BB Blasco MA Assessing cell and organ senescence biomarkers Circ Res 2012 111 97 109 10.1161/CIRCRESAHA.111.247866 22723221 Debacq-Chainiaux F Pascal T Boilan E Screening of senescence-associated genes with specific DNA array reveals the role of IGFBP-3 in premature senescence of human diploid fibroblasts Free Radic Biol Med 2008 44 1817 1832 10.1016/j.freeradbiomed.2008.02.001 18329388 Debacq-Chainiaux F Erusalimsky JD Campisi J Toussaint O Protocols to detect senescence-associated beta-galactosidase (SA-betagal) activity, a biomarker of senescent cells in culture and in vivo Nat Protoc 2009 4 1798 1806 10.1038/nprot.2009.191 20010931 Geißler S Textor M Kühnisch J Functional comparison of chronological and in vitro aging: differential role of the cytoskeleton and mitochondria in mesenchymal stromal cells PLoS ONE 2012 7 e52700 10.1371/journal.pone.0052700 23285157 Infante A Gago A de Eguino GR Prelamin A accumulation and stress conditions induce impaired Oct-1 activity and autophagy in prematurely aged human mesenchymal stem cell Aging (Albany NY) 2014 1 1 17 Kim J Choi YS Lim S Comparative analysis of the secretory proteome of human adipose stromal vascular fraction cells during adipogenesis Proteomics 2010 10 394 405 10.1002/pmic.200900218 19953544 Koch CM Joussen S Schellenberg A Monitoring of cellular senescence by DNA-methylation at specific CpG sites Aging Cell 2012 11 366 369 10.1111/j.1474-9726.2011.00784.x 22221451 Kong Y Cui H Ramkumar C Zhang H Regulation of senescence in cancer and aging J Aging Res 2011 2011 963172 10.4061/2011/963172 21423549 Kosar M Bartkova J Hubackova S Senescence-associated heterochromatin foci are dispensable for cellular senescence, occur in a cell type- and insult-dependent manner and follow expression of p16ink4a Cell Cycle 2011 10 457 468 10.4161/cc.10.3.14707 21248468 Lattanzi G Ortolani M Columbaro M Lamins are rapamycin targets that impact human longevity: a study in centenarians J Cell Sci 2014 127 147 157 10.1242/jcs.133983 24155329 Lee BY Han JA Im JS Senescence-associated β-galactosidase is lysosomal β-galactosidase Aging Cell 2006 5 187 195 10.1111/j.1474-9726.2006.00199.x 16626397 Malashicheva A Bogdanova M Zabirnyk A Various lamin A/C mutations alter expression profile of mesenchymal stem cells in mutation specific manner Mol Genet Metab 2015 Mattioli E Columbaro M Capanni C Drugs affecting prelamin A processing: effects on heterochromatin organization Exp Cell Res 2008 314 453 462 10.1016/j.yexcr.2007.11.012 18093584 Ragnauth CD Warren DT Liu Y Prelamin A acts to accelerate smooth muscle cell senescence and is a novel biomarker of human vascular aging Circulation 2010 121 2200 2210 10.1161/CIRCULATIONAHA.109.902056 20458013 Righolt CH Van’t Hoff MLR Vermolen BJ Robust nuclear lamina-based cell classification of aging and senescent cells Aging (Albany NY) 2011 3 1192–1201 1004 Scaffidi P Misteli T Lamin A-dependent misregulation of adult stem cells associated with accelerated ageing Nat Cell Biol 2008 10 452 459 10.1038/ncb1708 18311132 Schellenberg A Lin Q Schüler H Replicative senescence of mesenchymal stem cells causes DNA-methylation changes which correlate with repressive histone marks Aging (Albany NY) 2011 3 873 888 10.18632/aging.100391 22025769 Sepúlveda JC Tomé M Fernández ME Cell senescence abrogates the therapeutic potential of human mesenchymal stem cells in the lethal endotoxemia model Stem Cells 2014 Shay JW Wright WE Hayflick, his limit, and cellular ageing Nat Rev Mol Cell Biol 2000 1 72 76 10.1038/35036093 11413492 Sherley JL Asymmetric cell kinetics genes: the key to expansion of adult stem cells in culture Stem Cells 2002 20 561 572 10.1634/stemcells.20-6-561 12456964 Shibata KR Aoyama T Shima Y Expression of the p16INK4A gene is associated closely with senescence of human mesenchymal stem cells and is potentially silenced by DNA methylation during in vitro expansion Stem Cells 2007 25 2371 2382 10.1634/stemcells.2007-0225 17569790 Stenderup K Justesen J Clausen C Kassem M Aging is associated with decreased maximal life span and accelerated senescence of bone marrow stromal cells Bone 2003 33 919 926 10.1016/j.bone.2003.07.005 14678851 Vlcek S Foisner R Lamins and lamin-associated proteins in aging and disease Curr Opin Cell Biol 2007 19 298 304 10.1016/j.ceb.2007.04.001 17466505 Wagner W Horn P Castoldi M Replicative senescence of mesenchymal stem cells: a continuous and organized process PLoS ONE 2008 3 e2213 10.1371/journal.pone.0002213 18493317 Wagner W Bork S Horn P Aging and replicative senescence have related effects on human stem and progenitor cells PLoS ONE 2009 4 e5846 10.1371/journal.pone.0005846 19513108 Whitfield MJ Lee WCJ Van Vliet KJ Onset of heterogeneity in culture-expanded bone marrow stromal cells Stem Cell Res 2013 11 1365 1377 10.1016/j.scr.2013.09.004 24103495 Wuchter P Bieback K Schrezenmeier H Standardization of Good Manufacturing Practice-compliant production of bone marrow-derived human mesenchymal stromal cells for immunotherapeutic applications Cytotherapy 2014 17 128 139 10.1016/j.jcyt.2014.04.002 24856898 Yang NC Hu ML The limitations and validities of senescence associated-β- galactosidase activity as an aging marker for human foreskin fibroblast Hs68 cells Exp Gerontol 2005 40 813 819 10.1016/j.exger.2005.07.011 16154306 Yoo JK Choi S-J Kim JK Expression profiles of subtracted mRNAs during cellular senescence in human mesenchymal stem cells derived from bone marrow Exp Gerontol 2013 Yu K-R Kang K-S Aging-related genes in mesenchymal stem cells: a mini-review Gerontology 2013 Yu KR Lee S Jung JW MicroRNA-141-3p plays a role in human mesenchymal stem cell aging by directly targeting ZMPSTE24 J Cell Sci 2013 126 5422 5431 10.1242/jcs.133314 24101728

PMC005xxxxxx/PMC5001960.txt

==== Front AMB ExpressAMB ExpressAMB Express2191-0855Springer Berlin Heidelberg Berlin/Heidelberg 23410.1186/s13568-016-0234-8Original ArticleAdaptation to low pH and lignocellulosic inhibitors resulting in ethanolic fermentation and growth of Saccharomyces cerevisiae Narayanan Venkatachalam venkatachalam.narayanan@tmb.lth.se 1Sànchez i Nogué Violeta violeta.sanchezinogue@nrel.gov 2van Niel Ed W. J. ed.van_niel@tmb.lth.se 1Gorwa-Grauslund Marie F. marie-francoise.gorwa-grauslund@tmb.lth.se 11 Division of Applied Microbiology, Department of Chemistry, Lund University, P.O. Box 124, 221 00 Lund, Sweden 2 National Bioenergy Center, National Renewable Energy Laboratory, 15013 Denver West Parkway, Golden, CO 80401 USA 26 8 2016 26 8 2016 2016 6 1 596 7 2016 18 8 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Lignocellulosic bioethanol from renewable feedstocks using Saccharomyces cerevisiae is a promising alternative to fossil fuels owing to environmental challenges. S. cerevisiae is frequently challenged by bacterial contamination and a combination of lignocellulosic inhibitors formed during the pre-treatment, in terms of growth, ethanol yield and productivity. We investigated the phenotypic robustness of a brewing yeast strain TMB3500 and its ability to adapt to low pH thereby preventing bacterial contamination along with lignocellulosic inhibitors by short-term adaptation and adaptive lab evolution (ALE). The short-term adaptation strategy was used to investigate the inherent ability of strain TMB3500 to activate a robust phenotype involving pre-culturing yeast cells in defined medium with lignocellulosic inhibitors at pH 5.0 until late exponential phase prior to inoculating them in defined media with the same inhibitor cocktail at pH 3.7. Adapted cells were able to grow aerobically, ferment anaerobically (glucose exhaustion by 19 ± 5 h to yield 0.45 ± 0.01 g ethanol g glucose−1) and portray significant detoxification of inhibitors at pH 3.7, when compared to non-adapted cells. ALE was performed to investigate whether a stable strain could be developed to grow and ferment at low pH with lignocellulosic inhibitors in a continuous suspension culture. Though a robust population was obtained after 3600 h with an ability to grow and ferment at pH 3.7 with inhibitors, inhibitor robustness was not stable as indicated by the characterisation of the evolved culture possibly due to phenotypic plasticity. With further research, this short-term adaptation and low pH strategy could be successfully applied in lignocellulosic ethanol plants to prevent bacterial contamination. Electronic supplementary material The online version of this article (doi:10.1186/s13568-016-0234-8) contains supplementary material, which is available to authorized users. Keywords Saccharomyces cerevisiaeLow pHLignocellulosic inhibitorsPhenotypic robustnessAdaptationEthanol yieldSwedish National Energy AgencyP35350-1Gorwa-Grauslund Marie F. issue-copyright-statement© The Author(s) 2016 ==== Body Introduction Increasing concerns over the need for sustainable and scalable fuels as a means to curb global warming has led to focus on bioethanol production from renewable biomass, such as agricultural and industrial residues (Limayem and Ricke 2012). Due to the decrease in costs of petroleum as a response to recent discoveries of fossil fuel reserves, there is an intense emphasis on lowering the costs of renewable bioethanol by overcoming challenges connected to high substrate costs, low titers and low production rates accompanied by low yields (Papoutsakis and Pronk 2015). A significant challenge in the fuel ethanol production is acute and chronic bacterial contamination, since the incoming substrate might contain microorganisms and the fermentation is carried out in non-aseptic conditions (Skinner and Leathers 2004). Bacterial contamination is predominantly due to lactic and acetic acid bacteria leading to loss of fermentable sugars and micronutrients, increased by-product formation (lactic acid and acetic acid), reduced ethanol yields and productivities and stuck fermentations (Beckner et al. 2011; Bischoff et al. 2009). Bacterial contamination has been studied extensively (Bischoff et al. 2009; Skinner and Leathers 2004) and several antimicrobial strategies, including usage of antibiotics, have been adopted in the first generation bioethanol production (Muthaiyan et al. 2011). These methods are expensive and some are environmentally invasive when used in large-scale fermentations (Muthaiyan et al. 2011). Contamination might pose a bigger threat to lignocellulosic ethanol owing to the versatility in sugar substrates. As Saccharomyces cerevisiae displays glucose repression, it takes longer time to assimilate other sugars including xylose and arabinose. Thus any contamination might be able to utilise the other sugars swiftly and more efficiently than S. cerevisiae leading to reduction in ethanol production. Several attempts have been made to study and control bacterial contamination in lignocellulosic ethanol production including: (1) adding NaCl and ethanol to wood hydrolysate (Albers et al. 2011), (2) high solid loading in simultaneous saccharification and fermentation (SSF) (Ishola et al. 2013), (3) usage of an antibiotic like gentamicin and biomass autoclaving (Serate et al. 2015), and (4) usage of bacteriophages (Worley-Morse et al. 2015). These strategies encounter challenges including: (1) additional cost and need for extensive fine tuning and testing of concentrations of NaCl and ethanol (Albers et al. 2011), (2) loss of cell viability due to mechanical stress caused by solid particles in high cell loading (Ishola et al. 2013), (3) cost and environmental challenges posed by gentamicin, energy expenditure and formation of inhibitors due to autoclaving (Serate et al. 2015), and (4) rise of bacteriophage-insensitive mutants and possibilities of gene transfer from bacteriophages to yeast (Worley-Morse et al. 2015). One of the potentially scalable and economically feasible solutions to control bacterial contamination is to run the lignocellulosic fermentation at low pH, around pH 4 where the growth and viability of bacteria are drastically reduced (Kádár et al. 2007). Additionally, yeast cells are recycled in several commercial ethanol production processes up to 6 months to reduce fermentation time and cost of yeast propagation, increasing the chances of contamination (Basso et al. 2011). To prevent contamination, yeast cells are treated with dilute sulphuric acid (H2SO4) at pH between pH 1.8 and 2.5 for 1–2 h (Basso et al. 2011), which results in reduction in intracellular pH (Beales 2004), yeast viability and low ethanol yield (De Melo et al. 2010). Hence, it might be efficient to develop S. cerevisiae strains tolerant to lower pH induced by inorganic acids. Apart from bacterial contamination, inhibitors pose another obstacle to yeast in ethanol production, formed from the components of lignocellulose including cellulose, hemicellulose and lignin due to the harsh conditions of biomass pre-treatment (Almeida et al. 2007). They include (1) weak organic acids such as acetic acid, formic acid and levulinic acid, (2) furans, including furfural and hydroxymethylfurfural (HMF), and (3) phenolic compounds such as vanillin, coniferyl aldehyde and 4-hydroxybenzoic acid (Palmqvist and Hahn-Hägerdal 2000; Taherzadeh and Karimi 2011). S. cerevisiae endures inhibitors through different mechanisms, including detoxification by enzymatic reduction, efflux and membrane repair (Piotrowski et al. 2014). Advancements in pre-treatment processes resulted in reduction of furans, phenolics, formic acid and levulinic acid concentrations in the hydrolysate (Jönsson et al. 2013). However, acetic acid is naturally bound to lignocellulose in the form of acetyl sugars in the hemicellulose fraction and becomes de-acetylated during the hydrolysis treatment (Almeida et al. 2007). As a weak organic acid, its effect is more pronounced at low pH and may facilitate synergy between furans and phenolics (Ding et al. 2011). Hence, it is crucial to focus on yeast tolerance in acidic environments inflicted by the combination of inorganic and weak organic acids in the presence of lignocellulosic inhibitors for cost competitive ethanol production. Different rational engineering strategies have been pursued with S. cerevisiae to understand the molecular mechanisms involved in coping with one or several inhibitors, thereby creating inhibitor tolerant S. cerevisiae strains (Alriksson et al. 2010; Caspeta et al. 2015; Koichi et al. 2012; Lei et al. 2011; Liu 2011; Parawira and Tekere 2011; Taherzadeh and Karimi 2011; Takuya et al. 2013). As Meijnen et al. (2016) found that tolerance towards acetic acid is a result of a polygenic response from yeast, evolutionary adaptation might be a suitable strategy to improve tolerance towards low pH and acetic acid with other lignocellulosic inhibitors since yeast might accrue beneficial properties under stress conditions over the time of evolution. Evolutionary engineering strategies have been successfully pursued to obtain yeast strains with enhanced tolerance against individual or combinations of several inhibitors in defined media (Dominik and Uwe 2008; Wright et al. 2011) or in hydrolysates (Almario et al. 2013; Hanqi et al. 2014) with beneficial properties including better growth, improved viability, higher yield and ethanol productivity in comparison to the control strains. Yeast tolerance towards inhibitors could also be induced by pre-cultivation with lower concentration of inhibitors in defined media or diluted hydrolysates. This induces the general stress response leading to improved growth and fermentation performance in the inhibitory medium or hydrolysate (Nielsen et al. 2015; Tomás-Pejó and Olsson 2015). Kádár et al. (2007) had improved the tolerance of S. cerevisiae towards lignocellulosic inhibitors at pH 4, but with lower concentration of inhibitors to potentially threaten yeast growth and fermentation. Lowering the culture pH to pH 4 in the presence of different concentrations of acetic acid in corn-mash has resulted in complete inhibition of ethanol production at acetic acid concentrations greater than 0.8 % weight/volume (w/v) (Graves et al. 2006). Yet, other industrial strains have been analysed for their low pH tolerance, such as JP1 and PE-2 from commercial ethanol production in Brazil (De Melo et al. 2010; Della-Bianca et al. 2014). We have previously demonstrated a strain-independent pre-cultivation strategy where S. cerevisiae cells can grow and ferment at pH 3.7 with lethal concentrations of 6 g L−1 acetic acid after a short-term adaptation with 6 g L−1 acetic acid at pH 5.0 (Sànchez i Nogué et al. 2013). Though numerous endeavours have been pursued to develop a robust S. cerevisiae strain in the presence of inhibitors by targeted, evolutionary and pre-cultivation approaches, to our knowledge, there has been no investigation on improving phenotypic robustness of S. cerevisiae to low pH with acetic acid and other lignocellulosic inhibitors. Herein, we aimed at developing a short-term adaptation strategy and an ALE of yeast in a chemostat to investigate the nature of adaptability of S. cerevisiae to the harsh conditions of low pH and lignocellulosic hydrolysates and the stability of the acquired robustness. Furthermore, understanding the interactive effects among the inhibitors at low pH could pave the way for developing new strains and strategies to lignocellulosic ethanol production. Materials and methods Yeast strain and media The commercial S. cerevisiae brewer’s strain, Coobra 6 Magnum (CBF Drinkit AB, Mölndal, Sweden) was renamed to TMB3500 (Almeida et al. 2009a) and used in this study. It was stored at −80 °C in yeast peptone dextrose (YPD) medium containing 10 g L−1 yeast extract, 20 g L−1 peptone and 20 g L−1 glucose supplemented with 30 % (v/v) glycerol and maintained on YPD medium with 20 g L−1 agar. All the chemicals were purchased from Sigma Aldrich, Sweden, unless mentioned otherwise. A chemically defined medium (Verduyn et al. 1992) with 20 g L−1 glucose, buffered with 50 mM potassium hydrogen phthalate and 20 mM potassium hydroxide (KOH) (Hahn-Hägerdal et al. 2005) was used in all aerobic growth experiments. The pH of the defined medium was adjusted using 3 M H2SO4 and 3 M KOH. Ergosterol and Tween80 were added in anaerobic experiments, in the final concentrations of 0.01 and 0.42 g L−1, respectively. A silicone based antifoam (0.5 mL L−1) was added in the experiments performed in fermentors to avoid excessive foaming (Dow Corning Antifoam RD emulsion, VWR International Ltd., Poole, UK). Compounds including 6 g L−1 acetic acid, 1.5 g L−1 furfural, 0.5 g L−1 HMF and 1 g L−1 vanillin were used as inhibitors in this study, hereafter mentioned as inhibitor cocktail (IC), unless mentioned otherwise. The concentrations of different inhibitors chosen in this study were in the ranges found in different pre-treated lignocellulose hydrolysates obtained from barley straw, dilute spruce and wheat straw (Almeida et al. 2009a). The used vanillin concentration of 1 g L−1 was 5–10 times higher than in the lignocellulosic hydrolysate to account for different kinds of phenolic compounds and lignosulfonates. Defined media were chosen over hydrolysates to have better control over the experiments performed since hydrolysate might contain many unknown inhibitors (Palmqvist and Hahn-Hägerdal 2000) and they might change in composition over time. All media components were sterile filtered to avoid changes in composition due to evaporation. Culture conditions Aerobic cultures were performed at 30 °C in a rotary shake-incubator (New Brunswick, Enfield, CT, USA) at 180 rpm with cell concentrations determined as optical density (OD) at 620 nm (Spectrophotometer U-1800, Hitachi, Berkshire, UK). Seed cultures were grown from single colonies of TMB3500 (YPD agar plate) in 5 mL defined medium in a 50 mL conical tube to reach late exponential phase. Pre-cultures were started from the seed culture in defined medium with an initial OD of 0.5, grown till late exponential phase, unless mentioned otherwise. All the aerobic batch cultures were cultivated in baffled shake flasks with a medium volume equivalent to 10 % of the volume of the baffled shake flask to maintain adequate aeration. Cells for inoculation were obtained after centrifuging the pre-culture at 4000 rpm for 5 min at 4 °C, washing the cells with saline and repeating the centrifugation process. Gas proof neoprene tubes (Masterflex™, Cole-Parmer, Sweden) were used for connections in the anaerobic experiments to avoid oxygen diffusion. All the growth and fermentation experiments except the ALE were carried out at least in biological replicates and measurements were carried out in technical triplicates. Data represented in figures include standard deviations from the replicates. Short-term adaptation Pre-cultures were grown aerobically until late exponential phase in 25 mL of defined medium with the IC at pH 5.0 from the seed culture, termed as short-term adaptation step being used for subsequent cultivations. Cells grown until late exponential phase in defined medium without the IC at pH 5.0 was used as negative control, termed as non-adapted cells. Aerobic batch growth in short-term adaptation experiments were followed for 5 days. Aerobic batch growth Short-term adapted TMB3500 cells were inoculated into 25 mL of defined media at pH 3.7 with three different inhibitor combinations: (1) 6 g L−1 acetic acid and 0.5 g L−1 HMF; (2) 6 g L−1 acetic acid and 1.5 g L−1 furfural; and (3) 6 g L−1 acetic acid and 1 g L−1 vanillin. Defined media (25 mL) with the IC at different pH values (pH 5.0, 4.5, 4.0 and 3.7) were inoculated with short-term adapted cells of strain TMB3500 on defined medium. Short-term adapted TMB3500 cells were inoculated at different cell dry weights (gdw L−1) namely 0.5, 1 and 3 gdw L−1 into 25 mL of defined medium with the IC at pH 3.7. Anaerobic fermentation Short-term adapted and non-adapted TMB3500 cells were inoculated at 3 gdw L−1 cells into 500 mL of defined medium with the IC at pH 3.7 in a 1 L Infors fermentor (InforsHT, Switzerland). Anaerobic conditions were obtained by sparging with nitrogen gas (200 mL min−1), the stirring was set at 200 rpm and the temperature was maintained at 30 °C. The fermentation profile was followed by sampling for metabolites, residual inhibitors and OD. Adaptive lab evolution of TMB3500 A pre-culture of strain TMB3500 was used to inoculate an aerobic batch of 1 L defined medium without inhibitors at pH 5.0 in 1.4 L Infors fermentors (InforsHT, Switzerland). The culture was operated at 30 °C, with a stirring rate of 200 rpm and air was sparged at a flow rate of 200 mL min−1. At the end of the exponential phase, the feed containing defined medium with the IC was connected to the fermentor at a dilution rate of 0.1 h−1 and the fermentor was rendered anaerobic by sparging nitrogen gas at a flow rate of 200 mL min−1. After the culture reached steady state, the pH was reduced by 0.2 units using 3 M H2SO4. A 1.4 L Infors fermentor stirring at 200 rpm and nitrogen gas sparged at 200 mL min−1 was used between the fermentor with yeast culture and feed bottle to facilitate a slow transition to the new pH. The culture pH was maintained using 3 M KOH. Once the pH was reduced to 4.5, the dilution rate was increased to 0.15 h−1. Gradual reduction in pH was continued until it reached a value of 3.7. The experiment was carried out for 3600 h thereby obtaining 709 generations. Samples were taken for OD and metabolite analysis. Cell dry weight analyses were performed when the culture was in steady state. At the end of the evolution experiment, the cell suspension was transferred to a new chemostat running under the same conditions. Characterisation of evolved population The evolved population (CC156) at pH 3.7 in the ALE and the parental strain TMB3500 were compared for their robustness towards low pH and acetic acid with other inhibitors. In addition, a biofilm was formed when the pH of the chemostat was reduced to 4.1. The biofilm and CC156 population were collected at the end of the chemostat culture and stored as glycerol stocks. They were streaked on YPD plates to obtain single colonies for DNA fingerprinting and used as a population from glycerol stock in the pre-cultures for liquid media growth experiments. Agar plates were prepared by mixing autoclaved 20 g L−1 agar with filter sterilised chemically defined media of different conditions including pH 5.0 and 3.7 (with or without the IC) and pH 4.5 (50 % of the IC). Cell suspensions of strain CC156 and strain TMB3500 (100 μL) were streaked evenly on the agar plates and incubated at 30 °C aerobically and anaerobically for 3–7 days. Experiments were carried out in triplicate to enumerate the colony forming units (CFUs). Aerobic and anaerobic liquid batch growth experiments were carried out with pre-cultures of strain TMB3500 (from a single colony) and the evolved population CC156. Different conditions were tested in 25 mL defined media, including pH 5.0 and 3.7 (with and without the IC) and pH 4.5 (50 % of the IC). Cells were grown at 30 °C in baffled shake flasks and sealed glass vials with magnetic stirrers and rubber stoppers to follow aerobic and anaerobic growth, respectively. Cells of CC156 population, biofilm and strain TMB3500 from the short-term adaptation step were inoculated into 25 mL of defined medium with inhibitors at the concentrations similar to the residual inhibitor concentrations present in the CC156 population of the ALE chemostat at pH 3.7 i.e. 0.32 g L−1 HMF, 0.46 g L−1 furfural, 0.53 g L−1 vanillin, 6.0 g L−1 acetic acid, and this aerobic batch growth was followed for 2 days. Cells of strain CC156, biofilm and strain TMB3500 were propagated in 5 mL of YPD liquid medium containing 10 g L−1 yeast extract, 20 g L−1 peptone and 20 g L−1 glucose. Genomic DNA was extracted (Harju et al. 2004) from three colonies in each strain and amplified by PCR (C1000 Touch™ thermocycler, Bio-rad, USA) using Dream taq polymerase (Life technologies, Sweden) and selected primers targeting (1) TY1, TY3 elements (Transposable elements, individually and combined) (i Nogué et al. 2012; Schofield et al. 1995). TY elements were chosen due to their presence in a wide variation in the yeast genome distribution making them ideal for intraspecies discrimination (Schofield et al. 1995), (2) randomly amplified polymorphic DNA (S1254 random primer) (Akopyanz et al. 1992) and (3) (GACA)4 (Andrade et al. 2006) and (GTG)5 repeats (da Silva-Filho et al. 2005). They were chosen owing to their usefulness in differentiating a strain during the evolution monitoring process in wine fermentation (da Silva-Filho et al. 2005). PCR was performed adopting to conditions from the respective literature. The PCR products were separated in an agarose gel (0.8 %) at 100 V for 60 min with a gene ruler DNA ladder (100 bp–10 kb) and gene ruler 100 bp plus ladder (100 bp–3 kb) (Thermo Scientific, Sweden) as standards. Metabolite analysis Cell dry weight was determined in triplicate by filtering 5 mL of the culture on a pre-weighed 0.45 μm pore size Supor® membrane disc filter (Pall Corporation, Port Washington, NY, USA). Filters were washed with distilled water and dried for 8 min at 350 W in a microwave oven. To analyze the metabolites, cells were separated by centrifugation at 13,200 rpm for 2 min; the supernatant was filtered through 0.20 μm membrane filters (Toyo Roshi Kaish, Tokyo, Japan) and stored at −20 °C until analysis. Concentrations of glucose, glycerol, acetate, ethanol, HMF, furfural and vanillin were determined by high performance liquid chromatography (Waters, Milford, MA, USA) using a HPX-87H resin-based column (Bio-Rad, Hercules, CA, USA) preceded by a Micro-Guard Cation-H guard column (Bio-Rad). Separation was performed at 45 °C with 5 mM H2SO4 at a flow rate of 0.6 mL min−1. All compounds were quantified by refractive index detection (Shimadzu, Kyoto, Japan). For each HPLC run, a seven-point calibration curve was made for each compound. Results In this study, the significance of pre-cultivation and ALE towards low pH and inhibitor tolerance were explored. The TMB3500 yeast strain employed in this study was previously shown to possess high tolerance to lignocellulosic inhibitors by proliferating in the presence of high amounts of non-detoxified hydrolysates, including barley straw (40 % w/v), dilute spruce (60 % w/v) and wheat straw hydrolysates (60 % w/v) (Almeida et al. 2009a). Robust phenotype through short-term adaptation and role of individual inhibitors at low pH To test the influence of short-term adaptation on tolerance to low pH and individual inhibitors, TMB3500 cells short-term adapted with the IC at pH 5.0 and non-adapted cells were inoculated in defined media at pH 3.7 with acetic acid and individual inhibitors. Short-term adapted TMB3500 was able to grow at pH 3.7 with acetic acid and HMF at a maximum specific growth rate (μmax) of 0.09 ± 0.01 h−1 without any lag phase (Fig. 1). Cells grown at pH 3.7 with acetic acid and furfural had a μmax of 0.10 ± 0.01 h−1 with a lag phase of 13 ± 0.5 h whereas cells grown at pH 3.7 with acetic acid alone had a similar μmax (0.11 ± 0.01 h−1), but without any lag phase (Sànchez i Nogué et al. 2013). This indicates that furfural had a combined inhibitory effect along with acetic acid. Short-term adapted cells inoculated at pH 3.7 with acetic acid and vanillin had just marginal growth of 0.01 ± 0.00 h−1 after 160 h (Fig. 1). Similarly, minimal growth of 0.01 ± 0.00 h−1 was observed in the culture at pH 3.7 with all the inhibitors. Minimal to no growth was observed with non-adapted cells in any of the above conditions, hence subsequent aerobic batch experiments were performed only with short-term adaptation.Fig. 1 Effect of low pH with 6 g L−1 acetic acid and individual inhibitors (1.5 g L−1 furfural or 0.5 g L−1 HMF or 1 g L−1 vanillin) on short-term adapted (black) and non-adapted (white) cells. pH 3.7 with inhibitors includes the IC Combined effect of different pH and multiple inhibitors on growth To further map the inhibitory effect of low pH and acetic acid in the presence of a cocktail of lignocellulosic inhibitors, a short-term adapted TMB3500 culture was grown at different pH between 3.7 and 5.0 in media with the IC. Whereas the cells at pH 5.0 started growing almost immediately with a μmax of 0.20 ± 0.00 h−1, a decrease in pH by 0.5 units resulted in a long lag phase (20 ± 1 h) and a reduction in μmax to 0.08 ± 0.00 h−1. At pH 4.0 there was just a marginal growth and no growth was observed at pH 3.7 even after 150 h (Fig. 2). This illustrates the critical role played by pH in the presence of acetic acid and inhibitors on yeast growth.Fig. 2 Effect of the pH on growth of strain TMB3500 after short-term adaptation at pH 5 with the IC Influence of initial cell density in growth performance at low pH with inhibitors To investigate whether the observed growth inhibition could be relieved by increasing the initial cell density and if there was a critical cell concentration that allowed for growth to occur, short-term adapted TMB3500 cultures were inoculated at different cell concentrations (0.5, 1 and 3 gdw L−1) in a medium at pH 3.7 with the IC. When inoculated with 3 gdw L−1, cells grew at a μmax of 0.04 ± 0.00 h−1 (Fig. 3) in comparison to an inoculum size of 0.5 or 1 gdw L−1, where no growth was observed even after 130 h of incubation. The short-term adapted biomass at higher concentration might possess the required volumetric reductase activity to efficiently detoxify HMF and furfural to their corresponding alcohols (Modig et al. 2008) reaching inhibitor threshold levels for allowing growth at the low pH in presence of acetic acid.Fig. 3 Effect of initial cell density on growth and inhibitor tolerance of strain TMB3500 at pH 3.7 with the IC after short-term adaptation. Cell dry weight-3 gdw L−1 (triangles), 1 gdw L−1 (diamonds) and 0.5 gdw L−1 (squares) Effect of initial cell density in fermentation performance at low pH with inhibitors To test the influence of the short-term adaptation strategy to maintain metabolic activity and the ability to ferment at pH 3.7 in the presence of IC, an anaerobic batch culture was inoculated with 3 gdw L−1 of a TMB3500 culture with (Fig. 4a) and without pre-adaptation (Fig. 4b). In the pre-adapted system, glucose was consumed completely within 19 ± 5 h at a specific consumption rate of 0.47 ± 0.01 g glucose g cells−1 h−1 to produce 0.45 ± 0.01 g ethanol g glucose−1 (Table 1). There was a significant detoxification of inhibitors including 65 ± 10 % of HMF, 85 ± 12 % of furfural and 60 ± 1 % of vanillin to their respective alcohols (Fig. 5). In the non-adapted system, the cell concentration started to decrease and there was no significant sugar consumption or ethanol production even after 110 h and no inhibitor detoxification, except for 75 ± 12 % of furfural (Fig. 5; Table 1).Fig. 4 Anaerobic batch fermentation of strain TMB3500 with 3 gdw L−1 cells in pH 3.7 with the IC with (a) and without (b) short-term adaptation. Cell dry weight (squares), glucose (diamonds), ethanol (circles), acetate (triangles) and glycerol (crosses) Table 1 Anaerobic batch fermentation at pH 3.7 with the IC with and without short-term adaptation at pH 5 Condition Ethanol titre (g L−1) Yield (g g glucose−1) Specific consumption/production rate (g g−1 h−1) Biomass Glycerol Acetate Ethanol Glucose Ethanol Short-term adapted 8.93 ± 0.38 0.01 ± 0.00 0.01 ± 0.01 0.00 ± 0.01 0.45 ± 0.01 −0.47 ± 0.01 0.18 ± 0.02 Non-adapted 0.73 ± 0.84 −0.58 ± 0.56 0.03 ± 0.04 0.30 ± 0.15a 0.15 ± 0.13a 0.01 ± 0.00 0.00 ± 0.00 aYield values based on glucose consumption of 1.80 ± 2.11 g in total Fig. 5 Inhibitor conversion profile in anaerobic batch fermentation with (black) and without (white) short-term adaptation ALE of TMB3500 for tolerance towards low pH with acetic acid and other inhibitors Though short-term adaptation is successful in obtaining a stable growth and fermentation profile at low pH under inhibitory conditions, a stable robust yeast strain that does not require short-term adaptation towards low pH would be more ideal. Cells of strain TMB3500 were evolved to improve their tolerance towards low pH with inhibitory concentrations of acetic acid (6 g L−1) and other inhibitors typically present in lignocellulose hydrolysates including furfural (1.5 g L−1), HMF (0.5 g L−1) and vanillin (1 g L−1). They were grown in an anaerobic continuous culture at a dilution rate of 0.1 h−1 in a defined medium with the IC at pH 5.0. Once the culture attained steady state, the pH was reduced stepwise from pH 5.0 to 3.7. At pH 4.5, cells were further stressed to grow faster by increasing the dilution rate to 0.15 h−1. A biofilm which was formed at pH 4.1 in the chemostat covering the glass walls and baffles increased proportionally during further reduction in pH. At pH 3.7, the cell culture CC156 growing in the presence of lignocellulosic inhibitors was obtained after 709 generations (Fig. 6a, b). The cell suspension of the chemostat was transferred into a new chemostat operating under similar conditions to test its capacity for inhibitor tolerance at pH 3.7 and to identify the role of the biofilm in ALE. The chemostat culture washed out (data not shown) and the CC156 cells stored at −80 °C were chosen as the evolved population.Fig. 6 a ALE of strain TMB3500. Optical density (solid squares) and cell dry weight (spotted triangles). Dotted vertical lines indicate a change in pH. b Metabolite profile of ALE of strain TMB3500. Glucose (diamonds), ethanol (circles), acetate (triangles) and glycerol (crosses) The increase in un-dissociated acetic acid concentration in the chemostat due to stepwise lowering the pH led to a series of effects over the course of cultivation: (1) From pH 5.0 to 4.5 an initial decrease in optical density was followed by recovery of the culture, but at lower pH, the optical density gradually decreased (Fig. 6a); (2) this was accompanied by a temporary increase in the glucose concentration due to a lower glucose rate of consumption. After each pH change, the glucose concentration increased sequentially in the new steady states (Fig. 6b); (3) HMF, furfural and vanillin most probably have been detoxified by NAD(P)H-dependent reductases in strain TMB3500 as observed by Modig et al. (2008) and furfural was detoxified with a conversion efficiency of 75 % on average throughout the chemostat cultivation and HMF and vanillin were being detoxified in varying efficiencies based on the cell concentrations at the given time point (Fig. 7). On the other hand, the ethanol yield was maintained at 0.42 ± 0.04 g ethanol g glucose−1 throughout the evolution process in spite of the increase in dilution rate and decrease in pH. Moreover, the concentration of acetate remained constant at 6 g L−1, as it could not be consumed as a source of carbon by S. cerevisiae under anaerobic conditions (Henningsen et al. 2015) (Fig. 6b).Fig. 7 Inhibitor conversion profile throughout the ALE at the start and end of every pH shift. (% conversion, calculated based on initial and measured time point). HMF (white), furfural (grey) and vanillin (black) Characterisation of long-term evolution Characterisation of the CC156 population from the chemostat was performed in an attempt to isolate a stable strain displaying robustness towards low pH with inhibitors in comparison to the parental strain, TMB3500. To isolate a strain displaying an inhibitor tolerant phenotype at low pH, cells of CC156 and TMB3500 strains were plated in defined solid media with different concentrations of inhibitors and pH without short-term adaptation. Surprisingly, both the strains displayed similar colony growth characteristics: (1) colonies formed in less inhibitory conditions including defined media at pH 5.0, 3.7 and 4.5 with 50 % of the IC. (2) No colonies were formed in severe inhibitory conditions including pH 5.0 or 3.7 with the IC, indicating that the evolved population could not retain the phenotypic robustness against inhibitors. To validate the strain robustness, cells from the CC156 population and TMB3500 were grown in chemically defined liquid media with different pH and concentrations of inhibitors, aerobically and anaerobically. However, growth characteristics for both the parental and evolved strain were similar in liquid media as well. Cells were considered to have positive growth characteristics if their OD value had doubled at least threefold. Based on this, there was a difference between the growth patterns in solid and liquid media where growth was displayed in low to medium inhibitory conditions, including chemically defined media at pH 5.0, 3.7 and 4.5 with 50 % of the IC and pH 5.0 with IC. Nevertheless, both parental and adapted strain did not grow in the medium with pH 3.7 and IC like in the severest conditions applied in the chemostat. Since the evolved strain did not display inhibitor robustness in conditions of the ALE chemostat, it was hypothesised that the thick biofilm generated over the course of the long-term adaptation could have a role in detoxifying the medium and cell proliferation, whereas the free cells present in the suspension also might be growing and fermenting, tolerating the residual inhibitors. To verify this hypothesis and to test whether short-term adaptation helps to retain tolerance in the evolved strain, cells from the biofilm, CC156 and TMB3500 were inoculated in liquid media with inhibitor concentrations similar to the residual inhibitor concentrations in the liquid suspension of the ALE chemostat. Unexpectedly, strain TMB3500 could grow at a maximum specific growth rate of 0.07 ± 0.00 h−1 with a lag phase of 12 ± 0.5 h, biofilm cells and the CC156 population did not grow even after 150 h. To investigate if large rearrangements in the genome had occurred, DNA fingerprinting was carried out with different sets of primers in the cells from the biofilm, population CC156 and strain TMB3500. Interestingly, the gel bands were similar for all analysed strains (Additional file 1: Figure S1) indicating that there were no major genetic re-arrangements in the evolved strain in comparison to the parental strain. ALE of S. cerevisiae TMB3500 strain resulted in a population accompanied by a biofilm displaying robustness at pH 3.7 with the IC coupled with efficient ethanol production. However, when the selection pressure was removed through storage, the parental strain and the evolved population had similar growth characteristics under the various inhibitory conditions. This led to the conclusion that the robustness towards low pH and lignocellulosic inhibitors displayed by the evolved strain from ALE might be a result of population heterogeneity. Discussion Developing inhibitor tolerant S. cerevisiae strains fermenting at low pH is very attractive to the cellulosic bioethanol industry owing to challenges including bacterial contamination, reduction in cell viability, longer lag phase due to inhibitor detoxification, formation of undesirable products, lower ethanol yield, productivity and titre (Almeida et al. 2009b; Palmqvist and Hahn-Hägerdal 2000; Skinner and Leathers 2004). In this study, two different adaptation strategies were successfully applied to attain tolerance in S. cerevisiae towards more severe conditions than investigated before, i.e. low pH at 3.7 and inhibitory concentrations of acetic acid, furfural, HMF and vanillin. When a short-term adaptation strategy was applied with the IC, cells were able to grow at pH 5.0 and 4.5, but not at pH 4.0 and 3.7 (Fig. 2). This could be due to synergistic effects of low pH and the different inhibitors, i.e. (1) increased passive diffusion of undissociated acetic acid into the cells leading to acidification of the cytoplasm. Once inside the cell, acetate, being a weak acid, will reduce the intracellular pH. Plasma membrane ATPases and multiple drug resistance transporters pump the protons and acetate anions, respectively, out of the cell by utilising ATP, leading to an energy drain (Caspeta et al. 2015; Piotrowski et al. 2014; Taherzadeh and Karimi 2011), (2) deactivation of key cellular and glycolytic enzymes, membrane and DNA damage caused by furfural and HMF (Piotrowski et al. 2014); (3) vanillin may cause damage to cell membrane integrity (Piotrowski et al. 2014; Trinh Thi My et al. 2014); and (4) reduction of the furans demands excessive reducing power of NAD(P)H that is directed away from ethanol production and anabolism (Almeida et al. 2007, 2009b). All four mechanisms might have contributed to reduced growth capacity, elongated lag phases, lower growth rates and ethanol yields (Almeida et al. 2011; Piotrowski et al. 2014). As the mechanism of detoxification, tolerance, energy and co-factor requirements are different among weak acids, furans and phenolics, the presence of more than one class of inhibitor in the substrate results in an additional burden to the yeast cell. Short-term adaptation to the inhibitors at pH 5 might have led to a reduction in the cytosolic pH that subsequently has increased the tolerance to acetic acid at pH 4.5, as was earlier observed in another S. cerevisiae strain (Fernández-Niño et al. 2015). Hence during incubation at pH 3.7 with the IC, the effect of low pH and diffusion of undissociated acetic acid inside the cell must have been less pronounced. Increasing the cell density combined with short-term adaptation improved the potential of growth and ethanol fermentation capacity at pH 3.7 with the IC in 12 h compared to the non-adapted culture (Figs. 4, 5; Table 1). Higher inoculum concentrations could have several positive effects towards inhibitor detoxification including: (1) ‘Safety in numbers’, i.e. lowering the ratio of inhibitor concentration over cell concentration, which leads to lower detoxification demand per cell; and (2) population heterogeneity, where more representatives of several sub populations could be dedicatedly involved in detoxification of inhibitors, thereby facilitating growth and ethanol production by other sub populations. The heterogeneity could be due to (a) variations in cell growth phase, cycle and cell ageing, and (b) stochasticity in gene expression impacting enzymatic activities leading to variations in metabolic reactions (Avery 2006; Delvigne et al. 2014). Of all the inhibitors, vanillin had a major impact on the growth performance of strain TMB3500 at pH 3.7 in the presence of acetic acid as previously observed by Klinke et al. (2004) affecting the growth of S. cerevisiae at a concentration of 0.5 g L−1, followed by furfural with an increased lag phase, whereas HMF had no effect (Fig. 1) when compared with cells grown at pH 3.7 with acetic acid (Sànchez i Nogué et al. 2013). Since vanillin is involved in membrane damage, it might aid the cellular entry of acetic acid, thereby affecting the fitness and metabolism rapidly. Synergistic effects of acetic acid and furfural have been reported to negatively affect specific growth rate and ethanol yield (Palmqvist et al. 1999). In addition, effects of the individual and combination of inhibitors including furfural, phenol and acetic acid have been analysed using metabolic profiling. The synergistic negative effect on amino acids and central carbon metabolism was more pronounced than the sum of individual inhibitors with acetic acid playing a key role in the combined inhibition (Ding et al. 2011). Evolutionary engineering is a useful tool to obtain a desired phenotype by acquiring a stable genotype in an organism by applying constant or increasing selection pressure (Almario et al. 2013; Sauer 2001). The natural evolution process towards desirable properties like adaptability to low pH and inhibitor tolerance is prominent among industrial yeast strains where cells are repeatedly washed with dilute H2SO4 and recycled in the fermentation process along with pre-cultures as observed with the PE-2 strain used in Brazilian ethanol production plants (Della-Bianca et al. 2014). Exposure of strain TMB3500 to step-wise reduction in pH from 5 to 3.7 over 3600 h in a chemostat led to successful growth and ethanol production in the presence of inhibitory concentrations of acetic acid, furfural, HMF and vanillin over the whole pH range. Interestingly, at pH 4.1, biofilm was formed in the fermentor, possibly to protect the cells against harsher conditions as was observed for Zymomonas mobilis on rice bran hydrolysate forming a protective layer around cells (Todhanakasem et al. 2014). Yeast cells form biofilm through cell–cell adhesion in response to stress through triggering the Ras/cAMP/protein kinase A (PKA) and mitogen activated protein kinase pathways and expressing the FLO genes (Verstrepen and Klis 2006). Hence, the biofilm might have contributed to the majority of the detoxification and cell proliferation. In addition, cells released from the biofilm into the liquid suspension may have contributed to ethanol production, but without proliferation. Therefore, when the cell suspension was transferred into another chemostat under the same condition, the culture simply washed out immediately, adding credibility to this interpretation. Surprisingly, characterisation of the adapted strain acquired from the suspension after 709 generations showed that it possessed a similar phenotype as the parental strain TMB3500 in response to the exposure of low pH and the IC. The apparent robust phenotype had disappeared as soon as the adapted strain was exposed again to the harsh conditions as seen by Wright et al. (2011) when adapting for acetic acid tolerance. However, one drawback of the experimental setup of characterisation could have been the low inoculum size of OD 0.5 (OD of 2.5 corresponds to 1 gdw L−1 of strain TMB3500) used to start the aerobic and anaerobic liquid batches, which may very well have been below the critical cell mass to initially detoxify the inhibitors at low pH when compared with the size of inoculum in the chemostat (Fig. 6a). All in all, this study indicated that the phenotype of both the short-term and long-term adaptation, i.e. growth at pH 3.7 with inhibitors, turned out to be similar with no rigorous genetic changes. Copy number variations of specific genes and single nucleotide polymorphisms obtained from 5 months of evolution might be very specific to be visualised in the broad genomic DNA fingerprinting techniques used for the current analysis. Indeed, short-term adaptation at acidic pH might have a positive influence in epigenetic expression of various stress response genes and transcription factors/activators including YAP1, HAA1 (Anneli et al. 2006; Modig et al. 2008; Wright et al. 2011). Moreover, De Melo et al. (2010) used an industrial strain named JP1 to show that an acidic environment affected cell growth and induced general stress response. The cell tolerance to acidic environment may involve down regulation of transcription and protein synthesis due to PKA based glucose signalling (De Melo et al. 2010). Interestingly, a cross-tolerance phenomenon has been observed in S. cerevisiae, meaning that tolerance acquired to one stress enhances resistance to other forms of stress (Gibson et al. 2007). For instance, low pH stress due to inorganic and weak organic acids induces the expression of genes involved in tolerance to heat shock, cell wall assembly, trehalose biosynthesis, tolerance to osmotic stress and glycerol production (Kapteyn et al. 2001; Kawahata et al. 2006). Also, the resulting enhancement in population performance that we observed in ALE could be due to polygenic response as observed by Meijnen et al. (2016) in the case of tolerance towards acetic acid, preserving beneficial mutations and by avoiding any undesirable pleiotropic response as in case of any targeted genetic manipulation (Sauer 2001). The adaptation patterns observed in our study, including the long-term adaptation, could thus be due to a combination of a variety of genetic changes and stochastic switching that are either triggered by the harsh environment (Acar et al. 2008) or are already present in a subpopulation (Delvigne and Goffin 2014; Levy et al. 2012). In the latter case, phenotypic variability or plasticity exhibited in the subpopulation might enhance the survival of the species when confronted with diverse hostile environments (Delvigne and Goffin 2014; Levy et al. 2012). Most likely, in TMB3500 cultures only the subpopulation readily adapted to the inhibitors at low pH were selected for, which is underlined by the need to apply thick inocula to provide a critical mass of this subpopulation. We have performed a short-term adaptation and an ALE of an industrial S. cerevisiae strain to grow and produce ethanol at low pH in the presence of lignocellulosic inhibitors. The next step would be to perform fermentations with yeast and simulated contaminations of bacteria in a medium stressed with inorganic and weak organic acids in the presence of lignocellulosic inhibitors to analyze the effect of low pH fermentations. This might be a key step towards reduction of bacterial contamination in large-scale lignocellulosic ethanol production. Though the evolved strain obtained from long-term evolution did not maintain stable inhibitor robustness, the short-term adaptation strategy to pH 5 and all inhibitors made it possible for the first time to successfully ferment glucose to ethanol at pH 3.7 in the presence of lignocellulosic inhibitors. Inhibitor tolerance and cell growth in yeast is thus achieved through phenotypic plasticity, i.e. a delicate phenotypic balance underlined by expression of genes relating to stress tolerance, growth and fermentation. Understanding the complex genetic regulation behind phenotypic plasticity and stochastic expression in adaptive evolution might enable us to steer the yeast cells to desirable metabolic responses. Additional file 10.1186/s13568-016-0234-8 DNA fingerprinting with genomic DNA extracted from parental strain TMB3500 (first column of gel), evolved strain CC156 (middle column of gel) and biofilm (third column of gel). Primers used include a) (GACA)4 b) (GTG)5 c) S1254 random primer d) TY1 primer e) TY3 primer f) TY1 + TY3 primer. One gel is displayed from each strain as a representative from triplicate analysis. Abbreviations ALEadaptive lab evolution SSFsimultaneous saccharification and fermentation HMFhydroxymethylfurfural YPDyeast peptone dextrose KOHpotassium hydroxide H2SO4sulphuric acid ICinhibitor cocktail ODoptical density TMB3500Brewer’s yeast strain CC156evolved population TYtransposable elements μmaxmaximum specific growth rate PKAprotein kinase A Authors’ contributions All the authors participated in the design of the study, VN and VSN performed the experimental work, VN and EvN wrote the manuscript and all the authors critically read and approved the manuscript. All authors read and approved the final manuscript. Acknowledgements We would like to thank Dr. Magnus Carlquist for the discussions and critically reading the manuscript. Competing interests The authors declare that they have no competing interests. Funding This research was funded by Swedish National Energy Agency (www.energimyndigheten.se, Project No. P35350-1). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. ==== Refs References Acar M Mettetal JT van Oudenaarden A Stochastic switching as a survival strategy in fluctuating environments Nat Genet 2008 40 4 471 475 10.1038/ng.110 18362885 Akopyanz N Bukanov NO Westblom TU Kresovich S Berg DE DNA diversity among clinical isolates of Helicobacter pylori detected by PCR-based RAPD fingerprinting Nucl Acids Res 1992 20 19 5137 5142 10.1093/nar/20.19.5137 1408828 Albers E Johansson E Franzén CJ Larsson C Selective suppression of bacterial contaminants by process conditions during lignocellulose based yeast fermentations Biotechnol Biofuels 2011 4 59 10.1186/1754-6834-4-59 22185398 Almario MPP Reyes LH Kao KC Evolutionary engineering of Saccharomyces cerevisiae for enhanced tolerance to hydrolysates of lignocellulosic biomass Biotechnol Bioeng 2013 110 10 2616 2623 10.1002/bit.24938 23613173 Almeida JR Modig T Petersson A Hähn-Hägerdal B Lidén G Gorwa-Grauslund MF Increased tolerance and conversion of inhibitors in lignocellulosic hydrolysates by Saccharomyces cerevisiae J Chem Technol Biotechnol 2007 82 4 340 349 10.1002/jctb.1676 Almeida JRM Karhumaa K Bengtsson O Gorwa-Grauslund M-F Screening of Saccharomyces cerevisiae strains with respect to anaerobic growth in non-detoxified lignocellulose hydrolysate Bioresour Technol 2009 100 14 3674 3677 10.1016/j.biortech.2009.02.057 19329297 Almeida JRR Bertilsson M Gorwa-Grauslund MF Gorsich S Lidén G Metabolic effects of furaldehydes and impacts on biotechnological processes Appl Microbiol Biotechnol 2009 82 4 625 638 10.1007/s00253-009-1875-1 19184597 Almeida JRM Runquist D Sànchez Nogué V Lidén G Gorwa-Grauslund MF Stress-related challenges in pentose fermentation to ethanol by the yeast Saccharomyces cerevisiae Biotechnol J 2011 6 3 286 299 10.1002/biot.201000301 21305697 Alriksson B Horváth IS Jönsson LJ Overexpression of Saccharomyces cerevisiae transcription factor and multidrug resistance genes conveys enhanced resistance to lignocellulose-derived fermentation inhibitors Process Biochem 2010 45 2 264 271 10.1016/j.procbio.2009.09.016 Andrade M Rodriguez M Sánchez B Aranda E Córdoba J DNA typing methods for differentiation of yeasts related to dry-cured meat products Int J Food Microbiol 2006 107 1 48 58 10.1016/j.ijfoodmicro.2005.08.011 16257467 Anneli P João RMA Tobias M Kaisa K Bärbel HH Marie FGG Gunnar L A 5-hydroxymethyl furfural reducing enzyme encoded by the Saccharomyces cerevisiae ADH6 gene conveys HMF tolerance Yeast 2006 23 6 455 464 10.1002/yea.1370 16652391 Avery SV Microbial cell individuality and the underlying sources of heterogeneity Nat Rev Micro 2006 4 8 577 587 10.1038/nrmicro1460 Basso LC, Basso TO, Rocha SN. Ethanol production in Brazil: the industrial process and its impact on yeast fermentation. In: dos Santos Bernardes MA, editor. Biofuel production-recent developments and prospects. Rijeka: InTech; 2011. p. 85–100. doi:10.5772/17047. Beales N Adaptation of microorganisms to cold temperatures, weak acid preservatives, low ph, and osmotic stress: a review Compr Rev Food Sci F 2004 3 1 1 20 10.1111/j.1541-4337.2004.tb00057.x Beckner M Ivey M Phister T Microbial contamination of fuel ethanol fermentations Lett Appl Microbiol 2011 53 4 387 394 10.1111/j.1472-765X.2011.03124.x 21770989 Bischoff KM Liu S Leathers TD Worthington RE Rich JO Modeling bacterial contamination of fuel ethanol fermentation Biotechnol Bioeng 2009 103 1 117 122 10.1002/bit.22244 19148876 Caspeta L Castillo T Nielsen J Modifying yeast tolerance to inhibitory conditions of ethanol production processes Front Bioeng Biotechnol 2015 3 184 10.3389/fbioe.2015.00184 26618154 da Silva-Filho EAA Brito dos Santos SK Resende AdM de Morais JOO de Morais MA Ardaillon Simões D Yeast population dynamics of industrial fuel-ethanol fermentation process assessed by PCR-fingerprinting Antonie Van Leeuwenhoek 2005 88 1 13 23 10.1007/s10482-004-7283-8 15928973 De Melo HF Bonini BM Thevelein J Simões DA Morais MA Physiological and molecular analysis of the stress response of Saccharomyces cerevisiae imposed by strong inorganic acid with implication to industrial fermentations J Appl Microbiol 2010 109 1 116 127 20002866 Della-Bianca BE de Hulster E Pronk JT van Maris AJA Gombert AK Physiology of the fuel ethanol strain Saccharomyces cerevisiae PE-2 at low pH indicates a context-dependent performance relevant for industrial applications FEMS Yeast Res 2014 14 8 1196 1205 10.1111/1567-1364.12217 25263709 Delvigne F Goffin P Microbial heterogeneity affects bioprocess robustness: dynamic single-cell analysis contributes to understanding of microbial populations Biotechnol J 2014 9 1 61 72 10.1002/biot.201300119 24408611 Delvigne F Zune Q Lara AR Al-Soud W Sørensen SJ Metabolic variability in bioprocessing: implications of microbial phenotypic heterogeneity Trends Biotechnol 2014 32 12 608 616 10.1016/j.tibtech.2014.10.002 25457387 Ding MZ Wang X Yang Y Yuan YJ Metabolomic study of interactive effects of phenol, furfural, and acetic acid on Saccharomyces cerevisiae OMICS 2011 15 10 647 653 10.1089/omi.2011.0003 21978393 Dominik H Uwe S Identification of furfural as a key toxin in lignocellulosic hydrolysates and evolution of a tolerant yeast strain Microb Biotechnol 2008 1 6 497 506 10.1111/j.1751-7915.2008.00050.x 21261870 Fernández-Niño M Marquina M Swinnen S Rodríguez-Porrata B Nevoigt E Ariño J The cytosolic pH of individual Saccharomyces cerevisiae cells is a key factor in acetic acid tolerance Appl Environ Microbiol 2015 81 22 7813 7821 10.1128/AEM.02313-15 26341199 Gibson BR Lawrence SJ Leclaire JPR Powell CD Smart KA Yeast responses to stresses associated with industrial brewery handling FEMS Microbiol Rev 2007 31 5 535 569 10.1111/j.1574-6976.2007.00076.x 17645521 Graves T Narendranath NV Dawson K Power R Effect of pH and lactic or acetic acid on ethanol productivity by Saccharomyces cerevisiae in corn mash J Ind Microbiol Biotechnol 2006 33 6 469 474 10.1007/s10295-006-0091-6 16491359 Hahn-Hägerdal B Karhumaa K Larsson CU Gorwa-Grauslund M Görgens J van Zyl WH Role of cultivation media in the development of yeast strains for large scale industrial use Microb Cell Fact 2005 4 31 10.1186/1475-2859-4-31 16283927 Hanqi G Jian Z Jie B Inhibitor analysis and adaptive evolution of Saccharomyces cerevisiae for simultaneous saccharification and ethanol fermentation from industrial waste corncob residues Bioresour Technol 2014 157 6 13 10.1016/j.biortech.2014.01.060 24518544 Harju S Fedosyuk H Peterson KR Rapid isolation of yeast genomic DNA: bust n’Grab BMC Biotechnol 2004 4 1 8 10.1186/1472-6750-4-8 15102338 Henningsen BM Hon S Covalla SF Sonu C Argyros AD Barrett TF Wiswall E Froehlich AC Zelle RM Increasing anaerobic acetate consumption and ethanol yields in Saccharomyces cerevisiae with NADPH-specific alcohol dehydrogenase Appl Environ Microbiol 2015 81 23 8108 8117 10.1128/AEM.01689-15 26386051 i Nogué V Bettiga M Gorwa-Grauslund MF Isolation and characterization of a resident tolerant Saccharomyces cerevisiae strain from a spent sulfite liquor fermentation plant AMB Express 2012 2 1 68 10.1186/2191-0855-2-68 23237549 Ishola MM Babapour AB Gavitar MN Brandberg T Taherzadeh MJ Effect of high solids loading on bacterial contamination in lignocellulosic ethanol production Bioresources 2013 8 3 4429 4439 10.15376/biores.8.3.4429-4439 Jönsson LJ Alriksson B Nilvebrant N-O Bioconversion of lignocellulose: inhibitors and detoxification Biotechnol Biofuels 2013 6 1 16 10.1186/1754-6834-6-16 23356676 Kádár Z Maltha SF Szengyel Z Réczey K De Laat W Ethanol fermentation of various pretreated and hydrolyzed substrates at low initial pH Appl Biochem Biotechnol 2007 137 1–12 847 858 18478439 Kapteyn J Ter Riet B Vink E Blad S De Nobel H Van Den Ende H Klis F Low external pH induces HOG1-dependent changes in the organization of the Saccharomyces cerevisiae cell wall Mol Microbiol 2001 39 2 469 480 10.1046/j.1365-2958.2001.02242.x 11136466 Kawahata M Masaki K Fujii T Iefuji H Yeast genes involved in response to lactic acid and acetic acid: acidic conditions caused by the organic acids in Saccharomyces cerevisiae cultures induce expression of intracellular metal metabolism genes regulated by Aft1p FEMS Yeast Res 2006 6 6 924 936 10.1111/j.1567-1364.2006.00089.x 16911514 Klinke HB Thomsen AB Ahring BK Inhibition of ethanol-producing yeast and bacteria by degradation products produced during pre-treatment of biomass Appl Microbiol Biotechnol 2004 66 1 10 26 10.1007/s00253-004-1642-2 15300416 Koichi T Yukari I Jun O Jun S Enhancement of acetic acid tolerance in Saccharomyces cerevisiae by overexpression of the HAA1 gene, encoding a transcriptional activator Appl Environ Microbiol 2012 78 22 8161 8163 10.1128/AEM.02356-12 22961896 Lei J Yu S Lili X Bingyin P Yazhong X Xiaoming B Enhanced resistance of Saccharomyces cerevisiae to vanillin by expression of lacA from Trametes sp. AH28 -2 Bioresour Technol 2011 102 17 8105 8109 10.1016/j.biortech.2011.06.057 21727001 Levy SF Ziv N Siegal ML Bet hedging in yeast by heterogeneous, age-correlated expression of a stress protectant PLoS Biol 2012 10 5 e1001325 10.1371/journal.pbio.1001325 22589700 Limayem A Ricke SC Lignocellulosic biomass for bioethanol production: current perspectives, potential issues and future prospects Prog Energy Combust Sci 2012 38 4 449 467 10.1016/j.pecs.2012.03.002 Liu ZL Molecular mechanisms of yeast tolerance and in situ detoxification of lignocellulose hydrolysates Appl Microbiol Biotechnol 2011 90 3 809 825 10.1007/s00253-011-3167-9 21380517 Meijnen J-P Randazzo P Foulquié-Moreno MR van den Brink J Vandecruys P Stojiljkovic M Dumortier F Zalar P Boekhout T Gunde-Cimerman N Kokošar J Štajdohar M Curk T Petrovič U Thevelein JM Polygenic analysis and targeted improvement of the complex trait of high acetic acid tolerance in the yeast Saccharomyces cerevisiae Biotechnol Biofuels 2016 9 5 10.1186/s13068-015-0421-x 26740819 Modig T Almeida JR Gorwa-Grauslund MF Lidén G Variability of the response of Saccharomyces cerevisiae strains to lignocellulose hydrolysate Biotechnol Bioeng 2008 100 3 423 429 10.1002/bit.21789 18438882 Muthaiyan A Limayem A Ricke SC Antimicrobial strategies for limiting bacterial contaminants in fuel bioethanol fermentations Prog Energy Combust Sci 2011 37 3 351 370 10.1016/j.pecs.2010.06.005 Nielsen F Tomás-Pejó E Olsson L Wallberg O Short-term adaptation during propagation improves the performance of xylose-fermenting Saccharomyces cerevisiae in simultaneous saccharification and co-fermentation Biotechnol Biofuels 2015 8 219 10.1186/s13068-015-0399-4 26697108 Palmqvist E Grage H Meinander NQ Hahn-Hagerdal B Main and interaction effects of acetic acid, furfural, and p-hydroxybenzoic acid on growth and ethanol productivity of yeasts Biotechnol Bioeng 1999 63 1 46 55 10.1002/(SICI)1097-0290(19990405)63:1<46::AID-BIT5>3.0.CO;2-J 10099580 Palmqvist E Hahn-Hägerdal B Fermentation of lignocellulosic hydrolysates. II: inhibitors and mechanisms of inhibition Bioresour Technol 2000 74 1 25 33 10.1016/S0960-8524(99)00161-3 Papoutsakis ET Pronk JT Editorial overview: energy biotechnology Curr Opin Biotechnol 2015 33 8 11 10.1016/j.copbio.2015.04.001 25305544 Parawira W Tekere M Biotechnological strategies to overcome inhibitors in lignocellulose hydrolysates for ethanol production: review Crit Rev Biotechnol 2011 31 1 20 31 10.3109/07388551003757816 20513164 Piotrowski JS Zhang Y Sato T Ong I Keating D Bates D Landick R Death by a thousand cuts: the challenges and diverse landscape of lignocellulosic hydrolysate inhibitors Front Microbiol 2014 5 90 10.3389/fmicb.2014.00090 24672514 Sànchez i Nogué V Narayanan V Gorwa-Grauslund M Short-term adaptation improves the fermentation performance of Saccharomyces cerevisiae in the presence of acetic acid at low pH Appl Microbiol Biotechnol 2013 97 16 7517 7525 10.1007/s00253-013-5093-5 23872959 Sauer U Evolutionary engineering of industrially important microbial phenotypes. Advances in biochemical engineering/biotechnology 2001 Berlin Springer 129 169 Schofield MA Rowe SM Hammond JR Molzahn SW Quain DE Differentiation of brewery yeast strains by DNA fingerprinting J Inst Brewing 1995 101 2 75 78 10.1002/j.2050-0416.1995.tb00850.x Serate J Xie D Pohlmann E Donald C Shabani M Hinchman L Higbee A Mcgee M La Reau A Klinger GE Li S Myers CL Boone C Bates DM Cavalier D Eilert D Oates LG Sanford G Sato TK Dale B Landick R Piotrowski J Ong RG Zhang Y Controlling microbial contamination during hydrolysis of AFEX-pretreated corn stover and switchgrass: effects on hydrolysate composition, microbial response and fermentation Biotechnol Biofuels 2015 8 1 1 17 10.1186/s13068-015-0356-2 25642283 Skinner KA Leathers TD Bacterial contaminants of fuel ethanol production J Ind Microbiol Biotechnol 2004 31 9 401 408 10.1007/s10295-004-0159-0 15338420 Taherzadeh MJ Karimi K Fermentation inhibitors in ethanol processes and different strategies to reduce their effects. Biofuels, alternative feedstocks and conversion processes 2011 Cambridge Academic Press 287 311 Takuya I Daisuke W Yoko Y Koichi T Jun O Hiroshi T Hitoshi S Jun S An organic acid-tolerant HAA1 -overexpression mutant of an industrial bioethanol strain of Saccharomyces cerevisiae and its application to the production of bioethanol from sugarcane molasses AMB Express 2013 3 1 74 10.1186/2191-0855-3-74 24373204 Todhanakasem T Sangsutthiseree A Areerat K Young GM Thanonkeo P Biofilm production by Zymomonas mobilis enhances ethanol production and tolerance to toxic inhibitors from rice bran hydrolysate New Biotechnol 2014 31 5 451 459 10.1016/j.nbt.2014.06.002 Tomás-Pejó E Olsson L Influence of the propagation strategy for obtaining robust Saccharomyces cerevisiae cells that efficiently co-ferment xylose and glucose in lignocellulosic hydrolysates Microb Biotechnol 2015 8 6 999 1005 10.1111/1751-7915.12280 25989314 Trinh Thi My N Sakihito K Shingo I Importance of glucose-6-phosphate dehydrogenase (G6PDH) for vanillin tolerance in Saccharomyces cerevisiae J Biosci Bioeng 2014 118 3 263 269 10.1016/j.jbiosc.2014.02.025 24725964 Verduyn C Postma E Scheffers WA Van Dijken JP Effect of benzoic acid on metabolic fluxes in yeasts: a continuous-culture study on the regulation of respiration and alcoholic fermentation Yeast 1992 8 7 501 517 10.1002/yea.320080703 1523884 Verstrepen KJ Klis FM Flocculation, adhesion and biofilm formation in yeasts Mol Microbiol 2006 60 1 5 15 10.1111/j.1365-2958.2006.05072.x 16556216 Worley-Morse TO Deshusses MA Gunsch CK Reduction of invasive bacteria in ethanol fermentations using bacteriophages Biotechnol Bioeng 2015 112 8 1544 1553 10.1002/bit.25586 25788328 Wright J Bellissimi E de Hulster E Wagner A Pronk JT van Maris AJ Batch and continuous culture-based selection strategies for acetic acid tolerance in xylose-fermenting Saccharomyces cerevisiae FEMS Yeast Res 2011 11 3 299 306 10.1111/j.1567-1364.2011.00719.x 21251209

PMC005xxxxxx/PMC5001961.txt

==== Front J Headache PainJ Headache PainThe Journal of Headache and Pain1129-23691129-2377Springer Milan Milan 66810.1186/s10194-016-0668-zResearch ArticleATP-sensitive muscle afferents activate spinal trigeminal neurons with meningeal afferent input in rat – pathophysiological implications for tension-type headache Nöbel Moritz noebel-willing@gmx.de 1Feistel Stephan stephan-feistel@web.de 1Ellrich Jens jens.ellrich@fau.de 12Messlinger Karl karl.messlinger@fau.de 11 Institute of Physiology and Pathophysiology, Friedrich-Alexander-University of Erlangen-Nürnberg, Universitätsstr. 17, 91054 Erlangen, Germany 2 Department of Health Science and Technology, Medical Faculty, Aalborg University, Aalborg, Denmark 26 8 2016 26 8 2016 2016 17 1 753 5 2016 17 8 2016 © The Author(s). 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Background Tension-type headache and other primary headaches may be triggered or aggravated by disorders of pericranial muscles, which is possibly due to convergent or collateral afferent input from meningeal and muscular receptive areas. In rodent models high extracellular concentrations of ATP caused muscle nociception and central sensitization of second order neurons. In a rat model of meningeal nociception we asked if spinal trigeminal activity induced by ATP can be modulated by local anaesthesia of distinct muscles. Methods Ongoing activity was recorded from spinal trigeminal neurons with afferent input from the cranial dura mater, the temporal muscle and neck muscles. The stable ATP analogue α,β-methylene adenosine 5′-triphosphate (α,β-meATP, 10 mM) was injected into the ipsilateral temporal muscle, 30 min later followed by injection of local anaesthetics (lidocaine, 2 %) into the ipsilateral neck muscles and/or the temporal muscle. Results Injection of α,β-meATP into the temporal muscle caused progressive increase in ongoing activity of most of the spinal trigeminal neurons within 30 min. Injection of lidocaine into the neck muscles and/or the temporal muscle reduced this activation to previous levels within 10 min. Conclusions Distinct spinal trigeminal neurons processing meningeal nociceptive information are under the control of convergent afferent input from several pericranial muscles. Blockade of at least one of these inputs can normalize central trigeminal activity. This may explain why therapeutic manipulations of head muscles can be beneficial in primary headaches. Keywords Tension-type headachePericranial musclesMeningeal nociceptionα,β-meATPLocal anaesthesiahttp://dx.doi.org/10.13039/501100004963Seventh Framework Programme Grant 602633 (EUROHEADPAIN)Messlinger Karl issue-copyright-statement© The Author(s) 2016 ==== Body Background Tension-type headache (TTH) and migraine are the most frequent types of primary headache [1, 2]. Psychophysical studies in patients have indicated an important role of neck muscle nociception inducing central sensitization in the pathophysiology of TTH [3, 4], which has been confirmed in animal experiments [5, 6]. Pericranial tenderness and hardness of head and neck muscles are significantly pronounced in TTH and migraine patients compared to healthy volunteers [7–9]. Moreover, pericranial tenderness is positively associated with both the intensity and frequency of TTH [8] and more pronounced in patients with co-occurrence of migraine [10]. Accordingly, pressure pain and tolerance thresholds of head and neck regions are significantly decreased in TTH patients compared with healthy controls [11]. During cervical examination of patients suffering from migraine or TTH patients reported referred pain similar to the headache they usually experienced [12]. Adenosine 5′-triphosphate (ATP) is essential for an effective muscle contraction [13]. Natural concentrations of ATP in muscle cytosol are in the millimolar range buffered by creatine phosphate [14]. Interstitial ATP levels rise measurably dependent on experimental muscle tension [15] and compression [16] but also during natural exercise in humans associated with intensity of muscle contraction [17]. Upon cell damage and inflammation [18] high extracellular levels of ATP may induce purinergic signaling [19] and contribute to muscle pain [6, 20, 21]. In the rat ATP at concentrations present in muscle cells activated the majority of unmyelinated nociceptive and non-nociceptive afferent fibres [22] but high concentrations of ATP may decrease mechanical sensitivity of thin muscle afferents [23]. The stable ATP analogue α,β-methylene adenosine 5′-triphosphate (α,β-meATP) is most frequently used in animal studies on vascular and neuronal functions [24, 25]. Intra-arterial application of α,β-meATP in cats stimulated preferably C and slowly conducting Aβ fibres of triceps surae muscle [26] and caused inward currents in putative nociceptive dorsal root ganglion neurons [27] and sustained facilitation of the jaw opening reflex in mice as a measure of nociceptive processing in the trigeminal brainstem [28]. The present study was made to explore if electrophysiological recordings from spinal trigeminal neurons may reveal information about the processes underlying trigeminal reflex facilitation and headache associated with activation of pericranial muscle nociceptors. We used an approved model of meningeal nociception to record extracellularly from spinal trigeminal neurons with afferent input from the parietal dura mater, which is regarded to be involved in headache generation. The aim of the study was to find out whether and to which extent α,β-meATP injection into the temporal muscle modulates the neuronal activity of these neurons. The convergence of afferent information to second order neurons in the cervical dorsal horn from the supratentorial dura mater and the greater occipital nerve innervating major parts of neck muscles and skin has earlier been demonstrated [5]. Similar convergent mechanisms should be expected in the trigeminal nucleus caudalis, which is part of the trigeminal brainstem nuclear complex receiving both meningeal and cervical spinal afferent input [29, 30]. However, in the light of recent tracing experiments [31] collateral innervation of meningeal and pericranial structures by same primary trigeminal ganglion neurons may also contribute to the convergence of afferent information from different compartments to spinal trigeminal neurons. Methods The study was performed in accordance with the ethical issues of the International Association for the Study of Pain and with the guidelines and regulations of animal care provided by the Council Directive 2010/63EU of the European Parliament. The protocol was reviewed by an ethics committee and authorized by the local district government. Sixteen adult male Wistar rats with body weights ranging from 250 to 410 g were used. Rats were bred and held in groups of 2–4 in the Institute’s animal house under a 12/12 h light–dark cycle with food and water ad libitum. They were initially anaesthetised using 4 % isoflurane (Forene, Abbott, Wiesbaden, Germany) administered with an evaporator system (Vapor 19.3, Dräger, Lübeck, Germany) in a closed box. After the initial anaesthesia, 2 % isoflurane administration was maintained through a mask on the animal’s snout. The right femoral artery and vein were catheterized for systemic blood pressure recording and substance administration. Saline supplemented with 1 IU/ml heparin (Heparin-Natrium-5000-ratiopharm, Ulm, Germany) was permanently infused through the arterial catheter (rate 0.2 ml/h) to prevent blood clotting. Subsequently a tracheal tube (intravenous catheter, Vasuflo-T, Dispomed, Gelnhausen, Germany) was inserted in order to ventilate the animals (Rodent Ventilator, Ugo Basile, Comerio VA, Italy) with 2 % isoflurane in oxygen-enriched room air. The expiratory CO2 was monitored (Artema MM 200, Karl Heyer, Bad Ems, Germany) and kept steady at 3 % by modulating the ventilation frequency between 70 and 100 strokes per minute to suppress spontaneous breathing. Using a feedback controlled heating pad (TKM 0902, Föhr Medical Instruments, Frankfurt, Germany) and a rectal probe, the animal’s body temperature was maintained at 37–37.5 °C. Mean arterial pressure ranged from 80 to 100 mmHg (Pressure Monitor BP-1, World Precision Instruments, Sarasota, Florida) and was constant within each experiment. Accumulation of viscous mucus in the tracheal tube was prevented by intraperitoneal injection of 0.05 mg/kg atropine sulfate (Braun, Melsungen, Germany). The eyes of the animals were protected by an ointment (Bepanthen, Bayer, Leverkusen, Germany). During recordings the animals were paralyzed with i.v. administration of 50 mg/kg gallamine triethiodide (Sigma, Steinheim, Germany). The narcosis was maintained at constant depth during the experiment so that noxious pinch stimuli did not evoke nociceptive reflexes or changes in arterial blood pressure that may signal pain. After the experiments the animals were euthanized by administering an overdose of thiopental (Trapanal, Nycomed, Konstanz, Germany) intravenously. Specific surgery The animals were placed in a stereotaxic frame and the head fixed using ear bars and a snout clamp. The skull was exposed by cutting the skin along the midline from the forehead to the neck and extending the skin flaps, so that the dorsal part of the temporal muscle was visible. Using a dental drill and saline cooling the right parietal skull was carefully trepanned, thus avoiding dural blood vessel bleeding. After the dura mater was exposed by a cranial window of about 6 × 4 mm, it was protected from drying by cotton pads soaked with 0.9 % saline. In order to gain access to the medullary brainstem, the neck muscles were separated in the midline, detached from their insertions and held apart with a clamp. Finally the atlanto-occipital ligament was cut, exposing the spinal dura underneath. Recordings In order to record neuronal activity from the spinal trigeminal nucleus caudalis, custom-made carbon fibre glass microelectrodes (impendance 0.1–5 MΩ) were inserted into the ipsilateral medulla within the region of the spinal trigeminal nucleus caudalis. A microstepper was used to advance the electrodes through the brainstem at steps of 2.5 μm, while single units with meningeal receptive fields were detected by their firing of action-potentials to mechanical probing of the exposed dura mater with von Frey filaments. Receptive fields in the temporal muscle and neck muscles, facial skin, whiskers, and cornea were located using a fine glass rod. Mechanical stimuli were occasionally applied to the receptive fields to assure that the same neuron was recorded throughout the experiment. The responses to these stimuli were spared in the analysis. Recorded signals were band-pass filtered (0.5–1 kHz), amplified and processed (CED 1401, Cambridge Electronic Design, Cambridge, UK). Spike-analysis was performed offline using the discharges generated by mechanical stimulation of meningeal receptive fields as templates (Spike 2 software application, Cambridge Electronic Design). At the end of the experiment the x-y-position of the recording site relative to the caudal extension of the obex was noted according to the readings of the microdrives. Experimental protocols When the ongoing (unstimulated) neuronal activity of a unit was visibly stable, it was recorded for a control period of 30 min without any treatment (baseline activity). Subsequently, saline (0.9 % NaCl, 50 μl) was injected with a U-100 insulin syringe (BD, Franklin Lakes, NJ, USA) deep into the temporal muscle close to the crista temporalis within one minute. After another 30 min of recording, 50 μl of 10 mM α,β-meATP (α,β-methylene adenosine 5′-triphosphate disodium salt hydrate, Sigma-Aldrich, Taufkirchen, Germany) was injected into the temporal muscle and the recording continued for 60 min. After another 10 min without further treatment (new baseline), in most of the experiments 60 μl of 2 % lidocaine (Xylocain, Astra Zeneca, Wedel, Germany) was injected into the ipsilateral medial (semispinalis and rectus capitis) neck muscles followed by a recording period of 20 min. Next the temporal muscle was injected with 60 μl lidocaine and the activity recorded for another 20 min. Finally, 60 μl of lidocaine was applied onto the exposed cranial dura mater followed by a final recording period of 20 min. In some experiments 60 μl of lidocaine was injected only into the ipsilateral temporal muscle immediately followed by 50 μl of α,β-meATP, then the recording was continued for 60 min. Data analysis Discharges per minute were counted throughout the experiments. For comparison of data the neuronal activity was averaged across 10 min intervals, the injection minutes were excluded from analysis. Statistical analysis was made with Statistica software 7.1 (Statsoft GmbH, Hamburg, Germany). Data were analysed using analysis of variance (ANOVA) with repeated measurements followed by Fisher’s least significant difference (LSD) post hoc test comparing the averaged activity within intervals of 10 and 30 min. Differences were considered significant at p ≤ 0.05. Results General properties of neurons The recording sites of 16 units were located 1.64–3.4 mm caudal to the obex and 0.35–1.64 mm lateral to the midline and at depths of 302–972 μm below the dorsal surface of the medulla. The ongoing activity of units varied between 0.1 and 1304 (mean 183, SEM ± 83) spikes per minute during the control period. The meningeal receptive fields, mostly located close to the middle meningeal artery, were mapped. The mechanical threshold determined with graded von Frey filaments ranged from 0.49 to 11.8 mN (mean 5.0 mN). All recorded units received also convergent input from facial areas in the ophthalmic, maxillary or mandibular division of the trigeminal nerve as well as from the temporal muscle and/or the neck muscles and the periosteum around the cranial window and were characterized as wide-dynamic range (WDR) units according to their responses to touch stimuli of the facial skin. Mechanical threshold of muscle input ranged from 8.8 to 14.7 mN (mean temporal muscle 12.7 mN, neck muscles 13.6 mN). The units were activated by electrical pulses (duration 1 ms) with thresholds ranging from 0.4 to 2.1 mA applied to the meningeal receptive field. The latencies after a single electrical pulse close above threshold ranged from 12 to 25 ms. Taking a distance of 25 mm from the dura mater to the caudal medulla the units received afferent input from meningeal slowly conducting Aδ and/or C-fibres. Effects of vehicle and α,β-meATP After a control period of 30 min (baseline) vehicle was slowly injected into the temporal muscle, which was not followed by a significant increase in activity in the whole sample of 16 units within the recording time of 30 min, though the activity increased visibly in five (see example Fig. 1a) and decreased in 3 units. After the injection of α,β-meATP, the activity in half of the sample increased within 10 min and stayed at a higher level (Fig. 1a), in the other half it increased more or less gradually within the following 30 min. Four units of the sample showed virtually no activity (< 2 spikes/min) at baseline; two of them were only transiently activated during injection of saline and/or α,β-meATP ending up with no activity at 50 min after the injection. Repeated measures ANOVA applied to all 16 units during the 30 min periods before and after the injection of substances indicated a significant difference (F3,45 = 3.8; p < 0.02). Posthoc analysis revealed that the activity within the first 30 min after α,β-meATP was significantly different to the baseline (p = 0.01), while the activity within the period of 30–60 min after α,β-meATP was different to the baseline (p < 0.01) and to the activity after the vehicle injection (p < 0.05). Analysis of 10 min intervals of the sample of 12 units with spontaneous activity revealed the same differences (Fig. 2a).Fig. 1 Typical experiment showing the ongoing activity of a spinal trigeminal neuron displayed in discharges per minute. Injection of saline and α, β-meATP into the ipsilateral temporal muscle increased the activity (a), which was reversed by injection of lidocaine into the neck muscles (b), while lidocaine injection into the temporal muscle had no additional effect in this case Fig. 2 a Activity of spinal trigeminal neurons with afferent input from the dura mater and the temporal muscle. The normalized activity displayed in 10 min intervals increased more and more after injection of vehicle (saline) and α,β-meATP (ATP) into the ipsilateral temporal muscle (* significant difference to baseline, # to baseline and intervals after vehicle). The inset shows four additional units (activity displayed in 30 min intervals) which did not fit to the normalized sample because of their low spontaneous activity and relatively high activation following α,β-meATP injection. b Activity of spinal trigeminal neurons pre-treated by injection of α,β-meATP into the ipsilateral temporal muscle (experiments continued from (a)). Left: The activity (normalized to the 10 min interval following the experiment in (a)) is significantly (*) reduced after injection of lidocaine into the occipital muscles and further after injection into the temporal muscle (left) but not more after application of lidocaine onto the dura mater. Right: Four additional units recorded during lidocaine injection only into the temporal muscle, three of them showing decreased activity (displayed in 20 min intervals) Effects of lidocaine In 12 units the experiment was continued with local anaesthesia. After 60 min of recording following α,β-meATP injection, the mean activity within 10 min (new baseline) ranged from 20 to 1357 (mean 425, SEM ± 129) spikes/min. To test if the spinal trigeminal activity induced by local α,β-meATP could be blocked and reversed by anaesthesia of the homologous muscle, lidocaine was injected into the temporal muscle immediately followed by injection of α,β-meATP in four of the experiments (Fig. 2b right). In two units the neuronal activity decreased dramatically after the injections and did not recover during the recording time of further 60 min. In one unit the activity was slightly decreased, in another unit increased. In the remaining 8 experiments the effect of local anaesthesia of the heterologous structures delivering convergent input to the activated spinal trigeminal neurons was tested. Lidocaine was injected first into the neck muscles, 20 min later into the temporal muscle and after another 20 min finally applied onto the dura mater (Figs. 1b, 2b left). In all units but two the activity decreased already after the first lidocaine injection and continued to decrease in the whole sample of units (repeated measures ANOVA, F3,21 = 5.1; p < 0.01). Posthoc analysis revealed significant differences to the baseline after each lidocaine application (p < 0.01) (Fig. 2b left). Discussion The present experiments show that spinal trigeminal neurons with afferent input from the temporal muscle and the cranial dura mater increase their ongoing activity after injecting α,β-meATP deeply into the temporal muscle. This activation developed gradually in half of the neurons. The injection itself may have contributed to the increased activity in some neurons, likely caused by mechanical irritation, but did not play a significant role in the whole sample of units. The activating effect of α,β-meATP was persistent and even increased further beyond the 60 min after injection into the temporal muscle, before local anaesthesia was applied. Lidocaine injection into this muscle had a dramatical effect in lowering the activity of spinal trigeminal neurons suggesting that this activity was mainly maintained by afferent input from the temporal muscle. However, this could have been expected, while the main question was whether or not the ongoing activity was also dependent on the afferent input from the other structures, in which receptive fields were located. Therefore we anaesthetized first the neck muscles and later the dura mater in the main sample of units. A rapid decrease in α,β-meATP-driven activity was seen after anaesthesia of the neck muscles, indicating significant convergent input in accordance with the location of receptive fields. This clear effect could be explained by assuming that the afferents supplying the neck muscles had been sensitized during surgery (exposure and separation of muscles, dissection of the atlanto-occipital ligament) causing ongoing afferent activity that contributed to maintain the activity of second order neurons. After additional anaesthesia of the temporal muscle the activity further declined but local anaesthesia of the dura mater was not further effective, because the activity had already returned to the basal level. It seems that removal of a critical portion of afferent input to these multireceptive units is sufficient to restore their basal activity. This is reminiscent of previous recordings in which the ongoing central activity was blocked by anaesthesia of the trigeminal ganglion [32]. The activity evoked by α,β-meATP was probably due to purinergic signaling involving P2X receptor channels. Functional experiments indicate the preference of P2X receptors in skeletal muscle nociception. Activation of P2X3 receptors caused hyperalgesia of rat masseter muscle, which involved phosphorylation of TRPV1 receptors [33]. A variety of experiments investigating craniofacial nociceptive reflexes revealed various purinergic mechanisms in neck muscle nociception including P2X receptor activation that drive the nociceptive input to the trigeminal brainstem [6, 34]. The relevance of this translational model of tension-type headache is underlined by the suppression of α,β-meATP-induced facilitation of neck muscle nociception through inhibition of nitric oxide generation and acetylsalicylic acid [35, 36]. Activation of ATP-sensitive P2X receptors evoked also skeletal muscle afferent-mediated pressor responses [37]. Moreover, activation of P2X receptors in orofacial tissues have been found to play a critical role in central sensitization of second order neurons in the spinal trigeminal nucleus [38]. The fact that the elevated spinal trigeminal activity in our experiments could be normalized with the abrogation of peripheral input may suggest that central sensitization was not yet fully developed. Spinal trigeminal neurons with meningeal afferent input have frequently been recorded to study mechanisms of meningeal nociception [39, 40], though many of these neurons receive convergent cutaneous afferent input and show a wide-dynamic range (WDR) character, i.e. respond to cutaneous stimuli of increasing strength in a graded fashion [41, 42]. The units recorded in this study can be regarded to signal muscle pain, since they are activated by pressure onto the temporal muscle and respond to the noxious chemical stimulus α,β-meATP as a surrogate of high extracellular concentrations of ATP. In addition, due to their meningeal receptive input, they may signal meningeal nociceptive events, and respective neurons in humans may be those signaling head pain. Earlier, Bartsch and Goadsby [5] recorded from second-order neurons in the rat dorsal horn of the C2 cervical segment with afferent input from the supratentorial dura mater and the greater occipital nerve, which innervates major parts of the neck muscles and skin. Electrical stimulation of the greater occipital nerve as well as application of the TRPA1/V1 agonist mustard oil to the dura or the muscle enhanced the response of most of the spinal neurons to electrical stimuli applied to the dura or the nerve, revealing convergent input from both structures to the spinal trigeminal second order neurons. Neurons were also sensitized by this treatment showing decreased mechanical thresholds of the dural receptive sites and enlarged facial and cervical cutaneous mechanoreceptive fields as well as increased responses to noxious mechanical stimulation of paraspinal muscles [43]. These results, as well as the data of our study, may reveal a pathophysiological basis for pain referred from meningeal to cervical innervations territories, which may also underlie the clinical phenomena of cervical hypersensitivity and neck muscle tension in primary headaches like migraine [44]. Conversely, the convergent nociceptive input from neck muscles and meninges may form the basis for tension-type headache which can be induced or aggravated by muscle tension. Extracellular accumulation of ATP, which is measurably increased during experimental muscle tension and compression [15, 16] and intense natural muscle contraction [17], may contribute to this pathology. Recent findings extended the concept of afferent convergence to second order neurons in the trigeminal brainstem complex. Through histological examinations, tracing studies and electrophysiological recordings in rodent and human calvaria convincing evidence was collected showing that the periosteum and deep layers of pericranial muscles including the temporal muscle are innervated by collaterals of meningeal afferents that traverse the dura mater and project through fissures and emissary canals of the skull to these extracranial tissues [31, 45]. Several of these single afferents could be activated both from the inside and the outside of the rat skull, noxious stimulation of the temporal muscle caused neuropeptide release from the dura mater, and collision experiments proved that at least some of these intra- and extracranial afferent fibres belong to same afferent neurons [46]. Thus, in addition to convergent afferent input, the second order neurons processing muscle pain may be driven by primary nociceptive afferents that innervate both the dura mater and the temporal muscle. Recently there is evidence that also neck muscles can be supplied by collaterals of meningeal afferents, since retrograde tracing from neck muscles and the dura mater of the occipital cranial fossa showed convergence of the tracer in some trigeminal ganglion neurons (unpublished data from our group). Inhibition of the afferent muscle input could be rather effective in reducing the activity of the central neurons, as evidenced in the present study using local injection of anesthetics into the temporal muscle. Similarly, afferent nerve blockade such as anaesthesia of the greater occipital nerve has been proposed to reduce nociceptive input and hence headache and has been used with equivocal success in primary headaches [47, 48] but is recently used as a therapeutic option in refractory headache [49], cluster headache [50] and chronic migraine [51]. This concept is supported by clinical experiences and therapeutic efforts in tension-type headache and other primary headaches, which aim at decreasing muscle tension and nociceptive input from pericranial muscles [52, 53]. Conclusions Injection of α,β-meATP into the temporal muscle in rat induces ongoing activity of spinal trigeminal neurons with meningeal receptive fields and local anaesthesia of single neck muscles and/or the temporal muscle can abolish this activation. This shows that distinct spinal trigeminal neurons processing meningeal nociceptive information are under the control of convergent afferent input from several pericranial muscles, which may already occur at the primary afferent level by collaterals of meningeal nociceptors innervating pericranial muscle compartments. Blockade of one of these inputs can normalize central trigeminal activity providing an explanation for the beneficial effect of nerve blockade and extracranial therapeutic manipulations in primary headaches. Abbreviations ANOVAAnalysis of variance ATPAdenosine 5′-triphosphate LSD testFisher’s least significant difference test TTHTension-type headache α,β-meATPα,β-methylene adenosine 5′-triphosphate Acknowledgements Authors thank Jana Schramm for excellent technical assistance. The study was supported by the FP7grant 602633 (EUROHEADPAIN) of the European Union. Authors’ contributions MN performed the experiments, analyzed the data and outlined the manuscript. SF contributed to the experiments, the data analysis and the manuscript. JE contributed to the study design and the manuscript. KM designed the study, supervised the experimental work and the data analysis and refined the manuscript. All authors read and approved the final manuscript. Competing interests Authors declare no competing interests and have nothing to disclose. The first author (M.N.) performed this work for obtaining the degree “Dr. med. dent.” at the Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU). ==== Refs References 1. Bendtsen L Jensen R Tension-type headache: the most common, but also the most neglected, headache disorder Curr Opin Neurol 2006 19 305 309 16702840 2. Headache Classification Committee of the International Headache Society (IHS) The International Classification of Headache Disorders, 3rd edition (beta version) Cephalalgia 2013 33 629 808 10.1177/0333102413485658 23771276 3. Jensen R Pathophysiological mechanisms of tension-type headache: a review of epidemiological and experimental studies Cephalalgia Int J Headache 1999 19 602 621 10.1046/j.1468-2982.1999.019006602.x 4. Bendtsen L Central sensitization in tension-type headache--possible pathophysiological mechanisms Cephalalgia Int J Headache 2000 20 486 508 10.1046/j.1468-2982.2000.00070.x 5. Bartsch T Goadsby PJ Stimulation of the greater occipital nerve induces increased central excitability of dural afferent input Brain J Neurol 2002 125 1496 1509 10.1093/brain/awf166 6. Makowska A Panfil C Ellrich J ATP induces sustained facilitation of craniofacial nociception through P2X receptors on neck muscle nociceptors in mice Cephalalgia Int J Headache 2006 26 697 706 10.1111/j.1468-2982.2006.01095.x 7. Jensen K Tuxen C Olesen J Pericranial muscle tenderness and pressure-pain threshold in the temporal region during common migraine Pain 1988 35 65 70 10.1016/0304-3959(88)90277-1 3200599 8. Jensen R Rasmussen BK Pedersen B Olesen J Muscle tenderness and pressure pain thresholds in headache. A population study Pain 1993 52 193 199 10.1016/0304-3959(93)90131-8 8455967 9. Ashina M Bendtsen L Jensen R Muscle hardness in patients with chronic tension-type headache: relation to actual headache state Pain 1999 79 201 205 10.1016/S0304-3959(98)00167-5 10068165 10. Aaseth K Grande RB Lundqvist C Russell MB Pericranial tenderness in chronic tension-type headache: the Akershus population-based study of chronic headache J Headache Pain 2014 15 58 10.1186/1129-2377-15-58 25193401 11. Kim HS Chung SC Kim YK Lee SW Pain-pressure threshold in the head and neck region of episodic tension-type headache patients J Orofac Pain 1995 9 357 364 8995907 12. Watson DH Drummond PD Head pain referral during examination of the neck in migraine and tension-type headache Headache 2012 52 1226 1235 10.1111/j.1526-4610.2012.02169.x 22607581 13. Cooke R Bialek W Contraction of glycerinated muscle fibers as a function of the ATP concentration Biophys J 1979 28 241 258 10.1016/S0006-3495(79)85174-7 262549 14. Davies NW Modulation of ATP-sensitive K+ channels in skeletal muscle by intracellular protons Nature 1990 343 375 377 10.1038/343375a0 2153936 15. Li J King NC Sinoway LI ATP concentrations and muscle tension increase linearly with muscle contraction J Appl Physiol Bethesda Md 1985 2003 95 577 583 16. Taguchi T Kozaki Y Katanosaka K Mizumura K Compression-induced ATP release from rat skeletal muscle with and without lengthening contraction Neurosci Lett 2008 434 277 281 10.1016/j.neulet.2008.01.068 18313220 17. Hellsten Y Maclean D Rådegran G Adenosine concentrations in the interstitium of resting and contracting human skeletal muscle Circulation 1998 98 6 8 10.1161/01.CIR.98.1.6 9665052 18. Bodin P Burnstock G Increased release of ATP from endothelial cells during acute inflammation Inflamm Res 1998 47 351 354 10.1007/s000110050341 9754870 19. Bodin P Burnstock G Purinergic Signalling: ATP Release Neurochem Res 2001 26 959 969 10.1023/A:1012388618693 11699948 20. Mense S The pathogenesis of muscle pain Curr Pain Headache Rep 2003 7 419 425 10.1007/s11916-003-0057-6 14604500 21. Shinoda M Ozaki N Sugiura Y Involvement of ATP and its receptors on nociception in rat model of masseter muscle pain Pain 2008 134 148 157 10.1016/j.pain.2007.04.006 17521813 22. Reinöhl J Hoheisel U Unger T Mense S Adenosine triphosphate as a stimulant for nociceptive and non-nociceptive muscle group IV receptors in the rat Neurosci Lett 2003 338 25 28 10.1016/S0304-3940(02)01360-5 12565132 23. Matsuda T Kubo A Taguchi T Mizumura K ATP decreases mechanical sensitivity of muscle thin-fiber afferents in rats Neurosci Res 2015 97 36 44 10.1016/j.neures.2015.04.001 25862944 24. Cheng DY DeWitt BJ Suzuki F Adenosine A1 and A2 receptors mediate tone-dependent responses in feline pulmonary vascular bed Am J Physiol 1996 270 H200 H207 8769752 25. Dang K Bielefeldt K Gebhart GF Differential responses of bladder lumbosacral and thoracolumbar dorsal root ganglion neurons to purinergic agonists, protons, and capsaicin J Neurosci Off J Soc Neurosci 2005 25 3973 3984 10.1523/JNEUROSCI.5239-04.2005 26. Hanna RL Kaufman MP Activation of thin-fiber muscle afferents by a P2X agonist in cats J Appl Physiol Bethesda Md 1985 2004 96 1166 1169 27. Burgard EC Niforatos W van Biesen T P2X receptor-mediated ionic currents in dorsal root ganglion neurons J Neurophysiol 1999 82 1590 1598 10482772 28. Ellrich J Makowska A Nerve growth factor and ATP excite different neck muscle nociceptors in anaesthetized mice Cephalalgia Int J Headache 2007 27 1226 1235 10.1111/j.1468-2982.2007.01431.x 29. Goadsby PJ Zagami AS Lambert GA Neural processing of craniovascular pain: a synthesis of the central structures involved in migraine Headache 1991 31 365 371 10.1111/j.1526-4610.1991.hed3106365.x 1889975 30. Sugimoto T Fujiyoshi Y He YF Trigeminal primary projection to the rat brain stem sensory trigeminal nuclear complex and surrounding structures revealed by anterograde transport of cholera toxin B subunit-conjugated and Bandeiraea simplicifolia isolectin B4-conjugated horseradish peroxidase Neurosci Res 1997 28 361 371 10.1016/S0168-0102(97)00064-3 9274832 31. Schueler M Neuhuber WL De Col R Messlinger K Innervation of rat and human dura mater and pericranial tissues in the parieto-temporal region by meningeal afferents Headache 2014 54 996 1009 10.1111/head.12371 24673461 32. Roch M Messlinger K Kulchitsky V Ongoing activity in trigeminal wide-dynamic range neurons is driven from the periphery Neuroscience 2007 150 681 691 10.1016/j.neuroscience.2007.09.032 18023985 33. Saloman JL Chung M-K Ro JY P2X3 and TRPV1 functionally interact and mediate sensitization of trigeminal sensory neurons Neuroscience 2013 232 226 238 10.1016/j.neuroscience.2012.11.015 23201260 34. Reitz M Makowska A Ellrich J Excitatory and inhibitory purinergic control of neck muscle nociception in anaesthetized mice Cephalalgia Int J Headache 2009 29 58 67 10.1111/j.1468-2982.2008.01700.x 35. Ristic D Spangenberg P Ellrich J Inhibition of nNOS prevents and inhibition of iNOS reverses α, β-meATP-induced facilitation of neck muscle nociception in mice Eur J Pharmacol 2010 647 55 61 10.1016/j.ejphar.2010.08.009 20813105 36. Ristic D Spangenberg P Ellrich J Acetylsalicylic acid inhibits α, β-meATP-induced facilitation of neck muscle nociception in mice--implications for acute treatment of tension-type headache Eur J Pharmacol 2011 673 13 19 10.1016/j.ejphar.2011.10.008 22032900 37. Li J Sinoway LI ATP stimulates chemically sensitive and sensitizes mechanically sensitive afferents Am J Physiol Heart Circ Physiol 2002 283 H2636 H2643 10.1152/ajpheart.00395.2002 12427602 38. Cherkas PS Dostrovsky JO Sessle BJ Activation of peripheral P2X receptors is sufficient to induce central sensitization in rat medullary dorsal horn nociceptive neurons Neurosci Lett 2012 526 160 163 10.1016/j.neulet.2012.08.007 22910609 39. Davis KD Dostrovsky JO Responses of feline trigeminal spinal tract nucleus neurons to stimulation of the middle meningeal artery and sagittal sinus J Neurophysiol 1988 59 648 666 3351579 40. Schepelmann K Ebersberger A Pawlak M Response properties of trigeminal brain stem neurons with input from dura mater encephali in the rat Neuroscience 1999 90 543 554 10.1016/S0306-4522(98)00423-0 10215158 41. Strassman A Mason P Moskowitz M Maciewicz R Response of brainstem trigeminal neurons to electrical stimulation of the dura Brain Res 1986 379 242 250 10.1016/0006-8993(86)90777-8 3742223 42. Burstein R Yamamura H Malick A Strassman AM Chemical stimulation of the intracranial dura induces enhanced responses to facial stimulation in brain stem trigeminal neurons J Neurophysiol 1998 79 964 982 9463456 43. Bartsch T Goadsby PJ Increased responses in trigeminocervical nociceptive neurons to cervical input after stimulation of the dura mater Brain J Neurol 2003 126 1801 1813 10.1093/brain/awg190 44. Goadsby PJ Bartsch T On the functional neuroanatomy of neck pain Cephalalgia Int J Headache 2008 28 Suppl 1 1 7 10.1111/j.1468-2982.2008.01606.x 45. Kosaras B Jakubowski M Kainz V Burstein R Sensory innervation of the calvarial bones of the mouse J Comp Neurol 2009 515 331 348 19425099 46. Schueler M Messlinger K Dux M Extracranial projections of meningeal afferents and their impact on meningeal nociception and headache Pain 2013 154 1622 1631 10.1016/j.pain.2013.04.040 23707274 47. Busch V Jakob W Juergens T Occipital nerve blockade in chronic cluster headache patients and functional connectivity between trigeminal and occipital nerves Cephalalgia Int J Headache 2007 27 1206 1214 10.1111/j.1468-2982.2007.01424.x 48. Ashkenazi A Levin M Greater occipital nerve block for migraine and other headaches: is it useful? Curr Pain Headache Rep 2007 11 231 235 10.1007/s11916-007-0195-3 17504651 49. Martelletti P Giamberardino MA Mitsikostas D-D Greater occipital nerve as target for refractory chronic headaches: from corticosteroid block to invasive neurostimulation and back Expert Rev Neurother 2016 16 1 2 10.1586/14737175.2016.1164599 26618372 50. Lambru G Abu Bakar N Stahlhut L Greater occipital nerve blocks in chronic cluster headache: a prospective open-label study Eur J Neurol 2014 21 338 343 10.1111/ene.12321 24313966 51. Inan LE Inan N Karadaş Ö Greater occipital nerve blockade for the treatment of chronic migraine: a randomized, multicenter, double-blind, and placebo-controlled study Acta Neurol Scand 2015 132 270 277 10.1111/ane.12393 25765043 52. van Ettekoven H Lucas C Efficacy of physiotherapy including a craniocervical training programme for tension-type headache; a randomized clinical trial Cephalalgia Int J Headache 2006 26 983 991 10.1111/j.1468-2982.2006.01163.x 53. Moshkani Farahani D Tavallaie SA Ahmadi K Fathi Ashtiani A Comparison of neurofeedback and transcutaneous electrical nerve stimulation efficacy on treatment of primary headaches: a randomized controlled clinical trial Iran Red Crescent Med J 2014 16 e17799 10.5812/ircmj.17799 25389484

PMC005xxxxxx/PMC5001962.txt

==== Front SpringerplusSpringerplusSpringerPlus2193-1801Springer International Publishing Cham 309210.1186/s40064-016-3092-6ResearchA split-optimization approach for obtaining multiple solutions in single-objective process parameter optimization Rajora Manik mrajora3@gatech.eduhttp://www.gatech.edu 1Zou Pan zoupzp123@gatech.eduhttp://www.dhu.edu.cn 2Yang Yao Guang max.yang@mail.chmer.org.twhttp://www.mirdc.org.tw 3Fan Zhi Wen zwfen@mail.mirdc.org.twhttp://www.mirdc.org.tw 3Chen Hung Yi hyc@mail.mirdc.org.twhttp://www.mirdc.org.tw 3Wu Wen Chieh wcwu@mail.mirdc.org.twhttp://www.mirdc.org.tw 3Li Beizhi lbzhi@dhu.edu.cnhttp://www.dhu.edu.cn 2Liang Steven Y. steven.liang@me.gatech.eduhttp://www.gatech.edu 121 George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA 30332 USA 2 Mechanical Engineering College, Donghua University, Songjiang District, Shanghai, 201620 China 3 Regional R&D Services Department, Metal Industries Research and Development Center, Taichung, 407 Taiwan, ROC 26 8 2016 26 8 2016 2016 5 1 142431 3 2016 17 8 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. It can be observed from the experimental data of different processes that different process parameter combinations can lead to the same performance indicators, but during the optimization of process parameters, using current techniques, only one of these combinations can be found when a given objective function is specified. The combination of process parameters obtained after optimization may not always be applicable in actual production or may lead to undesired experimental conditions. In this paper, a split-optimization approach is proposed for obtaining multiple solutions in a single-objective process parameter optimization problem. This is accomplished by splitting the original search space into smaller sub-search spaces and using GA in each sub-search space to optimize the process parameters. Two different methods, i.e., cluster centers and hill and valley splitting strategy, were used to split the original search space, and their efficiency was measured against a method in which the original search space is split into equal smaller sub-search spaces. The proposed approach was used to obtain multiple optimal process parameter combinations for electrochemical micro-machining. The result obtained from the case study showed that the cluster centers and hill and valley splitting strategies were more efficient in splitting the original search space than the method in which the original search space is divided into smaller equal sub-search spaces. Keywords Split-optimizationMultiple solutionsProcess parameter optimizationGenetic algorithm (GA)Electrochemical micro-machining (EMM)issue-copyright-statement© The Author(s) 2016 ==== Body Background In today’s rapidly changing scenario in the manufacturing industries, optimization of process parameters is essential for a manufacturing unit to respond effectively to the severe competitiveness and increasing demand for quality products in the market (Cook et al. 2000). Previously, to obtain optimal combinations of input process parameters, engineers used a trial-and-error-based approach, which relied on engineers’ experience and intuition. However, the trial-and-error-based approach is expensive and time consuming; thus, it is not suitable for complex manufacturing processes (Chen et al. 2009). Thus, researchers have focused their attention on developing alternate methods to the trial-and-error-based approach that can help engineers obtain the combination of process parameters that will minimize or maximize the desired objective value for a given process. The methods for obtaining these combinations of process parameters can be split into 2 main categories: 1. forward mapping of process inputs to a performance indicator with backwards optimization and 2. reverse mapping between the performance indicators and the process inputs. In forward mappings, first, a model is created between the process inputs and the performance indicators using either physics-based models, regressions models, or intelligent techniques. Once a satisfactory model has been created, it is then utilized to obtain the combination of process parameters that will lead to a desired value of the output using optimization techniques such as the Genetic algorithm (GA), Simulated Annealing (SA), Particle Swarm Optimization (PSO), etc. The desired output can either be to a. minimize a given performance indicator or b. reach a desired level of a performance indicator. Chen et al. (2009) utilized the back propagation neural network (BPNN) and GA to create a forward prediction model and optimize the process parameters of plastic injection molding. Ylidiz (2013) utilized a hybrid artificial bee colony-based approach for selecting the optimal process parameters for multi-pass turning that would minimize the machining cost. Senthilkumaar et al. (2012) used mathematical models and ANN to map the relationship between the process inputs and performance indicators for finish turning and facing of Inconel 718. GA was then used to find the optimal combination of process parameters, with the aim of minimizing surface roughness and flank wear. Pawar and Rao (2013) applied the teaching–learning-based optimization (TLBO) algorithm to optimize the process parameters of abrasive water jet machining, grinding, and milling. They created physics-based models between the input and output parameters of each process and then utilized TLBO to minimize the material removal rate in abrasive water jets, minimize production cost and maximize production rate with respect to grinding, and minimize the production time in milling. Fard et al. (2013) employed adaptive network-based fuzzy inference systems (ANFIS) to model the process of dry wire electrical discharge machining (WEDM). This model was then used to optimize, using artificial bee colony (ABC), the process inputs that would minimize surface roughness and maximize material removal rate. Teixidor et al. (2013) used particle swarm optimization (PSO) to obtain optimal process parameters that would minimize the depth error, width error, and surface roughness in the pulsed laser milling of micro-channels on AISI H13 tool steel. Katherasan et al. (2014) used ANN to model the process of flux cored arc welding (FCAW) and then utilized PSO to minimize bead width and reinforcement and maximize depth of penetration. Yusup et al. (2014) created a regression model for the process parameters and process indicators of an abrasive waterjet (AWJ) and then used ABC to minimize the surface roughness. Panda and Yadava (2012) used ANN to model the process of die sinking electrochemical spark machining (DS-ESM) and then used GA for multi-objective optimization of the material removal rate and average surface roughness. Maji and Pratihar (2010) combined ANFIS with GA to create forward and backward input–output relationships for the electrical discharge machining process (EDM). In their proposed methodology, GA was used to optimize the membership functions of the ANFIS, with the aim of minimizing the error between the predicted and actual outputs. Cus et al. (2006) developed an intelligent system for online monitoring and optimization of process parameters in the ball-end milling process. Their objective was to find the optimal set of process parameters, using GA to achieve the forces selected by the user. Raja et al. (2015) optimized the process parameters of electric discharge machining (EDM) using the firefly algorithm to obtain the desired surface roughness in the minimum possible machining time. Raja and Baskar (2012) used PSO to optimize the process parameters to achieve the desired surface roughness while minimizing machining time for face milling. Rao and Pawar (2009) developed mathematical models using response surface modeling (RSM) to correlate the process inputs and performance indicators of WEDM. They then used ABC to achieve the maximum machining speed that would give the desired value of the surface finish. Lee et al. (2007) modeled the process of high-speed finish milling using a 2 stage ANN and then used GA to maximize the surface finish while achieving the desired material removal rate. Teimouri and Baseri (2015) used a combination of fuzzy logic and the artificial bee colony algorithm to create a forward prediction model between input and output parameters for friction stir welding (FSW). This trained model was then utilized to find the optimal input parameters that would give the desired output value by minimizing the absolute error between the predicted and specified output using the imperialist competitive algorithm (ICA). An ample amount of work has also been done to create a reverse mapping model between the process parameters and the performance indicators. Parappagoudar et al. (2008) utilized the back-propagation neural network (BPNN) and a genetic-neural network (GA-NN) for forward and reverse mapping of the process parameters and performance indicators in a green sand mold system. Parappagoudar et al. (2008) also extended their application of BPNN and GA-NN to create forward and backward mappings for the process of the Sodium Silicate-Bonded, Carbon Dioxide Gas Hardened Molding Sand System. Amarnath and Pratihar (2009) used radial basis function neural networks (RBFNNs) for forward and reverse input–output mapping of the tungsten inert gas (TIG) welding process. In their proposed methodology, the structure and the parameters of the RBFNN were modified using a combination of GA and the fuzzy C-means (FCM) algorithm for both the forward and reverse mapping. Chandrashekarappa et al. (2014) used BPNN and GA-NN for forward and reverse mappings of the squeeze casting process. Kittur and Parappagoudar (2012) utilized BPNN and GA-NN for forward and reverse mapping in the die casting process. Because batch training requires a tremendous amount of data, they used previously generated equations to supplement the experimental data. Malakooti and Raman (2000) used ANN to create forward- and backward-direction mappings between the process outputs and inputs for the cutting operation on a lathe. Even though extensive research has been done regarding optimization of the process parameter for different processes, the current algorithms used for the optimization procedure are limited to finding only one set of optimal process parameter combinations for a single-objective optimization problem each time the algorithms are executed. Though this process parameter combination may achieve the desired output, it may not always be suitable for actual production or may lead to undesirable experimental conditions. It can also be observed from the experimental data of different processes that different process parameter combinations may lead to the same or similar performance indicators. For example, in turning, multiple combinations of process parameters may lead to the same or similar value of surface roughness. In EMM, multiple combinations of process parameters may lead to the same or similar value of taper and overcut. Therefore, there is a possibility to develop a method that can provide multiple optimal process parameter combinations for a single-objective optimization problem. In this paper, the presented method is to obtain multiple optimal process parameter combinations for a single-objective optimization problem by splitting the original search space into smaller sub-search spaces and finding the optimal process parameter combinations in each sub-search space. Two different methods are used to split the original search space, and GA is utilized to optimize the process parameters in each sub-search space. The optimization results obtained after using the two search space splitting methods are compared to the optimization results obtained when the original search space was divided equally into smaller sub-search spaces; GA was used to optimize the process parameters in each sub-search space. EMM of SUS 304 is used as a case study because its experimental data shows that multiple process parameter combinations can lead to the same performance indicators. Due to the lack of physics-based models, a general regression neural network (GRNN) is used to create a forward prediction model between the input process parameters and the performance indicators for the process of EMM. The rest of the paper is organized as follows: section “Modeling” describes the modeling stage of the method. Section “Case study” presents and discusses the results obtained. Section “Conclusion” presents conclusions from the presented work and mentions future directions for the proposed approach. Modeling Split-optimization approach Traditional GA, when used in a single-objective optimization, only converges to a single local optima or near-optimum solution, while the search space might consist of multiple local optima that can satisfy the given criteria. Multi-objective GA, on the other hand, does provide multiple solutions, but each solution satisfies each objective to a different degree. A possible method to obtain multiple solutions for a single-objective optimization problem is to split the original search space into several smaller sub-search spaces, with each sub-search space containing a possible solution to the given objective. Once these sub-search spaces have been identified, GA can then be used in each sub-search space to find the possible solution. The procedure of the proposed split-optimization approach consists of two parts: splitting of the original search spaces into sub-search spaces and the application of GA to find the solution in each sub-search space. Because any optimization function needs a fitness function as an input, in this paper, a generalized neural network (GRNN) was used as the fitness function due to a lack of physics-based models for the given process. The flow chart of this proposed split-optimization approach is shown in Fig. 1.Fig. 1 The basic structure of the split-optimization approach Because the results obtained after using GA depend on the training accuracy of the GRNN, it is important to train the GRNN sufficiently so that it can predict the performance indicators with a high degree of accuracy. As there will always be some degree of error associated with the outputs of the GRNN, a possible method to cope with these errors is to take into consideration the significance level of the optimization problem. The significance level here is defined as a customized parameter that allows solutions with a fitness value better than or equal to it to be counted as final optimal solutions. The significance level by default is regarded as zero, which indicates that only solutions with the same minimum fitness value can be regarded as the final optimum solutions. Splitting strategies As mentioned earlier, two strategies are used to split the original search space into sub-search spaces. The details of the two strategies are highlighted below.Hill and valley splitting strategy The steps of the splitting strategy are as follows:Identify two data points, A and B, from the experimental data set whose input values are furthest away from each other. Here, A=a1,a2,⋯,an,ya and B=b1,b2,⋯,bn,yb, indicating that all the data points have n inputs and 1 output. Select a random data point C1 from the remaining data points and determine whether it is a hill, valley, or neither compared to the initial points, i.e., A and B, based on the value of its output. For example, if ya < yb < yc1, then C1 is a hill; if yc1 < ya < yb, then C1 is a valley; if ya < yc1 < yb, then C1 is neither. Select a random data point C2 from the remaining data points; find a pair of previously selected data points whose input values encompass the input values of C2. Compare the output value of C2 with the data points selected in step c and determine whether it is a hill, valley, or neither. Repeat step c and d until all the data points have been identified as a hill, valley, or neither. After the classification of all the experimental data points is completed, the input values of the original points (A and B) and all the points classified as either a hill or valley are used to split the original search space into smaller sub-search spaces. This is done by dividing the original range of the input parameters of the experimental data into sub-ranges by using the input values of the points classified as hill or valley and then finding all the combinations of the sub-ranges for all the inputs. Once the search space has been split into sub-search spaces, GA is used to optimize each search space individually. Figure 2 shows the flow chart of the hill and valley splitting strategy. Fig. 2 The flow chart of the hill and valley splitting strategy (2) Cluster centers splitting strategy In this strategy, the k-means clustering algorithm is used to divide the experimental data set into k clusters. Once the k cluster centers are identified, they are used to split the original search space into smaller sub-search spaces. This is accomplished by dividing the original range of input process parameters of the experimental data into smaller sub-ranges using the values of the k cluster centers. Next, the original search space is divided by using all the combinations of the sub-ranges for all the inputs. Figure 3 shows the flow chart of the cluster centers splitting strategy.Fig. 3 The flow chart of the cluster centers splitting strategy GRNN-GA optimization As mentioned earlier, a forward prediction model was created using GRNN (Specht 1991). The inputs of the GRNN were voltage, pulse on time, and feed rate; the outputs were Din and Dout. During the training of the GRNN, the original data was split into training, validation, and testing data sets, and tenfold cross validation was used during the training phase of the GRNN to avoid overfitting and to find the optimal value of the spread parameter that would minimize the mean squared error (MSE). Once the GRNN was trained sufficiently, it was then utilized as the fitness function for GA during the optimization procedure. Case study An input parameter optimization problem in EMM was utilized as a case study because it can be seen from Table 1 that multiple combinations of input process parameters lead to the same or similar values of taper and overcut.Table 1 Original range of the controllable process parameters Process parameter Voltage (V) Pulse on time (µs) Feed rate (µm/s) Lower bound 8 25 4 Upper bound 20 70 12 Description of the case Figure 4 schematically depicts the EMM experimental setup. The system consists of a three-dimensional movement device, a small-scale power supply of 100 A, and an electrolyte pump and filter. The feeding system is controlled by a PC-Based CNC Controller, RTX real-time windows kernel program, and a motion card that drives the linear motor precisely. A pulse generator supplies a periodic current to the experimental model. A digital oscilloscope ensures that the pulse generator produces a rectangular waveform with accurate amplitude. If the tool feed rate is excessive, the tool will contact the workpiece and cause a short circuit; thus, an oscilloscope is employed to detect any short circuits. Whenever the oscilloscope detects a short circuit, a signal is sent rapidly to the PC and the tool is extracted automatically until the measured voltage returns to the applied voltage. The micro array holes electrode module includes the multiple nozzle tool electrodes, PVC mask and tool fixture. The electrolyte is pumped to a multiple electrode cell and exits through the small nozzle in the form of a free standing jet directed towards the anode workpiece.Fig. 4 Schematic diagram of electrochemical micromachining system (left) and micro array hole electrode module (right) Other basic information and settings are as follows: the electrolyte velocity was 10 m/s, electrolyte temperature was 27 °C, the initial gap between the tool and the workpiece was 100 µm, tool moving distance was 800 µm, the workpiece material was SUS 304, the electrolyte used was 10 %wt. NaNO3, the nominal diameter of the hole was 900 µm, and the depth of the hole was 500 µm. Voltage, pulse on time, and feed rate were used as the controllable process parameters, while the inner diameter of the micro-hole Din and the outer diameter Dout were the measurable performances. The range of each process parameter is shown in Table 1. The range of the variables was fixed by taking into consideration two factors: 1. limitation of the devices used for EMM and 2. making sure that the experimental conditions would be stable within the chosen range. The resolution of the process parameters were was 0.1 V for the voltage, 0.1 µs for pulse on time, and 0.1 µm/s for the feed rate. This indicates that there are close to 3 million possible combinations of all the process parameters. Therefore, the proposed method was applied for this particular case study. The process of EMM has two responses, i.e., taper and overcut. When drilling micro-size holes in thin metallic foils, a major requirement is for the holes to have straight walls. The straightness of a wall can be represented by the taper and is given by: 1 Taper=(Din-Dout)/depth In critical applications, particularly in micro instruments, the straightness of a drilled hole is also very important. Overcut, as given by Eq. (2), is the difference between the aim holes’ diameters and actual hole diameter and is a good representation of the straightness of a drilled hole. A small overcut value represents a more precise EMM process. 2 Overcut=(Din-D)/2 In the process of EMM, the aim is to find the set of process parameter combinations that will minimize both taper and overcut. Though EMM has two responses, for the purpose of this case study, the two responses were combined into a single-objective by the use of weight values. Before combining them into a single objective, the values of taper and overcut were normalized between 0 and 1. Equation (3) shows how the taper and overcut were normalized, while Eq. (4) shows the objective function. 3 Tnormalized=Tpredicted-Tmin,experimentalTmax,experimental-Tmin,experimental,Onormalized=Opredicted-Omin,experimentalOmax,experimental-Omin,experimental where Tpredicted is the taper value predicted by the GRNN, Tmin,experimental is the minimum taper value in the experimental data, Tmax,experimental is the maximum taper value in the experimental data, and Tnormalized is the normalized predicted taper value. Similarly, Opredicted is the overcut value predicted by the GRNN, Omin,experimental is the minimum overcut value in the experimental data, Omax,experimental is the maximum overcut value in the experimental data, and Onormalized is the normalized predicted overcut value. 4 Objective=0.5×Tnormalized+0.5×Onormalized To create a forward prediction model for the process of EMM, three different sets of experiments were created. In the first experimental set, voltage and feed rate had 3 levels each, while pulse on time was constant, which resulted in a total of 9 combinations of input parameters. These combinations of input experiments were used to perform the process of EMM, and for each combination, Din and Dout were recorded. In the second and third experimental sets, voltage, pulse on time, and feed rate had 3 levels each, which resulted in 27 combinations of input process parameters for both experimental sets 2 and 3. The process of EMM was performed using the combination of inputs; Din and Dout were again recorded. The levels of voltage, pulse on time, and feed rate are given in Table 2.Table 2 Levels of voltage, pulse on time, and feed rate values used for the three experimental sets Experimental set # Levels of voltage (V) Levels pulse on time (µs) Levels feed rate (µm/s) 1 [16, 18, 20] 25 [4, 6, 8] 2 [4, 6, 8] [50, 60, 70] [8, 10, 12] 3 [4, 6, 8] [50, 60, 70] [8–10] In the experiments, the Charge Coupled Device (CCD) camera was utilized to measure all the workpieces after the process of EMM. Figure 5 shows the pictures taken using the CCD camera. The CCD images were then processed through a software, which provided the average value of the diameters of the holes on the front and back of the workpiece. The experimental data obtained is shown in Table 3.Fig. 5 Pictures taken using the CCD camera. a The front of the workpiece, while b. The back of the workpiece Table 3 The 63 groups of experimental data No. Voltage (V) Pulse on time (µs) Feed rate (µm/s) D in (µm) D out (µm) Taper Overcut (µm) 1 16 25 8 893 860 0.066 3.5 2 18 25 8 929 913 0.032 14.5 3 20 25 8 923 910 0.026 11.5 4 16 25 6 904 892 0.024 2 5 18 25 6 934 931 0.006 17 6 20 25 6 999 977 0.044 49.5 7 16 25 4 983 979 0.008 41.5 8 18 25 4 1050 1045 0.01 75 9 20 25 4 1125 1123 0.004 112.5 10 8 50 8 657.5 627.5 0.06 121.25 11 10 50 8 809.5 807.25 0.0045 45.25 12 12 50 8 866.25 858 0.0165 16.875 13 8 50 6 760 741 0.038 70 14 10 50 6 828.5 829.5 0.002 35.75 15 12 50 6 908.75 905.5 0.0065 4.375 16 8 50 4 781.75 780.25 0.003 59.125 17 10 50 4 887.25 881.75 0.011 6.375 18 12 50 4 957.75 970 0.0245 28.875 19 8 60 8 771.33 759.33 0.024 64.335 20 10 60 8 806.75 799.5 0.0145 46.625 21 12 60 8 862.75 847 0.0315 18.625 22 8 60 6 756.5 739.75 0.0335 71.75 23 10 60 6 776.75 777.5 0.0015 61.625 24 12 60 6 840.25 841.25 0.002 29.875 25 8 60 4 769 771.5 0.005 65.5 26 10 60 4 854.75 865.25 0.021 22.625 27 12 60 4 928.25 945.5 0.0345 14.125 28 8 70 8 718 721.5 0.007 91 29 10 70 8 779 796.75 0.0355 60.5 30 12 70 8 841.5 849.75 0.0165 29.25 31 8 70 6 736.5 744.5 0.016 81.75 32 10 70 6 802 829.75 0.0555 49 33 12 70 6 858.75 865 0.0125 20.625 34 8 70 4 783.25 783.25 0 58.375 35 10 70 4 878.75 872 0.0135 10.625 36 12 70 4 946.25 955.25 0.018 23.125 37 8 50 8 874 704 0.34 13 38 9 50 8 914 789 0.25 7 39 10 50 8 999 827 0.344 49.5 40 8 50 6 922 765 0.314 11 41 9 50 6 955 807 0.296 27.5 42 10 50 6 1039 837 0.404 69.5 43 8 50 4 932 797 0.27 16 44 9 50 4 1044 790 0.508 72 45 10 50 4 1130 858 0.544 115 46 8 60 8 903 708 0.39 1.5 47 9 60 8 967 766 0.402 33.5 48 10 60 8 1084 817 0.534 92 49 8 60 6 917 760 0.314 8.5 50 9 60 6 1043 856 0.374 71.5 51 10 60 6 1115 871 0.488 107.5 52 8 60 4 1071 754 0.634 85.5 53 9 60 4 1087 972 0.23 93.5 54 10 60 4 1263 1044 0.438 181.5 55 8 70 8 875 789 0.172 12.5 56 9 70 8 1071 842 0.458 85.5 57 10 70 8 1158 862 0.592 129 58 8 70 6 987 846 0.282 43.5 59 9 70 6 1212 886 0.652 156 60 10 70 6 1243 1056 0.374 171.5 61 8 70 4 1134 877 0.514 117 62 9 70 4 1260 935 0.65 180 63 10 70 4 1348 1016 0.664 224 Results As stated earlier, to compensate for the errors associated with the trained GRNN, a significance level needs to be specified. In this case study, if the value of the objective function, given by Eq. (4), after optimization was less than 0.5, then the solution of that particular sub-search space was said to be a final optimal solution. The only changeable parameter for the GRNN was the spread value, which was obtained after the training process. The changeable parameters for GA are listed in Table 4.Table 4 Parameter values used for GA Number of generations Population size Crossover fraction Mutation fraction Elite count 100 50 0.85 0.15 3 The methods mentioned above were used to split and optimize the search space 10 times independently, and the average value of the objective function for the best solutions of each run was calculated. The run that had the lowest average value of the objective function was used as the best run; its results are presented here.Hill and valley spitting strategy As mentioned previously, the first step in this method is to find two data points that are furthest away from each other. To accomplish this task, the distance from the origin to every data point was obtained after each input was normalized using Eq. (6). The equations used to normalize the inputs are given in Eq. (5). The two data points with distances dmin and dmax were the inputs furthest away from each other. Then, the steps outlined in the previous section were followed to split the original search space into several sub-search spaces. 5 Vi,normalized=Vi-Vmin,experimentalVmax,experimental-Vmin,experimental,Pi,normalized=Pi-Pmin,experimentalPmax,experimental-Pmin,experimental,Fi,normalized=Fi-Fmin,experimentalFmax,experimental-Fmin,experimental where Vi,normalized is the normalized values of voltage in the ith experimental data, Vmin,experimental is the minimum voltage value in the experimental data, Vmax,experimental is the maximum voltage value in the experimental data, and Vi is the voltage in the ith experimental data. Similarly, Pi,normalized is the normalized values of pulse on time in the ith experimental data, Pmin,experimental is the minimum pulse on time value in the experimental data, Pmax,experimental is the maximum pulse on time value in the experimental data, and Pi is the pulse on time in the ith experimental data. Fi,normalized is the normalized values of feed rate time in the ith experimental data, Fmin,experimental is the minimum feed rate value in the experimental data, Fmax,experimental is the maximum feed rate value in the experimental data, and Fi is the feed rate in the ith experimental data. 6 Di=Vi,normalized2+(Pi,normalized)2+(Fi,normalized)2 Table 5 provides the ranges for each of the input values. For each of the sub-search spaces, GA was utilized to find the optimal process parameter combination. The optimization results are shown in Table 6.Table 5 Splitting result of the hill and valley splitting strategy Number of hills Number of valleys Sub-range of voltage (V) Sub-range of pulse on time (µs) Sub-range of feed rate (µm/s) Number of sub-search spaces 29 14 [8, 9]; [9, 10]; [10, 18]; [18, 20] [25,50]; [50,60]; [60,70] [4, 6]; [6, 8] 24 Table 6 Optimization results obtained after using the hill and valley splitting strategy No. Voltage (V) Pulse on Time (µs) Feed rate (µm/s) D in (µm) D out (µm) Taper Overcut (µm) 1 18.0 60.0 6.0 901.05 892.68 0.02 0.52 2 19.4 60.0 5.9 900.00 895.82 0.01 0.00 3 18 61.6 6.0 901.59 894.02 0.02 0.80 4 19.9 67.1 5.9 900.00 898.96 0.00 0.00 5 18.0 60.0 6.2 900.43 891.84 0.02 0.21 6 19.6 59.6 6.1 900.00 895.94 0.01 0.00 7 18.0 62.6 6.3 901.38 893.94 0.01 0.69 8 20.0 68.6 6.1 900.00 899.24 0.00 0.00 (2) Cluster centers splitting strategy The k value in the k-means clustering algorithm is a user dependent parameter; an inappropriate choice of k may yield poor results. However, so far there is no clear guideline for choosing the value of k. In this case study, the value of k was varied from 2 to 6; the corresponding splitting and optimization results are shown in Table 7. The maximum number of optimal solutions was obtained when the value of k was 6. These results are shown in Table 8.Table 7 Splitting result obtained using cluster centers strategy Value of k Cluster centers Sub-range of voltage (V) Sub-range of pulse on time (µs) Sub-range of feed rate (µm/s) Number of sub-spaces 2 [10 V,60 µs,6 µm/s]; [18 V,25 µs,6 µm/s] [8,10]; [10,18]; [18, 20] [25, 60]; [60,70] [4, 6]; [6, 8] 12 3 [10 V,55 µs,6 µm/s]; [10 V,70 µs,6 µm/s]; [18 V,25 µs,6 µm/s] [8,10]; [10,18]; [18,20] [25,55]; [55,70] [4, 6]; [6, 8] 12 4 [10 V,60 µs,6 µm/s]; [10 V,70 µs,6 µm/s]; [18 V,25 µs,6 µm/s]; [10 V,50 µs,6 µm/s] [8,10]; [10, 18]; [18,20] [25,50]; [50,60]; [60,70] [4,6]; [6,8] 18 5 [9 V,50 µs,5 µm/s]; [10 V,70 µs,6 µm/s]; [18 V,25 µs,6 µm/s]; [10 V,60 µs,6 µm/s]; [10 V,50 µs,8 µm/s] [8, 9]; [9, 10]; [10, 18]; [18, 20] [25,50]; [50,60]; [60,70] [4,5]; [5,6]; [6, 8] 36 6 [9 V,50 µs,5 µm/s]; [10 V,50 µs,8 µm/s]; [10 V,70 µs,7 µm/s]; [9 V,70 µs,5 µm/s]; [10 V,60 µs,6 µm/s]; [18 V,25 µs,6 µm/s] [8, 9]; [9, 10]; [10, 18]; [18, 20] [25,50]; [50,60]; [60,70] [4, 5]; [5, 6]; [6, 7]; [7, 8] 48 Table 8 The optimization results obtained using the cluster centers splitting strategy with k = 6 No. Voltage (V) Pulse on time (µs) Feed rate (µm/s) D in (µm) D out (µm) Taper Overcut (µm) 1 18.0 59.7 6.0 901.01 892.49 0.02 0.50 2 19.2 59.1 5.9 900.00 894.85 0.01 0.00 3 18.0 61.6 6.0 901.58 893.98 0.02 0.79 4 19.9 67.6 6.0 900.00 899.03 0.00 0.00 5 18.0 60.0 6.2 900.44 891.85 0.02 0.22 6 19.8 60.0 6.2 900.00 895.87 0.01 0.00 7 18.0 61.7 6.2 901.00 893.27 0.02 0.50 8 20.0 68.1 6.0 900.00 899.19 0.00 0.00 9 18.0 70.0 7.5 900.00 890.99 0.02 0.00 10 18.9 70.0 7.0 899.97 895.46 0.01 −0.01 (3) Equally splitting strategy The results obtained using the two previous splitting strategies were compared to the results obtained when the original search space was split equally into smaller sub-search spaces. In the equally splitting strategy, each process parameter was equally split into 4 sub-ranges, as shown in Table 9. The optimization results obtained using the equally splitting strategy are shown in Table 10.Table 9 Splitting result of equally splitting strategy Sub-range of voltage (V) Sub-range of pulse on time (µs) Sub-range of feed rate (µm/s) Number of sub-spaces [8, 11]; [11, 14]; [14, 17]; [17, 20] [25,36]; [36,47]; [47,58]; [58,70] [4, 5]; [5, 6]; [6, 7]; [7, 8] 64 Table 10 The optimization result obtained using the equally splitting strategy No. Voltage (V) Pulse on time (µs) Feed rate (µm/s) D in (µm) D out (µm) Taper Overcut (µm) 1 19.1 58.8 6.0 900.01 894.33 0.01 0.01 2 20.0 65.8 5.8 900.00 898.78 0.00 0.00 3 19.2 58.7 6.1 900.00 894.42 0.01 0.00 4 20.0 69.4 6.3 900.00 899.29 0.00 0.00 5 18.8 70.0 7.0 899.98 895.45 0.01 −0.01 Comparison and analysis These three splitting strategies provide different ways to split the search space into smaller sub-search spaces. To evaluate the efficiency of a strategy, the percentage of useful sub-search spaces was calculated using Eq. (7). 7 percentageofusefulsub-searchspaces=No.ofoptimalsolutionsNo.ofsub-searchspaces×100% Table 11 shows the comparison between the 3 strategies based on Eq. (7).Table 11 Comparison of three splitting strategies Name of splitting strategy Hill and valley splitting Cluster centers splitting Equally splitting Value of parameter – k = 2 k = 3 k = 4 k = 5 k = 6 4 No. of sub-spaces 24 12 12 18 36 48 64 No. of solutions 8 8 4 8 8 10 5 Percentage of useful sub-spaces 33.3 % 66.7 % 33.3 % 44.4 % 22.2 % 20.8 % 7.8 % It can be observed that the equally splitting strategy is the least efficient way because its percentage of useful sub-search spaces is the lowest (7.8 %). The efficiency of hill and valley splitting is fixed because it lacks any controllable parameters and because the sequence in which points are selected can affect their classification. It can be seen from Table 10 that there is a correlation between the efficiency of cluster centers splitting and the value of k. However, there is no clear understanding between the value of k and the efficiency of the method and there is also no guideline for selecting the optimal value of k. This case study utilized a trained NN prediction model in the evaluation of input parameter combinations. Therefore, to validate the optimization result, one additional experiment with the randomly chosen optimized input parameter combination was done. The data of the validation experiment is shown in Table 12.Table 12 Result of an additional validation experiment Voltage (V) Pulse on time (µs) Feed rate (µm/s) D in (µm) D out (µm) Taper Overcut (µm) Predicted 19.2 59.1 6 900.00 894.85 0.01 0.00 Experimental 899 894 0.01 0.00 Based on the validation experimental result, it can be seen that the prediction error of the NN prediction model used in this case study is quite low and the results obtained using the proposed approach are better than the results shown in the initial experimental data. Noteworthy, the optimized input process parameter combination was not in the initial training dataset and the optimization algorithm was able to find a better-than-ever objective value. Therefore, the optimization result is verified. Conclusion In this paper, a split-optimization approach was proposed for obtaining multiple solutions for a single-objective process parameter optimization problem. The proposed approach consisted of two stages: splitting of the original search space into smaller sub-search spaces and optimization of process parameters in each of the smaller sub-search spaces. Two splitting strategies, i.e., hill and valley splitting strategy and cluster centers splitting strategy, were used to split the original search space into smaller sub-search spaces efficiently. Next, GA was used in each sub-search space to find multiple combinations of process parameters that minimized the single-objective value, one from each sub-search space. The efficiency of these two strategies was verified by comparing them with a method in which the original search space is divided into smaller and equal sub-search spaces. The comparison of the results from the different splitting methods showed that the hill and valley splitting strategy and cluster centers splitting strategy were more efficient than the equal splitting strategy. Among all the methods, the cluster centers splitting strategy, for a k value of 6, was able to achieve the most optimal solutions. The results obtained from the hill and valley splitting strategy showed that though it is an efficient method, its efficiency depends on the order in which the points are classified as a hill or valley. Possible future work includes a study of the relationship between the efficiency of the cluster centers splitting strategy and the k value; a guideline should be to choose an optimal value of k. Future works also include experimentally validating the multiple solutions obtained using the proposed approach, applying the proposed approach to more case studies, and refining the proposed approach based on the results of the experimental validation and other case studies. Authors’ contributions MR and ZP—analysis of data, development of the required code, and writing of the manuscript. YGY, ZWF, HYC, and WCW—study, collection and analysis of data. BL—comments for the paper. SYL—guideline for the proposed approach and comments for the paper. All authors read and approved the final manuscript. Acknowledgements We would like to acknowledge the Metal Industries Research & Development Center for collecting the data and providing us background knowledge regarding EMM. Competing interests The authors declare that they have no competing interests. ==== Refs References Amarnath MV Pratihar DK Forward and reverse mappings of the tungsten inert gas welding process using radial basis function neural networks Proc Inst Mech Eng Part B: J Eng Manuf 2009 223 12 1575 1590 10.1243/09544054JEM1278 Chandrashekarappa MPG Krishna P Parappagoudar MB Forward and reverse process models for the squeeze casting process using neural network based approaches Appl Comput Intell Soft Comput 2014 2014 12 Chen WC Fu GL Tai PH Deng WJ Process parameter optimization for MIMO plastic injection molding via soft computing Expert Syst Appl 2009 36 2 1114 1122 10.1016/j.eswa.2007.10.020 Cook DF Ragsdale CT Major RL Combining a neural network with a genetic algorithm for process parameter optimization Eng Appl Artif Intell 2000 13 4 391 396 10.1016/S0952-1976(00)00021-X Cus F Milfelner M Balic J An intelligent system for monitoring and optimization of ball-end milling process J Mater Process Technol 2006 175 1 90 97 10.1016/j.jmatprotec.2005.04.041 Fard RK Afza RA Teimouri R Experimental investigation, intelligent modeling and multi-characteristics optimization of dry WEDM process of Al–SiC metal matrix composite J Manuf Process 2013 15 4 483 494 10.1016/j.jmapro.2013.09.002 Katherasan D Elias JV Sathiya P Haq AN Simulation and parameter optimization of flux cored arc welding using artificial neural network and particle swarm optimization algorithm J Intell Manuf 2014 25 1 67 76 10.1007/s10845-012-0675-0 Kittur JK Parappagoudar MB Forward and reverse mappings in die casting process by neural network-based approaches J Manuf Sci Prod 2012 12 1 65 80 Lee SW, Nam SH, Choi HJ, Kang EG, Ryu KY (2007) Optimized machining condition selection for high-quality surface in high-speed finish milling of molds. In: Key Engineering Materials vol 329, pp 711–718 Maji K Pratihar DK Forward and reverse mappings of electrical discharge machining process using adaptive network-based fuzzy inference system Expert Syst Appl 2010 37 12 8566 8574 10.1016/j.eswa.2010.05.019 Malakooti B Raman V An interactive multi-objective artificial neural network approach for machine setup optimization J Intell Manuf 2000 11 1 41 50 10.1023/A:1008999907768 Panda MC Yadava V Intelligent modeling and multiobjective optimization of die sinking electrochemical spark machining process Mater Manuf Process 2012 27 1 10 25 10.1080/10426914.2010.544812 Parappagoudar MB Pratihar DK Datta GL Forward and reverse mappings in green sand mould system using neural networks Appl Soft Comput 2008 8 1 239 260 10.1016/j.asoc.2007.01.005 Parappagoudar MB Pratihar DK Datta GL Neural network-based approaches for forward and reverse mappings of sodium silicate-bonded, carbon dioxide gas hardened moulding sand system Mater Manuf Processes 2008 24 1 59 67 10.1080/10426910802543681 Pawar PJ Rao RV Parameter optimization of machining processes using teaching–learning-based optimization algorithm Int J Adv Manuf Techn 2013 67 5–8 995 1006 10.1007/s00170-012-4524-2 Raja SB Baskar N Application of particle swarm optimization technique for achieving desired milled surface roughness in minimum machining time Expert Syst Appl 2012 39 5 5982 5989 10.1016/j.eswa.2011.11.110 Raja SB Pramod CS Krishna KV Ragunathan A Vinesh S Optimization of electrical discharge machining parameters on hardened die steel using Firefly Algorithm Eng Comput 2015 31 1 1 9 10.1007/s00366-013-0320-3 Rao RV Pawar PJ Modelling and optimization of process parameters of wire electrical discharge machining Proc Inst Mech Eng Part B: J Eng Manuf 2009 223 11 1431 1440 10.1243/09544054JEM1559 Senthilkumaar JS Selvarani P Arunachalam RM Intelligent optimization and selection of machining parameters in finish turning and facing of Inconel 718 Int J Adv Manuf Technol 2012 58 9–12 885 894 10.1007/s00170-011-3455-7 Specht DF A general regression neural network IEEE Trans Neural Netw 1991 2 6 568 576 10.1109/72.97934 18282872 Teimouri R Baseri H Forward and backward predictions of the friction stir welding parameters using fuzzy-artificial bee colony-imperialist competitive algorithm systems J Intell Manuf 2015 26 2 307 319 10.1007/s10845-013-0784-4 Teixidor D Ferrer I Ciurana J Özel T Optimization of process parameters for pulsed laser milling of micro-channels on AISI H13 tool steel Robotics Comput Integrated Manuf 2013 29 1 209 218 10.1016/j.rcim.2012.05.005 Yildiz AR Optimization of cutting parameters in multi-pass turning using artificial bee colony-based approach Inf Sci 2013 220 399 407 10.1016/j.ins.2012.07.012 Yusup N Sarkheyli A Zain AM Hashim SZM Ithnin N Estimation of optimal machining control parameters using artificial bee colony J Intell Manuf 2014 25 6 1463 1472 10.1007/s10845-013-0753-y

PMC005xxxxxx/PMC5001963.txt

==== Front Ann Intensive CareAnn Intensive CareAnnals of Intensive Care2110-5820Springer Paris Paris 18210.1186/s13613-016-0182-0ResearchUsefulness of video-laryngoscopy with the Airway Scope for intubation performance and learning: an experimental manikin controlled study Declercq Pierre-Louis pdeclercq@ch-dieppe.fr 1Bubenheim Michael michael.bubenheim@chu-rouen.fr 2Gelinotte Stéphanie sgelinotte@ch-dieppe.fr 1Guernon Kévin kevinguernon@hotmail.com 1Michot Jean-Baptiste jean-baptiste.michot@chi-elbeuf-louviers.fr 1Royon Vincent vincentroyon@hotmail.com 1Carpentier Dorothée dorothee.carpentier@chu-rouen.fr 1Béduneau Gaëtan gaetan.beduneau@chu-rouen.fr 13Tamion Fabienne fabienne.tamion@chu-rouen.fr 1Girault Christophe (33) 2-32-88-82-61Christophe.Girault@chu-rouen.fr 1341 Department of Medical Intensive Care, Rouen University Hospital, Rouen Cedex, France 2 Department of Clinical Research Support, Biostatistics Unit, Rouen University Hospital, Rouen Cedex, France 3 UPRES EA 3830-IRIB, Institute for Biomedical Research, Rouen University, Rouen Cedex, France 4 Service de Réanimation Médicale, Hôpital Charles Nicolle, Centre Hospitalier Universitaire-Hôpitaux de Rouen, 1, Rue de Germont, 76031 Rouen Cedex, France 26 8 2016 26 8 2016 2016 6 1 8327 2 2016 10 8 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Background Different video-laryngoscopes (VDLs) for endotracheal intubation (ETI) have recently been developed. We compared the performance of the VDL Airway Scope (AWS) with the direct laryngoscopy by Macintosh (DLM) for ETI success, time and learning. Methods We performed an experimental manikin controlled study. Twenty experienced (experts) and 40 inexperienced operators (novices) for DLM-ETI were enrolled. None of them had experience with the use of AWS-VDL. Novices were assigned to start learning with DLM or AWS, and two sub-groups of 20 novices were formed. Experts group constituted the control group. Each participant performed 10 ETI attempts with each device on the same standard manikin. The primary endpoint was the ETI success probability. Secondary endpoints were ETI time, technical validity and qualitative evaluation for each technique. We also assessed the learning order and the successive attempts effects for these parameters. Results Overall, 1200 ETI attempts were performed. ETI success probability was higher with the AWS than with the DLM for all operators (98 vs. 81 %; p < 0.0001) and for experts compared to novices using devices in the same order (97 vs. 83 %; p = 0.0002). Overall ETI time was shorter with the AWS than with the DLM (13 vs. 20 s; p < 0.0001) and for experts compared to novices using devices in the same order (11 vs. 21 s; p < 0.0001). Among novices, those starting learning with AWS had higher ETI success probability (89 vs. 83 %; p = 0.03) and shorter ETI time (18 vs. 21 s; p = 0.02). Technical validity was found better with the AWS than DLM for all operators. Novices expressed global satisfaction and device preference for the AWS, whereas experts were indifferent. Conclusions AWS-VDL permits faster, easier and more reliable ETI compared to the DLM whatever the previous airway ETI experience and could be a useful device for DLM-ETI learning. Keywords IntubationVideo-laryngoscopyAirway ScopeMacintoshSimulationLearningissue-copyright-statement© The Author(s) 2016 ==== Body Background Endotracheal intubation (ETI) is a routine life-saving procedure for airway management, and direct laryngoscopy with the Macintosh laryngoscope (DLM) is probably the most common way to perform ETI. Nevertheless, ETI using DLM (DLM-ETI) is a high-risk procedure, which can lead to vital complications, particularly in intensive care unit (ICU) patients [1, 2]. These risks are related to the critical setting as well as the underlying disease (hypoxemia, hemodynamic instability) and technical conditions of ETI [3]. Furthermore, learning and regular training are needed to sufficiently acquire skills and experience with the DLM technique [4–7]. In the past few years, new devices called video-laryngoscopes (VDLs) have been developed for improving airway management. With optical or video technique, most of these instruments allow to indirectly visualize the larynx and control the endotracheal tube (ETT) passing throughout the glottis [8]. To date, these different VDLs have been mainly used in the operating and emergency room or prehospital setting [7, 8]. Although VDL has been more recently applied in ICU setting [9–16], some authors have shown that ETI with VDL (VDL-ETI) could be associated with longer ETI time, lower oxygen saturation and even higher mortality as compared to conventional DLM-ETI [9, 13]. In fact, due to their technical conception, VDL may have their respective advantages and pitfalls, and their place may be still controversial in the ICU where airway management appears more hazardous and insufficiently studied [17]. Therefore, there is a need to rigorously assess each VDL device on an experimental and clinical basis in order to optimize ETI success. The Airway Scope (AWS®, Pentax Corp., Tokyo, Japan) is a new VDL device first reported in 2006 (Fig. 1) [18]. We hypothesized that ETI success rate, time and learning to perform a reliable ETI could be optimized by the use of the AWS as compared to the DLM, in experienced and inexperienced operators.Fig. 1 Airway Scope® (AWS; Pentax Corp., Tokyo, Japan) device. a The AWS is a portable and battery-operated video-laryngoscope (VDL) with an integrated and wide-viewing-angle (180°) liquid crystal display (LCD) monitor (6.1 cm) providing an indirect laryngoscopy of the airway via a charged coupled device attached to the blade tip of the AWS. The single-use intlock blade has a specific tube guide to accept the ETT (internal diameter between 6.5 and 8 mm). AWS-ETI requires to load and lift the epiglottis with the AWS blade tip. b Once the target signal on the LCD monitor is aligned with the glottis opening, the ETT is passed through the vocal cords (a). Then, the AWS is removed laterally, leaving the ETT in place Methods We conducted an experimental manikin controlled study to compare the AWS and DLM performance for ETI success and time, as well as ETI learning between experienced and inexperienced operators. All operators agreed to participate in this manikin study, and as it was a teaching study, which enrolled no patients and did not evaluate physiopathological process, our local ethics committee (Comité de Protection des Personnes du Nord Ouest I) stated that no ethical approval and no consent were required (conclusion date February 20, 2015, chairperson address: Pr F. Bauer, Hôpital Charles Nicole, Centre Hospitalier-Universitaire de Rouen, 1 rue de Germont 76031, France). No company sponsored the study nor supplied any device used. Experienced operators, named “experts,” were 20 senior ICU physicians (n = 9) or anesthesia residents (n = 11) who had previously performed more than 100 DLM-ETI but with no prior experience with AWS-ETI. Inexperienced operators, named “novices,” were 40 medical students with no prior experience with either of the two ETI techniques. Male and female participants were equally distributed for each novice sub-group (10 vs. 10), except for the expert sub-group (13 vs. 7, respectively). ETI was performed on a simple manikin head (Cparlene®, Adult Airway Larry manikin, CPR Savers & First Aid Supply, Scottsdale, USA) without any difficult ETI criteria, either with the DLM (blade size no. 3) or with the AWS. The AWS is a portable and battery-operated VDL with an integrated and wide-viewing-angle (180°) liquid crystal display (LCD) monitor (6.1 cm) providing an indirect laryngoscopy of the airway via a charged coupled device attached to the blade tip of the AWS. The single-use intlock blade has a specific tube guide to accept the ETT (Fig. 1a). Concerning AWS-ETI, in contrast to DLM-ETI, the ETT must be preloaded on the tube guide and ETI requires to load and lift the epiglottis with the blade tip. Once the target signal on the LCD monitor is aligned with the glottis opening, the ETT is passed through the vocal cords (Fig. 1b). Then, the AWS is removed laterally, leaving the ETT in place. For AWS-ETI, the ETT was already preload in the blade-side channel before starting the ETI attempt. All ETI were performed with no introducer and a lubricated low-pressure cuffed ETT (Portex®, Smiths Medical, St Paul, USA) with an internal diameter of 7.5 mm. The manikin head was maintained in a neutral position, i.e., not in improved Jackson’s position. The good ETT position was confirmed by lung inflation with a manual self-inflating bag (AMBU® Mark IV, Ambu Corp., Ballerup, Denmark). All inexperienced students underwent 30-min theoretical and practical training, including an ETI demonstration with each device by one experienced investigator in both techniques. During this training, students had only one attempt on the manikin with each device. The same program was given to all experts, for the AWS alone. Then, each participant performed 10 consecutive ETI attempts on the manikin with both devices. The novice operators were assigned to two different study sub-groups according to the device order of the 10 attempts, whereas the experts group performed all attempts in the same order. Therefore, three sub-groups of 20 operators were organized according to their ETI experience and device order, i.e., expert (DLM then AWS), novice 1 (DLM then AWS) and novice 2 (AWS then DLM) sub-groups. The primary endpoint of the study was the ETI success probability with each technique. ETI success was defined as an ETI performed within 60 s, with ETT in place in the trachea. Consequently, ETI failure could have been either due to a delayed ETI (>60 s) or due to esophageal intubation. Several secondary endpoints were assessed. The ETI time was defined as the time taken from the blade (DLM or AWS) first passing the incisors until ETT passage through the vocal cords. We also reported the time between the ETT passage through the vocal cords and lung inflation. The ETI technical validity for each attempt with both devices was scored and visually assessed by collecting the following adverse events: for the DLM: dental trauma based on the excessive pressure exerted by the blade on the upper incisors, epiglottis loading and default of epiglottis traction due to a lack of handle traction up and forward; for the AWS: ETT dislodgment (mobilization outside the blade channel or ablation when removing the AWS laterally), epiglottis not loaded and epiglottis luxation as well as blade malposition in the glottis. The number of esophageal intubations for both devices and the Cormack and Lehane grade for the DLM were also recorded [19]. After performing 10 ETI attempts with each device, all participants were asked to rate the following qualitative evaluation criteria using a 5-point Likert scale: ease of assembly of the device, difficulty with ETT manipulation, force of traction required for ETI, as well as the global satisfaction with the technique. It was also asked for their preference regarding the techniques for everyday use. Statistical analysis Given the dependence between successive attempts of the same volunteer, a mixed-effect logistic regression model was used [20]. As operator’s selection was based on a block design, sex was treated as a random block effect. Because observed times appeared skewed, a mixed-effect regression model with auto-correlated errors was used with time being lognormal. Point and interval estimates were then retransformed using the exponential function to present the results in conventional units. Since no failure was observed in all operator groups at several occasions, the p value for differences between attempts using AWS was derived by comparing the results of each attempt with respect to the first one, the one with the lowest number of successes for this device, and then correcting for multiple testing using Holm’s procedure. In case of undefined logits, the sign test was carried out for local testing and simple logistic regression was used in all other cases. As regards technical events, the number per operator was described using median, first and third quartiles [Q1–Q3]. In order to assess whether operator groups differed with respect to the occurrence of these events, Freeman–Halton’s test was used when results were heavily tied, i.e., events happened to less than 8 operators, otherwise the Kruskal–Wallis test was taken. In order to come to know whether operators assessed both devices differently, the sign test was used. A p value less than 0.05 was considered statistically significant. Results Sixty participants performed 20 ETI attempts each, yielding 400 ETI attempts for each sub-group, i.e., a total of 1200 ETI attempts were assessed. Considering all ETI attempts, we observed 111 failures with the DLM and 14 with the AWS (Table 1). In univariable analysis, the overall ETI success probability was significantly higher for the AWS compared to the DLM (98 vs. 81 %; p < 0.0001), for experts compared to novices 1 (97 vs. 83 %; p = 0.0002) and for novices 2 compared to novices 1 (89 vs. 83 %; p = 0.03). No significant learning effect of successive attempts was observed for ETI success. These findings were confirmed by multivariable analysis with independent factors (Table 1). As regards univariable analysis, ETI time was significantly shorter with AWS compared to DLM for all operators (13 vs. 20 s; p < 0.0001), for experts compared to novices 1 (11 vs. 21 s; p < 0.0001) and for novices 2 compared to novices 1 (18 vs. 21 s; p = 0.02) (Table 2). When compared to the first attempt, ETI time was significantly reduced at the third attempt (21 vs. 17 s; p = 0.003) and for all further attempts (Table 2). Multivariable analysis with independent factors led to the same conclusions (Table 2).Table 1 Factors potentially influencing endotracheal intubation success probability All attempts (n = 1200) Univariable analysis* Multivariable analysis*** Success** probability 95 % CI limits Success probability for novices 1 95 % CI limits Number of attempts, n Failure attempts, n (%) Lower Upper Odds ratio p value Odds ratio Lower Upper p value ETI device 1200 125 (10.4)  DLM 600 111 (18.5) 0.81 0.67 0.90 1.00 <0.0001 0.69 – 0.005 1.00 <0.0001  AWS 600 14 (2.3) 0.98 0.94 0.99 9.57 0.96 – 0.09 1.00 Operator sub-group (device order)  Experts (DLM then AWS) 400 10 (2.5) 0.97 0.94 0.99 8.04 0.0002 – 9.06 4.17 19.71 <0.0001  Novices 1 (DLM then AWS) 400 70 (17.5) 0.83 0.72 0.90 1.00 – – 1.00 – – –  Novices 2 (AWS then DLM) 400 45 (11.3) 0.89 0.81 0.94 1.68 0.03 – 1.80 1.12 2.89 0.02 Attempt number in the series of 10 attempts  1 120 17 (14.2) 0.87 0.73 0.94 1.00 Reference – – – – –  2 120 18 (15) 0.85 0.71 0.93 0.89 0.70 – – – – –  3 120 12 (10) 0.90 0.79 0.96 1.41 0.39 – – – – –  4 120 17 (14.2) 0.86 0.72 0.93 0.95 0.88 – – – – –  5 120 8 (6.7) 0.93 0.84 0.97 2.21 0.08 – – – – –  6 120 13 (10.8) 0.89 0.77 0.95 1.29 0.52 – – – – –  7 120 11 (9.2) 0.91 0.80 0.96 1.56 0.28 – – – – –  8 120 8 (6.7) 0.93 0.84 0.97 2.21 0.08 – – – – –  9 120 10 (8.3) 0.92 0.82 0.97 1.73 0.19 – – – – –  10 120 11 (9.2) 0.90 0.79 0.96 1.42 0.35 – – – – – ETI endotracheal intubation, DLM direct laryngoscopy with Macintosh, AWS video-laryngoscopy with Airway Scope® * Univariable analysis, all of the 1200 attempts have been analyzed separately according to 3 variables: ETI device (DLM vs. AWS), operator sub-group (experts vs. novices 1 and novices 1 vs. novices 2) and the attempt number in the series of 10 attempts ** Probability of a single successful attempt for one operator *** Novices 1 is the reference sub-group of the multivariable analysis, i.e., the simultaneous analysis of 2 variables: ETI device (DLM vs. AWS) and operator sub-group (experts vs. novices 1 and novices 1 vs. novices 2). For example, the probability of a successful attempt is 9.06 higher for an expert than a novice 1 Table 2 Factors potentially influencing endotracheal intubation time All attempts (n = 1200) Univariable analysis* Multivariable analysis** ETI Time (s) 95 % CI limits p value ETI time at first attempt for novices 1 (s) Time multiplier*** 95 % CI limits p value Number of attempts, n Sample arithmetic mean time (s) Lower Upper Lower Upper ETI device 1200 19.63  DLM 600 24.52 20.44 16.18 25.83 <0.0001 35.65 – 29.61 42.93 <0.0001  AWS 600 14.73 12.72 10.08 16.06 22.51 – 18.75 27.02 Operator sub-group (device order)  Experts (DLM then AWS) 400 12.53 11.01 9.28 13.07 <0.0001 – 0.54 0.49 0.58 <0.0001  Novices 1 (DLM then AWS) 400 25.35 20.51 17.29 24.32 – 1.00 – – –  Novices 2 (AWS then DLM) 400 21.00 18.37 15.49 21.79 0.02 – 0.83 0.77 0.90 <0.0001 Attempt number in the series of 10 attempts  1 120 24.92 20.99 16.42 26.84 Reference – 1.00 – – –  2 120 23.63 19.44 15.16 24.92 0.16 – 0.91 0.83 1.01 0.08  3 120 20.62 17.09 13.33 21.93 0.003 – 0.80 0.71 0.89 0.0001  4 120 19.56 15.92 12.41 20.42 <0.0001 – 0.74 0.66 0.84 <0.0001  5 120 18.76 15.22 11.87 19.51 <0.0001 – 0.71 0.63 0.80 <0.0001  6 120 18.22 14.87 11.59 19.09 <0.0001 – 0.69 0.62 0.77 <0.0001  7 120 18.65 15.34 11.94 19.72 <0.0001 – 0.71 0.63 0.80 <0.0001  8 120 17.81 14.59 11.35 18.75 <0.0001 – 0.68 0.60 0.76 <0.0001  9 120 17.03 14.06 10.96 18.03 <0.0001 – 0.65 0.58 0.73 <0.0001  10 120 17.06 13.69 10.61 17.66 <0.0001 – 0.63 0.57 0.70 <0.0001 ETI endotracheal intubation, DLM direct laryngoscopy with Macintosh, AWS video-laryngoscopy with Airway Scope® * Univariable analysis, all the 1200 attempts have been analyzed separately according to 3 variables: ETI device (DLM vs. AWS), operator sub-group (experts vs. novices 1 and novices 1 vs. novices 2) and the attempt number in the series of 10 attempts ** Novices 1 at the first attempt serve as reference situation of the multivariable analysis, i.e., the simultaneous analysis of 3 variables: ETI device (DLM vs. AWS), operator sub-group (experts vs. novices 1 and novices 1 vs. novices 2) and the attempt number in the series of 10 attempts *** Multiplicative factor allowing to estimate ETI time from the novices 1 ETI time for the others situations. For example, ETI time for an expert using AWS at first attempt is estimated to be 12.16 s, i.e., 22.51 × 0.54 × 1 All experts succeeded their 10 AWS-ETI attempts, whereas 9 novices failed at least once (3 and 6 novices in sub-groups 1 and 2, respectively). Fourteen experts (70 %) succeeded all DLM-ETI attempts in contrast to 3 (15 %) and 7 (35 %) novices in sub-groups 1 and 2, respectively (data not shown). Empirical ETI learning curves are shown separately for AWS (Figs. 2, 3) and DLM (Figs. 4, 5). The AWS-ETI success proportion curves for novices are close to that of experts (Fig. 2). Looking at AWS alone and noting that no failure was observed for several attempts, no evidence was found for an effect of experience, learning order and successive attempts (Fig. 2). ETI time for AWS alone was significantly shorter for experts when compared to novices 1 (9 vs. 16 s; p < 0.0001), when learning order began with AWS instead of DLM (13 vs. 16 s; p = 0.0001) and with increasing number of attempts (10 and 19 s, at first and tenth attempts, respectively; p < 0.0001) (Fig. 3). Learning curves were different for each sub-group with DLM (Figs. 4, 5). When considering DLM alone, DLM-ETI success proportion was significantly higher for experienced operators as compared to novices 1 (95 vs. 82 %; p = 0.0002), when learning started with AWS as compared to DLM (82 vs. 68 %; p = 0.002), but no significant effect of successive attempt was found (Fig. 4). ETI time for DLM alone changed significantly between the first and the tenth attempts (24 and 18 s, respectively; p = 0.048) and was shorter for experts (13 vs. 28 s; p < 0.0001) and for novices 2 (23 vs. 28 s; p = 0.0002) as compared to novices 1 (Fig. 5).Fig. 2 AWS-ETI learning curves according to ETI success rate by attempt and operator sub-group. ETI endotracheal intubation, DLM direct laryngoscopy with Macintosh, AWS video-laryngoscopy with Airway Scope®. Success proportion is the observed number of successes among 100 attempts for the 3 study sub-groups (experts, novices 1 and novices 2) according to the attempt number in a series of ten attempts and the device used (DLM or AWS). p values refer to homogeneity tests for experience (experts compared to novices 1), device learning order (novices 1 compared to novices 2), and to the global test for any difference between attempts, which evaluate the learning effect of successive attempts Fig. 3 AWS-ETI learning curves according to ETI time by attempt and operator sub-group. ETI endotracheal intubation, DLM direct laryngoscopy with Macintosh, AWS video-laryngoscopy with Airway Scope®. Mean time to vocal cords corresponds to ETI time (time taken from the blade (DLM or AWS) first passing the incisors until ETT passage through the vocal cords) in seconds according to the attempt number in a series of ten attempts and the device used (DLM or AWS). p values refer to homogeneity tests for experience (experts compared to novices 1), device learning order (novices 1 compared to novices 2), and to the global test for any difference between attempts, which evaluate the learning effect of successive attempts Fig. 4 DLM-ETI learning curves according to ETI success rate by attempt and operator sub-group. ETI endotracheal intubation, DLM direct laryngoscopy with Macintosh, AWS video-laryngoscopy with Airway Scope®. Success proportion is the observed number of successes among 100 attempts for the 3 study sub-groups (experts, novices 1 and novices 2) according to the attempt number in a series of ten attempts and the device used (DLM or AWS). p values refer to homogeneity tests for experience (experts compared to novices 1), device learning order (novices 1 compared to novices 2), and to the global test for any difference between attempts, which evaluate the learning effect of successive attempts Fig. 5 DLM-ETI learning curves according to ETI time by attempt and operator sub-group. ETI endotracheal intubation; DLM direct laryngoscopy with Macintosh, AWS video-laryngoscopy with Airway Scope®. Mean time to vocal cords corresponds to ETI time (time taken from the blade (DLM or AWS) first passing the incisors until ETT passage through the vocal cords) in seconds according to the attempt number in a series of ten attempts and the device used (DLM or AWS). p values refer to homogeneity tests for experience (experts compared to novices 1), device learning order (novices 1 compared to novices 2), and to the global test for any difference between attempts, which evaluate the learning effect of successive attempts In univariable analysis, the time between the ETT passage through the vocal cords and lung inflation was found significantly shorter for all operators when the DLM was used as compared to AWS (9 vs. 11 s; p < 0.0001), when ETI was performed by experts (7 vs. 13 s; p < 0.0001) or by novices 2 (12 vs. 13 s; p = 0.005) as compared to novices 1, and with increasing number of attempts (12 and 9 s, at first and tenth attempts, respectively; p < 0.001). All these findings were confirmed by multivariable analysis (data not shown). ETI technical validity and adverse events are reported in Tables 3 and 4. The DLM-ETI technical validity appeared higher for experts than for both novice sub-groups (p = 0.0005; Table 3). All adverse events occurred more frequently in these two sub-groups, and dental pressure was the main event reported. There were few adverse events with the AWS-ETI and virtually no difference in technical validity between the three sub-groups (Table 4). The main events reported were epiglottis luxation for experts (p = 0.04) and delayed ETI time for both novice sub-groups (p = 0.04).Table 3 Technical validity and adverse events for endotracheal intubation with the DLM by operator sub-group Operator sub-group (device order) p value* Experts (DLM then AWS) n = 20 Novices 1 (DLM then AWS) n = 20 Novices 2 (AWS then DLM) n = 20 Events ** (out of 200 attempts) Median [Q1–Q3]*** Events** (out of 200 attempts) Median [Q1–Q3]*** Events ** (out of 200 attempts) Median [Q1–Q3]*** Technically valid attempts, n 146 9 [7–9] 61 2 [1–4] 84 4 [2–6] 0.0005 Adverse technical events  Esophageal intubation, n 6 0 [0–0] 34 1 [0–2] 21 0 [0–2] 0.02  Delayed ETI time (>60 s), n 4 0 [0–0] 30 1 [0–2] 16 0 [0–1] 0.003  Dental pressure, n 49 1 [0–3] 78 4 [2–5] 76 3 [2–5] 0.02  Epiglottis loading, n 1 0 [0–0] 26 1 [0–2] 25 1 [0–2] 0.0001  Lack of traction, n 3 0 [0–0] 35 1 [0–4] 15 0 [0–1] 0.002  Cormack–Lehane grade (median) – 1 [1, 2] – 2 [1, 2] – 2 [1, 2] 0.008 ETI endotracheal intubation, DLM direct laryngoscopy with Macintosh, AWS video-laryngoscopy with Airway Scope® * Kruskal–Wallis test ** Total number of each event on 200 attempts in each sub-group with DLM *** Median [1st–3rd quartile] per operator of the number of events on 10 attempts with DLM in each sub-group Table 4 Technical validity and adverse events for endotracheal intubation with the AWS by operator sub-group Operator sub-group (device order) p value Experts (DLM then AWS) n = 20 Novices 1 (DLM then AWS) n = 20 Novices 2 (AWS then DLM) n = 20 Events* (out of 200 attempts) Median [Q1–Q3]** Events* (out of 200 attempts) Median [Q1–Q3]** Events* (out of 200 attempts) Median [Q1–Q3]** Technically valid attempts, n 179 9 [8–10] 177 9 [9, 10] 184 9 [9, 10] 0.59**** Adverse technical event  Esophageal intubation, n 0 0 [0–0] 1 0 [0–0] 1 0 [0–0] 1.0***  Delayed ETI time (>60 s.), n 0 0 [0–0] 5 0 [0–0] 7 0 [0–0] 0.04***  Epiglottis luxation, n 21 1 [0–2] 13 0 [0–1] 8 0 [0–1] 0.04****  Epiglottis unloaded, n 0 0 [0–0] 2 0 [0–0] 0 0 [0–0] 0.32***  ETT dislodgment, n 0 0 [0–0] 0 0 [0–0] 3 0 [0–0] 0.1***  Blade malposition in glottis, n 0 0 [0–0] 9 0 [0–0] 5 0 [0–0] 0.43*** ETI endotracheal intubation, DLM direct laryngoscopy with Macintosh, AWS video-laryngoscopy with Airway Scope® * Total number of each event on 200 attempts in each sub-group with AWS ** Median [1st–3rd quartile] per operator of the number of events on 10 attempts with AWS in each sub-group *** Freeman–Halton’s test **** Kruskal–Wallis test Concerning qualitative assessment of ETI technique (data not shown), DLM was considered easier to assemble than the AWS by expert and novice 2 sub-groups. ETT manipulation with DLM was considered easier by experts although novices reported no difference. All sub-groups agreed that DLM-ETI required more force than did AWS-ETI. ETI global satisfaction and preference were in favor of the AWS in both novice sub-groups, whereas experts were indifferent. Discussion To our knowledge, this is the largest controlled manikin study comparing a new device, AWS-VDL, with the DLM technique for ETI. The AWS allowed faster, easier and more reliable ETI performance than the DLM whatever the previous airway ETI experience. Interestingly, novice operators who started with the AWS performed DLM-ETI more efficiently than those starting with the DLM, which suggests AWS to be a useful device for DLM-ETI learning. Furthermore, there was no effect of successive attempts on ETI success probability with both devices. This could mean that for novice operators, AWS-ETI learning on manikin is achieved after the first attempts. By contrast, and as demonstrated in previous studies [4–6], 10 ETI attempts were found insufficient for novice operators to perform DLM-ETI with reliably. Finally, our results demonstrate that AWS-ETI acquisition may require less operator skill than the DLM. Our results for AWS-ETI performance are in agreement with those of one previous manikin study comparing the AWS with DLM in 31 inexperienced nurses [21]. Two prospective randomized clinical studies reported that the AWS allowed to decrease ETI time and failure by inexperienced residents as compared to DLM [22, 23]. In routine anesthesia, the AWS was also able to significantly improve the laryngeal view being more reliable in case of unanticipated difficult ETI as compared to DLM [24]. The time between the ETT passage through the vocal cords and lung inflation was found shorter with the DLM for all operators. The device itself could explain this finding in part. Indeed, the DLM can be more easily removed than the AWS, since the operator has to take care not to accidentally remove the tube inserted in the AWS side channel (Fig. 1a) while he removes the AWS blade laterally [18]. This adverse event occurred three times in six hundred ETI attempts with AWS in our study. In addition, the time required to remove the AWS from the mouth should not delay the lung inflation. In practice, it is possible to ventilate the more hypoxemic patients manually before removing the AWS device. The time between ETI and lung inflation was also found shorter for experts compared to novices. This could be explained by the experts’ greater experience to manage cuff inflation as well as manual self-inflating ventilation. Technical validity and adverse events with both devices, closely related to ETI acquisition and performance, have been poorly reported [21, 25]. We found a better ETI technical validity and less adverse events with the AWS, although epiglottis luxation and AWS blade malposition were frequently reported for experts and novices, respectively. In a clinical study, ETT impingement onto the laryngeal structures has also been observed in 4 % of 320 ETI, but this could be easily managed by adjusting the AWS blade direction [24]. Unlike previous manikin and clinical studies [21–24], we observed two cases of esophageal intubation by novice operators with the AWS. Numerous adverse technical events, a higher Cormack and Lehane grade and delayed DLM-ETI were expected with the DLM in inexperienced sub-groups and reflected the difficulty to acquire the DLM-ETI technique [4–6]. Interestingly, the DLM-ETI failure rate was not negligible (2.5 %) for experienced operators with a significant amount of dental pressure. This could be related to manikin head rigidity, which can be more difficult to intubate than patients [21, 25]. Furthermore, dental pressure by the DLM could have been overestimated as it was subjectively evaluated and not based on the number of audible teeth clicks [21, 25] or video-recording. Indeed, a video-recording for each attempt would have been useful to improve attempts analysis and their technical validity as well as limit the potential subjective interpretation of investigators, especially for DLM-ETI. Furthermore, watching video-recordings would have allowed to operators to see and understand their potential mistakes and therefore to improve themselves. However, such video-recordings would have been very time-consuming for all the 1200 ETI attempts. The simplicity and facility of AWS use were confirmed by an ETI qualitative evaluation by all operators including global satisfaction and preference. Nevertheless, AWS assembly and ETT manipulation were considered to be slightly more difficult by expert and novice 1 sub-groups. One explanation could be that AWS assembling can be less intuitive as the blade requires to be attached and locked to the monitor by a metallic collar and clip, and the ETT must be maintained in place while the AWS blade is removed laterally [18]. It is noteworthy that all operators considered that the DLM-ETI required more force of traction than did the AWS-ETI. This finding could explain, in part, the significant frequency of excessive dental pressure reported with the DLM including expert operators, as well as the higher ETI failure rate and longer ETI time for novices. Our findings on global satisfaction for the AWS confirm those of previous studies in inexperienced operators [7, 26, 27]. They also showed that DLM-ETI experts may rapidly appropriate the AWS-ETI technique. AWS-ETI performance and easiness could probably be explained by its technical features. First, the disposable AWS blade is anatomically designed to conform to the shape of the mouth and pharynx and to pass over the tongue dorsum, with the minimal displacement of soft tissues. Second, unlike the DLM, the AWS does not require alignment of the oral, pharyngeal and tracheal axes to visualize the vocal cords [18]. Third, once inserted into the mouth, the AWS requires minimal adjustment of its position, whereas the DLM requires coordinated movements to expose the glottis [21, 24]. Fourth, imposed by the anatomic shape of the blade, epiglottis loading with the AWS blade tip allows to get the best glottis view to perform reliable ETI. By contrast, the DLM-ETI does not require epiglottis loading, but the blade tip must be placed into the glossoepiglottic fold in order to tract epiglottis and getting direct view of the glottis space [28]. This technical features, notably the need for loading the epiglottis or not, could explain in part the difference between both devices regarding ETI performance such as handling and the risk of adverse events or injuries. Therefore, the entire ETI procedure being performed under visual control with the AWS monitor, ETI could be more secured and supervised. Simulation program and manikin studies are now essential for operator’s skill acquisition of life-saving procedures like ETI in the ICU environment [29–31]. Anyway, clinical studies are obviously needed. A recent meta-analysis, involving nine clinical ICU studies, demonstrated that VDL-ETI in ICU could be useful to decrease difficult ETI, esophageal intubation, Cormack 3/4 grades, and increase first-attempt success, but did not reduce severe complications [16]. However, these results should be interpreted with caution due to the between study heterogeneity depending on the outcome analyzed, operator experience and different VDL devices used. In fact, these ICU studies primarily assessed the GlideScope (GVL®; Verathon Medical, Bothell, WA, USA) [16], a VDL with a different conception and operating mode, mainly due to a deported screen and the use of a separate preformed metal stylet inserted into the ETT. In contrast, two prospective randomized clinical cohort studies performed by inexperienced providers outside the ICU found better performances for the AWS [22, 23]. Furthermore, due to their technical conception, the best glottis view provided by different VDL may not always match ETI outcomes [9, 13]. AWS advantages result from its integrated monitor and ETT side channel into the handle and blade, respectively, allowing to facilitate the ETT insertion and to visualize its entire progression [24]. These features are useful to avoid the difficulties due to the necessary coordination between a separate monitor and/or the use of an additional stylet. In addition, other VDL may partially blind the ETT progression and result in serious complications [32]. It must be underlined that our experimental manikin study has been conducted with a non-difficult airway head, and results cannot be expanded in cases of difficult ETI. Experimental and clinical studies have previously shown, however, that AWS-ETI could be a reliable technique in different difficult airways such as pharyngeal obstruction, cervical spine rigidity, tongue edema [26, 33] and limited mouth opening (≈20 mm) [34]. AWS can also be used according to a multimodal approach combining a flexible stylet or a fiberoptic bronchoscope, particularly if nasal ETI is required [35]. Nevertheless, there are some situations at risk of AWS-ETI difficulty or failure. The blade introduction in the mouth can be difficult due to the device length. However, the AWS blade introduction first, and secondarily connect to the AWS handle, may permit to introduce the entire device in the mouth. Despite a specific blade-side channel for introducing an aspiration catheter, abundant pharyngeal secretions can also obstruct the LCD glottis view, as well as condensation on the blade extremity [24]. In rare cases, the larynx can be difficult to reach due to a shorter length of the blade [36]. Finally, the minimum mouth opening required for the AWS-VDL insertion appears to be 20–25 mm [34], and AWS usefulness in cases of sub-glottic tumors needs to be assessed, likely in association with fiberoptic bronchoscope [37]. Therefore, all these features are strong arguments to initially perform manikin studies and to conduct then prospective clinical studies with each type of VDL, in no difficult as well as difficult ETI settings, before to determine the VDL-ETI’s place in the ICU [38]. Some limitations must be underlined in our study. First, it was a manikin study that may not strictly reproduce the human clinical conditions for ETI, particularly in the ICU regarding the underlying disease and technical circumstances. However, the manikin’s airway has been recognized as an acceptable and realistic condition for ETI evaluation [29–31]. Second, as it was obviously not possible to perform a blind study, a potential bias may exist, particularly for experts and investigators. Nevertheless, this bias was probably reduced, as our main endpoints were clearly defined. Lastly, we used subjective qualitative criteria as secondary endpoints for which the evaluation may widely change. However, this was probably unlikely as we found a good agreement between these criteria and our objective endpoints, particularly for ETI success and time. Conclusions This large experimental manikin controlled study demonstrates that the AWS allowed faster, easier and more reliable ETI than did the DLM whatever the previous airway ETI experience. It also confirms that DLM-ETI needs a prolonged learning, whereas AWS-ETI requires less operator skill than does the DLM to effectively and rapidly secure the airway. Our results further suggest that the AWS may be a useful device for ETI learning with the DLM. Nevertheless, further randomized clinical studies are still warranted to determine the respective place of different VDL devices for ETI practice, particularly in the ICU setting. Abbreviations AWSAirway Scope AWS-ETIEndotracheal intubation with Airway Scope AWS-VDLAirway Scope video-laryngoscope DLMDirect laryngoscopy with Macintosh DLM-ETIEndotracheal intubation with Macintosh laryngoscope ETIEndotracheal intubation ETTEndotracheal tube ICUIntensive care unit VDLVideo-laryngoscope VDL-ETIEndotracheal intubation with video-laryngoscope Authors’ contributions PLD contributed to conception and design of the study, acquisition of data, analysis and interpretation of data, drafted the submitted article and provided final approval of the version to be published. MB contributed to analysis and interpretation of the data, performed the statistical analysis, drafted the submitted article and provided final approval of the version to be published. SG, KG, JBM, VR, DC and FT contributed to data analysis and interpretation, critical revision of the manuscript for important intellectual content and provided final approval of the version to be published. CG was at the study’s origin, contributed to study design and conception, study supervision, data analysis and interpretation and provided final approval of the version to be published. All authors read and approved the final manuscript. Competing interests All the authors declare that they have no competing interests. ==== Refs References 1. Jaber S Amraoui J Lefrant JY Arich C Cohendy R Landreau L Clinical practice and risk factors for immediate complications of endotracheal intubation in the intensive care unit: a prospective, multiple-center study Crit Care Med 2006 34 2355 2361 10.1097/01.CCM.0000233879.58720.87 16850003 2. Cook TM Woodall N Harper J Benger J Major complications of airway management in the UK: results of the Fourth National Audit Project of the Royal College of Anaesthetists and the Difficult Airway Society. Part 2: intensive care and emergency departments Br J Anaesth 2011 106 632 642 10.1093/bja/aer059 21447489 3. De Jong A Molinari N Terzi N Mongardon N Arnal J-M Guitton C Early identification of patients at risk for difficult intubation in the intensive care unit: development and validation of the MACOCHA score in a multicenter cohort study Am J Respir Crit Care Med 2013 187 832 839 10.1164/rccm.201210-1851OC 23348979 4. Konrad C Schupfer G Wietlisbach M Gerber H Learning manual skills in anesthesiology: Is there a recommended number of cases for anesthetic procedures? Anesth Analg 1998 86 635 639 10.1213/00000539-199803000-00037 9495429 5. Mulcaster JT Mills J Hung OR MacQuarrie K Law JA Pytka S Laryngoscopic intubation: learning and performance Anesthesiology 2003 98 23 27 10.1097/00000542-200301000-00007 12502974 6. Garza AG Gratton MC Coontz D Noble E Ma OJ Effect of paramedic experience on orotracheal intubation success rates J Emerg Med 2003 25 251 256 10.1016/S0736-4679(03)00198-7 14585451 7. Griesdale DE Liu D McKinney J Choi PT Glidescope_video-laryngoscopy versus direct laryngoscopy for endotracheal intubation: a systematic review and meta-analysis Can J Anesth 2012 59 41 52 10.1007/s12630-011-9620-5 22042705 8. Niforopoulou P Pantazopoulos I Demestiha T Koudouna E Xantos T Video-laryngoscopes in the adult airway management: a topical review of the literature Acta Anaesthesiol Scand 2010 54 1050 1061 10.1111/j.1399-6576.2010.02285.x 20887406 9. Griesdale DE Chau A Isac G Ayas N Foster D Irwin C Video-laryngoscopy versus direct laryngoscopy in critically ill patients: a pilot randomized trial Can J Anaesth 2012 59 1032 1039 10.1007/s12630-012-9775-8 22932944 10. Noppens RR Geimer S Eisel N David M Piepho T Endotracheal intubation using the C-MAC_video laryngoscope or the Macintosh laryngoscope: a prospective, comparative study in the ICU Crit Care 2012 16 R103 10.1186/cc11384 22695007 11. Kory P Guevarra K Mathew JP Hegde A Mayo PH The impact of video laryngoscopy use during urgent endotracheal intubation in the critically ill Anesth Analg 2013 117 144 149 10.1213/ANE.0b013e3182917f2a 23687228 12. Lakticova V Koenig SJ Narasimhan M Mayo PH Video laryngoscopy is associated with increased first pass success and decreased rate of esophageal intubations during urgent endotracheal intubation in a medical intensive care unit when compared to direct laryngoscopy J Intensive Care Med 2015 30 44 48 10.1177/0885066613492641 23771876 13. Yeatts DJ Dutton RP Hu PF Chang YW Brown CH Chen H Effect of video laryngoscopy on trauma patient survival: a randomized controlled trial J Trauma Acute Care Surg 2013 75 212 219 10.1097/TA.0b013e318293103d 23823612 14. De Jong A Clavieras N Conseil M Coisel Y Moury PH Pouzeratte Y Implementation of a combo videolaryngoscope for intubation in critically ill patients: a before-after comparative study Intensive Care Med 2013 39 2144 2152 10.1007/s00134-013-3099-1 24045887 15. Mosier JM Whitmore SP Bloom JW Snyder LS Graham LA Carr GE Video laryngoscopy improves intubation success and reduces esophageal intubations compared to direct laryngoscopy in the medical intensive care Crit Care 2013 17 R237 10.1186/cc13061 24125064 16. De Jong A Molinari N Conseil M Coisel Y Pouzeratte Y Belafia F Video laryngoscopy versus direct laryngoscopy for orotracheal intubation in the intensive care unit: a systematic review and meta-analysis Intensive Care Med 2014 40 629 639 24556912 17. Cook T Behringer EC Benger J Airway management outside the operating room: hazardous and incompletely studied Curr Opin Anaesthesiol 2012 25 461 469 10.1097/ACO.0b013e32835528b1 22673785 18. Koyama J Aoyama T Kusano Y Seguchi T Kawagishi K Iwashita T Description and first clinical application of Airway Scope for tracheal intubation J Neurosurg Anesthesiol 2006 18 247 250 10.1097/00008506-200610000-00005 17006122 19. Cormack RS Lehane J Difficult tracheal intubation in obstetrics Anaesthesia 1984 39 1105 1111 10.1111/j.1365-2044.1984.tb08932.x 6507827 20. Hedeker D Gibbons RD Longitudinal data analysis 2006 Hoboken Wiley 21. Miki T Inagawa G Kikuchi T Koyama Y Goto T Evaluation of the Airway Scope, a new video laryngoscope, in tracheal intubation by naive operators: a manikin study Acta Anaesthesiol Scand 2007 51 1378 1381 10.1111/j.1399-6576.2007.01450.x 17944642 22. Hirabayashi Y Seo N Tracheal intubation by non-anaesthetist physicians using the Airway Scope Emerg Med J 2007 24 572 573 10.1136/emj.2007.048546 17652682 23. Hirabayashi Y Seo N Tracheal intubation by non-anesthesia residents using the Pentax-AWS Airway Scope and Macintosh laryngoscope J Clin Anesth 2009 21 268 271 10.1016/j.jclinane.2008.08.024 19502032 24. Suzuki A Toyama Y Katsumi N Kunisawa T Sasaki R Hirota K The Pentax-AWS rigid indirect video laryngoscope: clinical assessment of performance in 320 cases Anaesthesia 2008 63 641 647 10.1111/j.1365-2044.2008.05452.x 18477277 25. Malik MA Hassett P Carney J Higgins BD Harte BH Laffey JG A comparison of the Glidescope, Pentax AWS, and Macintosh laryngoscopes when used by novice personnel: a manikin study Can J Anaesth 2009 56 802 811 10.1007/s12630-009-9165-z 19655213 26. Koyama Y Inagawa G Miyashita T Kikuchi T Miura N Miki T Comparison of the Airway Scope, gum elastic bougie and fiberoptic bronchoscope in simulated difficult tracheal intubation: a manikin study Anaesthesia 2007 62 936 939 10.1111/j.1365-2044.2007.05152.x 17697222 27. Lewis AR Hodzovic I Whelan J Wilkes AR Bowler I Whitfield R A paramedic study comparing the use of the Airtraq, Airway Scope and Macintosh laryngoscopes in simulated prehospital airway scenarios Anaesthesia 2010 65 1187 1193 10.1111/j.1365-2044.2010.06514.x 20958276 28. Henderson JJ Questions about the Macintosh laryngoscope and technique of laryngoscopy Eur J Anaesthesiol 2000 17 2 5 10.1097/00003643-200001000-00002 10758437 29. Hesselfeldt R Kristensen MS Rasmussen LS Evaluation of the airway of the SimMan full-scale patient simulator Acta Anaesthesiol Scand 2005 49 1339 1345 10.1111/j.1399-6576.2005.00856.x 16146473 30. Piquette D Tarshis J Regehr G Fowler RA Pinto R LeBlanc VR Effects of clinical supervision on resident learning and patient care during simulated ICU scenarios Crit Care Med 2013 41 2705 2711 10.1097/CCM.0b013e31829a6f04 23963128 31. Kennedy CC Cannon EK Warner DO Cook DA Advanced airway management simulation training in medical education: a systematic review and meta-analysis Crit Care Med 2014 42 169 178 10.1097/CCM.0b013e31829a721f 24220691 32. Cooper RM Complications associated with the use of the GlideScope videolaryngoscope Can J Anaesth 2007 54 54 57 10.1007/BF03021900 17197469 33. Kariya T Inagawa G Nakamura K Fujimoto J Aoi Y Morita S Evaluation of the Pentax-AWS(® ) and the Macintosh laryngoscope in difficult intubation: a manikin study Acta Anaesthesiol Scand 2011 55 223 227 10.1111/j.1399-6576.2010.02363.x 21226865 34. Komatsu R Kamata K Hoshi I Sessler DI Ozaki M Airway Scope and gum elastic bougie with Macintosh laryngoscope for tracheal intubation in patients with simulated restricted neck mobility Br J Anaesth 2008 101 863 869 10.1093/bja/aen288 18835885 35. Xue FS Xiong J Yuan YJ Wang Q Asai T Pentax-AWS video laryngoscope for awake nasotracheal intubation in patients with a difficult airway Br J Anaesth 2010 104 505 506 10.1093/bja/aeq042 20228184 36. Sasano N Yamauchi H Fujita Y Failure of the Airway Scope to reach the larynx Can J Anaesth 2007 54 774 775 10.1007/BF03026884 17766752 37. Gaszyński T A combination of KingVision video-laryngoscope and flexible fibroscope for awake intubation in patient with laryngeal tumor-case report and literature review Anaesthesiol Intensive Ther 2015 47 433 435 10.5603/AIT.a2015.0019 25830936 38. Apfelbaum JL Hagberg CA Caplan RA Blitt CD Connis RT Nickinovich DG Practice guidelines for management of the difficult airway: an updated report by the American Society of Anesthesiologists Task Force on management of the difficult airway Anesthesiology 2013 118 251 270 10.1097/ALN.0b013e31827773b2 23364566

PMC005xxxxxx/PMC5001964.txt

==== Front Nanoscale Res LettNanoscale Res LettNanoscale Research Letters1931-75731556-276XSpringer US New York 159610.1186/s11671-016-1596-1Nano ExpressComparative Study on Single-Molecule Junctions of Alkane- and Benzene-Based Molecules with Carboxylic Acid/Aldehyde as the Anchoring Groups Chen Fang Peng Lin-Lu Hong Ze-Wen Mao Jin-Chuan Zheng Ju-Fang Shao Yong Niu Zhen-Jiang Zhou Xiao-Shun xszhou@zjnu.edu.cn Key Laboratory of the Ministry of Education for Advanced Catalysis Materials, Institute of Physical Chemistry, Zhejiang Normal University, Jinhua, Zhejiang 321004 China 26 8 2016 26 8 2016 2016 11 1 3808 7 2016 17 8 2016 © The Author(s). 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.We have measured the alkane and benzene-based molecules with aldehyde and carboxylic acid as anchoring groups by using the electrochemical jump-to-contact scanning tunneling microscopy break junction (ECSTM-BJ) approach. The results show that molecule with benzene backbone has better peak shape and intensity than those with alkane backbone. Typically, high junction formation probability for same anchoring group (aldehyde and carboxylic acid) with benzene backbone is found, which contributes to the stronger attractive interaction between Cu and molecules with benzene backbone. The present work shows the import role of backbone in junction, which can guide the design molecule to form effective junction for studying molecular electronics. Keywords Single-molecule junctionsECSTM-BJJunction formation probabilityCarboxylic acidAldehydeZhejiang Provincial Natural Science Foundation of ChinaLR15B030002Zhou Xiao-Shun http://dx.doi.org/10.13039/501100001809National Natural Science Foundation of China21573198Zhou Xiao-Shun National Natural Science Foundation of China (CN)21273204Zhou Xiao-Shun issue-copyright-statement© The Author(s) 2016 ==== Body Background In recent years, single-molecule junctions have attracted wide attention because of its potential application in nano-electronic and molecular electronic device [1–10]. At present stage, it is important to fully understand the electron transport of single-molecule junctions and its influence factors [11]. Many factors can affect the conductance of single-molecule junctions, such as anchoring group, molecule structure, the contact configuration between molecule and electrode, and temperature [5, 7, 12–17]. Among them, anchoring group is very important in forming the molecular junction, and it was found that different anchoring groups have different junction formation probabilities [14, 18–20]. However, another interesting question is still unclear that how molecular structure would influence the junction formation probability for same anchoring group. In this work, we will focus on the junction formation probability of same anchoring group with different molecular structures (saturated and conjugated structure) by using electrochemical jump-to-contact scanning tunneling microscopy break junction (ECSTM-BJ) approach (Fig. 1a) [21, 22]. Aldehyde and carboxylic acid anchoring groups binding to Cu electrode are used in the current study, for they have been demonstrated to form effective junctions [23–25]. We use 1,4-benzenedicarboxaldehyde, glutaraldehyde, 1,4-benzenedicarboxylic acid and pentanedioic acid as target molecules (Fig. 1b) to study the influence of different structures on the junction formation probability. Those molecules have different backbones with saturated (alkane) or conjugated (benzene) structure.Fig. 1 The schematic diagram of ECSTM-BJ and molecular structure. a Schematic diagram of electrochemical jump-to-contact scanning tunneling microscopy break junction (ECSTM-BJ) approach in solution containing target molecule. b Molecular structures of glutaraldehyde, 1,4-benzenedicarboxaldehyde, pentanedioic acid and 1,4-benzenedicarboxylic acid Methods Na2SO4 (99.995 %) and CuSO4 (99.999 %) were purchased from Alfa-Aesar, while petanedioic acid, 1,4-benzenedicarboxylic acid, glutaraldehyde, and 1,4-benzenedicarboxaldehyde were purchased from Sigma-Aldrich. Ultrapure water (≥18.2 MΩ cm) was used for preparing aqueous solutions. Naturally formed Au(111) was used as the substrate, and cut Pt-Ir STM tip was covered with thermosetting glue to reduce the electrochemical current. Meantime, Pt and Cu wires were used as the counter and reference electrodes, respectively. The conductance measurement was performed by ECSTM-BJ approach on the modified Nanoscope IIIa STM (Veeco, USA); especially, preamplifier with four-out linear current-to-voltage converters was used [23]. The experiment was carried out in aqueous solution containing saturated target molecule + 1 mM CuSO4 + 50 mM Na2SO4 as following: Firstly, the tip potential was set at −5 mV to allow the bulk deposition of Cu. Secondly, after applying the pulse voltage on z-piezo, the deposited Cu on the tip would transfer to the substrate and build a metallic contact due to the tip closed to the surface. Thirdly, Cu atomic wire could be formed during the separation of tip and substrate with 20 nm/s, and then, molecule could simultaneously bridge to the both electrodes upon the breaking of metal atomic wire. The tip current vs. distance curves were recorded with sampling frequency of 20 kHz. More detailed procedure can be seen in our previously reports [22, 23]. Results and Discussion Comparative Study on Single-Molecule Conductance of Glutaraldehyde and 1,4-Benzenedicarboxaldehyde with Cu Electrode We firstly measure the conductance of Cu-glutaraldehyde-Cu junction. Conductance curves with obvious step can be seen in Fig. 2a and then were treated by logarithm and binning to construct the histogram (Fig. 2c). A peak at 10−3.45 G0 (27 nS) is found in the Fig. 2c. Comparing with glutaraldehyde, pronounced peak at 10−3.6 G0 (19 nS) is found for 1,4-benzenedicarboxaldehyde (Fig. 2b, d), and this value is consistent with our previously report [23]. Obviously, the peak intensity of 1,4-benzenedicarboxaldehyde is higher than that of glutaraldehyde. And the different intensity of peaks between glutaraldehyde and 1,4-benzenedicarboxaldehyde in the histograms may show internal property of benzene and alkane backbone.Fig. 2 Single-molecule conductance histograms of pentanedioic acid and glutaraldehyde. A log-scale conductance curves of a Cu-glutaraldehyde-Cu junctions and b Cu-1,4-benzenedicarboxaldehyde-Cu junctions. Log-scale conductance histogram of c Cu-glutaraldehyde-Cu junctions and d Cu-terephthalaldehyde-Cu junctions Then, we also construct histograms by using linear bin-size. Obvious difference is again observed for those molecules in Fig. 3. While rather weak peak can be found for petanedioic acid (Fig. 3a), pronounced peak is shown for 1,4-benzenedicarboxaldehyde (Fig. 3b). Conductance values of 4 and 11 nS are found for 1,4-benzenedicarboxaldehyde, which is different from the conductance value shown in log-scale histogram (Fig. 2d). This is caused by the different statistical methods between linear and logarithm bin-size, and different molecule-electrode configurations are shown in those histograms [23]. As our previously report, we can also obtain all conductance values of 4, 11, and 20 nS using data selection with linear bin-size [23]. However, the linear-bin histograms show even large difference of intensity between molecules with benzene and alkane backbone.Fig. 3 Comparison of the linear-scale conductance histograms of glutaraldehyde and 1,4-benzenedicarboxaldehyde. a The linear-scale conductance histogram of a Cu-glutaraldehyde-Cu junctions and b Cu-1,4-benzenedicarboxaldehyde-Cu junctions Return back to the difference histograms between benzene and alkane backbone, the weak peak can be caused by the less probability in the forming the molecular junctions. Usually, the junction formation probability can be analyzed by stretched distance distribution [26, 27] or counting the number of curves with step [28, 29] as previously reports. Here, we manually analyze the opportunity of step (typically, the curve with step length longer than 0.05 nm) in conductance curves showing step value smaller than 10−2 G0, it is found that step opportunity is around 40 % in forming junction of 1,4-benzenedicarboxaldehyde, while around 22 % for glutaraldehyde. From above, we can conclude that the anchoring group of aldehyde with benzene backbone has high junction formation probability than that with alkane backbone connecting with Cu electrode, and we will discuss it later. Comparative Study on Single-Molecule Conductance of Pentanedioic Acid and 1,4-Benzenedicarboxylic Acid with Cu Electrode In order to prove the role of backbone in forming molecular junction, we also use carboxylic acid as the anchoring group to comparing difference between petanedioic acid and 1,4-benzenedicarboxylic acid. As shown in Fig. 4, similar behavior is also found that 1,4-benzenedicarboxylic acid shows more pronounced peak comparing with petanedioic acid. Again, different conductance values are found in different statistical method between linear and log bin-size. According to Fig. 4, the junction formation probability of 1,4-benzenedicarboxylic acid is higher than that of petanedioic acid in both linear-scale and log-scale statistical histograms. We found that the step opportunity of 1,4-benzenedicarboxylic acid and petanedioic acid is around 51 % and 33 %, respectively, which illustrates the similar results as the 1,4-benzenedicarboxaldehyde and glutaraldehyde. However, molecules with carboxylic acid have larger junction formation probability than those with aldehyde anchoring group; this may be caused by that carboxylic acid can also bind to the Cu through carboxylate form with two O atoms binding to the electrode, while only one O atom can bind to the electrode for aldehyde group.Fig. 4 Comparison of linear-scale and log-scale conductance histogram of 1,4-benzenedicarboxylic acid and petanedioic acid. The log-scale conductance histograms of a Cu-1,4-benzenedicarboxylic acid-Cu junctions and b Cu-petanedioic acid-Cu junctions. The linear-scale conductance histograms of c Cu-1,4-benzenedicarboxylic acid-Cu junctions and d Cu-petanedioic acid-Cu junctions The Role of Backbone in Forming Molecular Junction According to above results, those molecules with benzene backbone have higher junction formation probabilities than those with alkane backbone connecting with Cu electrode, which should be caused by the stronger interaction between anchoring group and Cu in 1,4-benzenedicarboxaldehyde and 1,4-benzenedicarboxylic acid. Taking carboxylic acid as example, carboxylic acid binds to the Cu electrode through carboxylate group [30]. It was reported that the bond length of Cu-O for benzene system is shorter than that of alkane system, which may reveal that benzene-based molecule and Cu system has stronger attractive interaction than that of alkane-based molecule and Cu system [30, 31]. This can explain our result that the junction formation probability of 1,4-benzenedicarboxylic acid is higher than that of petanedioic acid. We deduce the similar situation for molecules with aldehyde anchoring group, since similar Cu–O bond is formed in the junctions [23]. The current work shows the import role of backbone in forming molecular junctions and may help the design of molecule in studying the electron transport of single-molecule junction. Conclusions In this work, we have measured the single molecular junction conductance of molecules with aldehyde and carboxylic acid anchoring groups. It has been found that the structure of backbone can influence the junction formation probability for same anchoring group (aldehyde and carboxylic acid), which contributes to the stronger attractive interaction between Cu and molecules with benzene backbone. Those results can guide the design molecule to form effective junction for studying molecular electronics. Acknowledgements We gratefully thank Zhejiang Provincial Natural Science Foundation of China (No. LR15B030002) and the financial support by the National Natural Science Foundation of China (Nos. 21573198 and 21273204). Authors’ Contributions FC, LLP, and ZWH carried out the experiments; JCM and JFZ analyzed the results. FC, YS, ZJN, and XSZ conceived and designed the experiments, analyzed the results, and wrote the manuscript. All authors read and approved the final manuscript. Authors’ Information FC is a Master’s degree student under the supervision of XSZ in the Institute of Physical Chemistry, Zhejiang Normal University, China. Competing Interests The authors declare that they have no competing interests. ==== Refs References 1. Xiang D Wang X Jia C Lee T Guo X Molecular-scale electronics: from concept to function Chem Rev 2016 116 4318 4440 10.1021/acs.chemrev.5b00680 26979510 2. Huang C Rudnev AV Hong W Wandlowski T Break junction under electrochemical gating: testbed for single-molecule electronics Chem Soc Rev 2015 44 889 901 10.1039/C4CS00242C 25560965 3. Arroyo CR Frisenda R Moth-Poulsen K Seldenthuis JS Bjornholm T van der Zant HSJ Quantum interference effects at room temperature in OPV-based single-molecule junctions Nanoscale Res Lett 2013 8 1 6 10.1186/1556-276X-8-234 23279756 4. Venkataraman L Klare JE Nuckolls C Hybertsen MS Steigerwald ML Dependence of single-molecule junction conductance on molecular conformation Nature 2006 442 904 907 10.1038/nature05037 16929295 5. Sedghi G Garcia-Suarez VM Esdaile LJ Anderson HL Lambert CJ Martin S Bethell D Higgins SJ Elliott M Bennett N Long-range electron tunnelling in oligo-porphyrin molecular wires Nat Nanotechnol 2011 6 517 523 10.1038/nnano.2011.111 21804555 6. Guo CL Wang K Zerah-Harush E Hamill J Wang B Dubi Y Xu BQ Molecular rectifier composed of DNA with high rectification ratio enabled by intercalation Nat Chem 2016 8 484 490 10.1038/nchem.2480 27102683 7. Darwish N Aragones AC Darwish T Ciampi S Diez-Perez I Multi-responsive photo- and chemo-electrical single-molecule switches Nano Lett 2014 14 7064 7070 10.1021/nl5034599 25419986 8. Fujii S Tada T Komoto Y Osuga T Murase T Fujita M Kiguchi M Rectifying electron-transport properties through stacks of aromatic molecules inserted into a self-assembled cage J Am Chem Soc 2015 137 5939 5947 10.1021/jacs.5b00086 25900686 9. Manrique DZ Huang C Baghernejad M Zhao X Al-Owaedi OA Sadeghi H Kaliginedi V Hong W Gulcur M Wandlowski T A quantum circuit rule for interference effects in single-molecule electrical junctions Nat Commun 2015 6 6389 10.1038/ncomms7389 25731608 10. Zhou XS Mao BW Amatore C Compton RG Marignier J-L Mostafavi M Nierengarten J-F Maisonhaute E Transient electrochemistry: beyond simply temporal resolution Chem Commun 2016 52 251 263 10.1039/C5CC07953E 11. Sun L Diaz-Fernandez YA Gschneidtner TA Westerlund F Lara-Avila S Moth-Poulsen K Single-molecule electronics: from chemical design to functional devices Chem Soc Rev 2014 43 7378 7411 10.1039/C4CS00143E 25099384 12. Tao NJ Electron transport in molecular junctions Nat Nanotechnol 2006 1 173 181 10.1038/nnano.2006.130 18654182 13. Rascón-Ramos H Artés JM Li Y Hihath J Binding configurations and intramolecular strain in single-molecule devices Nat Mater 2015 14 517 522 10.1038/nmat4216 25686263 14. Chen F Li XL Hihath J Huang ZF Tao NJ Effect of anchoring groups on single-molecule conductance: comparative study of thiol-, amine-, and carboxylic-acid-terminated molecules J Am Chem Soc 2006 128 15874 15881 10.1021/ja065864k 17147400 15. Aradhya SV Meisner JS Krikorian M Ahn S Parameswaran R Steigerwald ML Nuckolls C Venkataraman L Dissecting contact mechanics from quantum interference in single-molecule junctions of stilbene derivatives Nano Lett 2012 12 1643 1647 10.1021/nl2045815 22352939 16. Capozzi B Xia J Adak O Dell EJ Liu Z-F Taylor JC Neaton JB Campos LM Venkataraman L Single-molecule diodes with high rectification ratios through environmental control Nat Nanotechnol 2015 10 522 527 10.1038/nnano.2015.97 26005998 17. Zhou XY Wang YH Qi HM Zheng JF Niu ZJ Zhou XS Single-molecule conductance of dipyridines binding to Ag electrodes measured by electrochemical scanning tunneling microscopy break junction Nanoscale Res Lett 2014 9 77 10.1186/1556-276X-9-77 24528610 18. Kaliginedi V Rudnev AV Moreno-Garcia P Baghernejad M Huang C Hong W Wandlowski T Promising anchoring groups for single-molecule conductance measurements Phys Chem Chem Phys 2014 16 23529 23539 10.1039/C4CP03605K 25285778 19. Cheng ZL Skouta R Vazquez H Widawsky JR Schneebeli S Chen W Hybertsen MS Breslow R Venkataraman L In situ formation of highly conducting covalent Au-C contacts for single-molecule junctions Nat Nanotechnol 2011 6 353 357 10.1038/nnano.2011.66 21552252 20. Mishchenko A Zotti LA Vonlanthen D Bürkle M Pauly F Cuevas JC Mayor M Wandlowski T Single-molecule junctions based on nitrile-terminated biphenyls: a promising new anchoring group J Am Chem Soc 2011 133 184 187 10.1021/ja107340t 21155561 21. Zhou XS Wei YM Liu L Chen ZB Tang J Mao BW Extending the capability of STM break junction for conductance measurement of atomic-size nanowires: an electrochemical strategy J Am Chem Soc 2008 130 13228 13230 10.1021/ja8055276 18788809 22. Zhou XS Liang JH Chen ZB Mao BW An electrochemical jump-to-contact STM-break junction approach to construct single molecular junctions with different metallic electrodes Electrochem Commun 2011 13 407 410 10.1016/j.elecom.2011.02.005 23. Hong ZW Chen F Wang YH Mao JC Li DF Tang Y Shao Y Niu ZJ Zhou XS The binding sites of carboxylic acid group contacting to Cu electrode Electrochem Commun 2015 59 48 51 10.1016/j.elecom.2015.07.003 24. Peng ZL Chen ZB Zhou XY Sun YY Liang JH Niu ZJ Zhou XS Mao BW Single molecule conductance of carboxylic acids contacting Ag and Cu electrodes J Phys Chem C 2012 116 21699 21705 10.1021/jp3069046 25. Wang YH Hong ZW Sun YY Li DF Han D Zheng JF Niu ZJ Zhou XS Tunneling decay constant of alkanedicarboxylic acids: different dependence on the metal electrodes between air and electrochemistry J Phys Chem C 2014 118 18756 18761 10.1021/jp505374v 26. Hong W Manrique DZ Moreno-García P Gulcur M Mishchenko A Lambert CJ Bryce MR Wandlowski T Single molecular conductance of tolanes: experimental and theoretical study on the junction evolution dependent on the anchoring group J Am Chem Soc 2012 134 2292 2304 10.1021/ja209844r 22175273 27. Yoo PS Kim T High probability of single molecule junction formation with Ag electrodes Curr Appl Phys 2015 15 124 128 10.1016/j.cap.2014.11.016 28. Li XL He J Hihath J Xu BQ Lindsay SM Tao NJ Conductance of single alkanedithiols: conduction mechanism and effect of molecule-electrode contacts J Am Chem Soc 2006 128 2135 2141 10.1021/ja057316x 16464116 29. Li ZH Pobelov I Han B Wandlowski T Blaszczyk A Mayor M Conductance of redox-active single molecular junctions: an electrochemical approach Nanotechnology 2007 18 044018 10.1088/0957-4484/18/4/044018 30. Lennartz MC Atodiresei N Müller-Meskamp L Karthäuser S Waser R Blügel S Cu-adatom-mediated bonding in close-packed benzoate/Cu(110)-systems Langmuir 2009 25 856 864 10.1021/la801822e 19177646 31. Barbosa LAMM Sautet P Stability of chiral domains produced by adsorption of tartaric acid isomers on the Cu(110) surface: a periodic density functional theory study J Am Chem Soc 2001 123 6639 6648 10.1021/ja004336k 11439051

PMC005xxxxxx/PMC5001965.txt

==== Front SpringerplusSpringerplusSpringerPlus2193-1801Springer International Publishing Cham 304810.1186/s40064-016-3048-xResearchAssociations between parental impulsivity and child body mass index Sleddens Ester F. C. Ester.Sleddens@maastrichtuniversity.nl 1ten Hoor Gill A. Gill.tenHoor@maastrichtuniversity.nl 23Kok Gerjo G.Kok@maastrichtuniversity.nl 3Kremers Stef P. J. S.Kremers@maastrichtuniversity.nl 11 Department of Health Promotion, NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University Medical Center+, P.O. Box 616, 6200 MD Maastricht, The Netherlands 2 Department of Human Biology, NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University Medical Center+, Maastricht, The Netherlands 3 Department of Work and Social Psychology, Maastricht University, P.O. Box 616, 6200 MD Maastricht, The Netherlands 26 8 2016 26 8 2016 2016 5 1 142223 10 2015 11 8 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Objective The aim of this study was to examine the association between parental impulsivity and (12–15 year old) child body mass index (BMI). Methods In total, 300 parents completed a survey regarding their own impulsivity level (Barratt impulsiveness scale) and that of their child (impulsivity scale of the temperament in middle childhood questionnaire), and supplied details of their own and their child’s height and weight. Partial correlations were computed to assess relationships between both parental and child impulsiveness scores and child BMI z-scores, independent of parental BMI. Mediation analyses were performed to assess the potential mediating role of child impulsivity on the relationship between parental impulsivity and child BMI z-score. Results For daughters, parental impulsivity was significantly correlated with BMI z-score. Parent-reported child impulsivity was not related to child BMI z-score, and no evidence was found for a mediating effect of parent-reported child impulsivity on the relationship between parental impulsivity and child BMI z-score. Conclusion There is a stronger association between parental impulsivity and child BMI z-score than between child impulsivity and child BMI z-score. The relationship between parental impulsivity and-child BMI z-score could possibly be explained by parenting styles and practices. The potentially mediating role of parenting should be taken into account in future studies investigating the role of personality in children becoming overweight or obese. Electronic supplementary material The online version of this article (doi:10.1186/s40064-016-3048-x) contains supplementary material, which is available to authorized users. Keywords AdolescentsBody mass indexChildImpulsivityParentsNetherlands Heart Foundation2014T037Sleddens Ester F. C. Netherlands Organization for Health Research and Development (ZonMw)525001004ten Hoor Gill A. issue-copyright-statement© The Author(s) 2016 ==== Body Significance What is already known Although parental impulsivity may play an important role in the positive association between child impulsivity and BMI z-score there is a lack of studies exploring the role of parental impulsivity on childhood overweight. What this study adds The current study did not confirm findings from previous studies linking child impulsivity to child BMI z-score. This study demonstrates that parental impulsivity scores are associated with daughters’ BMI z-scores. This relationship could possibly be explained by parenting styles and practices. No evidence for a mediating role of parent-reported child impulsivity was found. Background Parental impulsivity may play an important role in the positive association between child impulsivity and BMIz-score. Numerous studies have been conducted assessing the influence of a child’s temperament (i.e., individual differences in behavioral patterns) on the development of weight status (e.g., Braet et al. 2007; Anzman-Frasca et al. 2012; Thamotharan et al. 2013; Bergmeier et al. 2014). With regard to child impulsivity, a temperamental trait, the evidence thus far suggests that high impulsivity in childhood is associated with being overweight and obese (Thamotharan et al. 2013). There are also indications that children with attention-deficit/hyperactivity disorder (ADHD), in which impulsivity is common, are heavier than children not diagnosed with ADHD (Cortese et al. 2008; Flier et al. 2013). It has been suggested that impulsive behaviors could contribute to excessive food intake (Davis 2010) and unhealthy snack consumption (Scholten et al. 2014). A Dutch study investigating parents of 6–13 year olds supported this notion by demonstrating that child impulsivity was indirectly associated with child body mass index (BMI) through overeating (Van den Berg et al. 2011). Moreover, a recent review study found that food-related impulsivity was more prevalent in obese children (Schag et al. 2013). Although several studies have been carried out to investigate the influence of child impulsivity on weight status, there is a lack of studies exploring the role of parental impulsivity on childhood overweight. Impulsive traits may be inherited, but are likely to be influenced over time by maturation and experience, including parenting processes. Therefore, parental impulsivity may also influence a child’s BMI z-score via the parenting processes. It is well-known that parents play a key role in influencing the development of their child’s weight status (e.g., Monasta et al. 2010; Cislak et al. 2012). It is also recognized that mothers and fathers have a differential role in influencing child weight-related outcomes, and, although the exact mechanisms remain unclear, these effects are likely to differ with regard to boys and girls (Wake et al. 2007; Berge et al. 2010; Jansen et al. 2013; Gevers et al. 2015). In this study, we assess relationships between parental impulsivity and children’s BMI z-score, and examine the potential mediating role of parent-reported child impulsivity on this relationship (see Fig. 1 for a visual representation of our conceptual model).Fig. 1 Model of the relationship between parental impulsivity and child BMI: potential mediating role of child impulsivity Methods Following recent requests for full disclosure (Peters et al. 2012), all research materials, data, analyses and output are available in a combined.rar archive labelled as Additional file 1. This study (as well as the consent procedure) was approved by the Ethical Review Committee Psychology and Neuroscience (ERCPN), Maastricht University, the Netherlands. Participants In total, 600 parents of 12–15 year old adolescents were randomly invited to participate in our study via Flycatcher. The parents that were invited included both mothers and fathers. Flycatcher, a representative online panel for the Dutch population (http://www.flycatcher.eu/; ISO 26362 and ISO20252; Dutch quality label, certifying that the panel can be used for social-scientific research) has 1300 registered parents of adolescents in the age range 12–15 years. After dropout (wrong e-mail address (N = 10) and the exclusion of questionable data (the use of straight-lining, or other patterned response strategies; nonsense answers on open questions, N = 23), 314 parents completed the study (53.2 % response rate). Procedure and measures Before participation in this computer study, all parents provided informed consent by clicking on the appropriate button. They were instructed that all questions related to their youngest child in the age range 12–15 years. No children participated in this study. Information about gender and age of both the parent and child, and the highest level of education attained by the parent (categorized into low—no education or primary education; medium—intermediate/high general secondary education or intermediate vocational education; and high—college degree or higher) had already been collected by Flycatcher. In addition to these background characteristics, parents were asked questions about both their own and their child’s impulsiveness and anthropometrics (i.e., height and weight). Impulsivity in children To measure impulsivity in children, we used the impulsivity scale from the parent-reported version of the temperament in middle childhood questionnaire (TMCQ) (Simonds and Rothbart 2004), adapted from the children’s behavior questionnaire (CBQ) (Rothbart et al. 2001). The CBQ is an instrument used to assess three broad dimensions of temperamental traits including surgency/extraversion, negative affectivity, and effortful control in children between 3 and 6 years old (Rothbart et al. 2001). The TMCQ has previously been validated against the CBQ in a Dutch sample (Sleddens et al. 2013). This validation study supported the applicability of the TMCQ as compared with all three previously mentioned temperamental traits of the CBQ. Impulsivity, as measured by the TMCQ, is considered to be a subfactor of surgency/extraversion. Parents were asked to respond to 13 statements describing their child’s impulsive behaviors (e.g., ‘My child decides what she/he wants very quickly and then goes after it’). They were asked to choose the answer that applied to their child the most, using a 5-point Likert scale ranging from 1 (Never) to 5 (Always). Cronbach’s alpha for the TMCQ scale in the present sample was 0.89. Corrected-item total correlations (CITCs) ranged from 0.41–0.73 for 13 items, with an average of 0.58, indicating homogeneity of the items (Nunnally and Bernstein 1994). Impulsivity in parents For the measurement of impulsivity in parents, we used the 30-item Barratt impulsiveness scale (BIS-11) (Patton et al. 1995). Sample items include ‘I do things without thinking’, and ‘I plan tasks carefully (reverse coded)’. Cronbach’s alpha for the BIS in the present sample was 0.82. However, three items had CITCs below the critical cut-off point of 0.15, defined by Nunnally and Bernstein (1994); ‘I make-up my mind quickly’ (CITC = 0.11), ‘I am happy-go-lucky’ (CITC = −0.03), and ‘I change residences’ (CITC = 0.13). The Cronbach’s alpha of the resulting 27-item scale was 0.83, with CITCs ranging from 0.15 to 0.56, with an average of 0.37. Body mass index Parents were asked to indicate (specified to one decimal place) their own and their child’s weight (kg) and height (cm), which we used to calculate BMI (in kg/m2). Each child’s BMI was then recoded into an age and gender-specific BMI z-score and compared to the national reference population (Fredriks et al. 2000). A BMI z-score >85th percentile was considered to indicate overweight and a BMI z-score >95th percentile was considered to indicate obesity (Barlow 2007). BMI z-scores <−5.0 or >5.0 were considered to be unrealistic, as advised by the WHO (De Onis et al. 2007). Only two children had a BMI z-score below −5 and were therefore removed from further analyses involving child BMI z-scores. Additionally, 12 children had missing values for exact day of birth. We also removed data from these children from all further analyses. Data analyses IBM SPSS statistics 20 was used to analyze the data. Descriptive analyses—frequencies (N), means and standard deviations (SD)—were calculated to assess the background characteristics of the sample (i.e., parent and child age and BMI (z-score), and parental educational level). Associations between both parental and child impulsiveness scores and child BMI z-scores were examined by computing partial correlations for the total sample (N = 300), for mothers and fathers separately, and for sons and daughters separately. The partial correlations were controlled for the potentially confounding variable of parental BMI. Additionally, single mediation analyses were carried out using the PROCESS software including the bootstrapping method with bias-corrected confidence estimates (Mackinnon et al. 2004; Preacher and Hayes 2004) in order to test the direct and indirect associations linking parental impulsivity scores to child BMI z-scores. Bootstrapping, a non-parametric sampling procedure, was used to assess the significance of indirect effects. In the present study, the 95 % confidence interval of the indirect effects was obtained with 5000 bootstrap resamples; results are statistically significant when 95 % confidence intervals do not overlap zero. This procedure was repeated for different groups (i.e., total sample, fathers and mothers, and sons and daughters). Also the mediation analyses were controlled for the potentially confounding variable of parental BMI. Results Key characteristics of the sample (N = 300) are depicted in Table 1. An approximately equal number of males and females participated in the study; 44.3 % (N = 133) of the questionnaires were completed by mothers and 55.7 % (N = 167) were completed by fathers. Mean (SD) age of the participants was 45.8 (4.8) years. The majority of the participants reported medium or higher levels of education: 44.7 % reported having intermediate/high general secondary education or intermediate vocational education, and 32.3 % reported having a college degree or higher. Gender of the children was also equally divided (49.7 % girls). The mean age of the children was 13.4 (SD = 1.1) years. With regard to weight status, 62.0 % of the parents were overweight or obese, and for children this percentage was 21.7 %. These percentages seem representative for the Dutch population except for a slight overrepresentation of highly educated parents.Table 1 Background characteristics of the sample (N = 300) Parent Child Gender (female: male) 133:167 149:151 Mean age in years (SD) 45.8 (4.8) 13.4 (1.1) Education level  Low (%) 69 (23.0 %) –  Medium (%) 134 (44.7 %) –  High (%) 97 (32.3 %) –  Mean BMI (z) (SD)a 26.74 (4.49) −0.05 (1.32)  Underweight (%) 3 (1.0 %) 36 (12 %)  Normal-weight (%) 111 (37.0 %) 199 (66.3 %)  Overweight (%) 125 (41.7 %) 36 (12 %)  Obese (%) 61 (20.3 %) 29 (9.7 %) All values are N’s, unless otherwise indicated Education level: low = no, or primary education; medium = intermediate/high general secondary education or intermediate vocational education; high = college degree or higher aFor the parents a BMI score was calculated; for the youngsters a BMI z-score was calculated Table 2 presents partial correlation coefficients (controlled for parental BMI) between both parental and child impulsivity scores and child BMI z-scores for different groups (i.e., total sample and sample split by parental and child gender). With regard to the correlation between parental impulsivity and child impulsivity, we found significant positive relationships for the total sample, (r ranging from 0.18 to 0.23, p < 0.05). However, these positive relations are caused by the significant correlations between maternal impulsivity and child impulsivity (r 0.33–0.47, p < 0.01) while the correlations between paternal impulsivity and child impulsivity were not significant (r = 0.12–0.14, p > 0.12). A positive significant correlation was only found between parental impulsivity scores and daughters’ BMI z-scores (p < 0.05). The correlation between parental impulsivity scores and child BMI z-scores was marginally significant for the total sample. The relationship between maternal impulsivity scores and child BMI z-scores was also marginally significant. Child impulsivity scores were only marginally significantly related to daughters’ BMI z-scores.Table 2 Partial correlations between parental impulsivity, child impulsivity and child BMI z-score Child impulsivity Child BMI z-score T ♂ ♀ T ♂ ♀ Parent impulsivity  Total sample 0.23 (<0.001) 0.27 (0.001) 0.18 (0.03) 0.10 (0.10)a 0.01 (0.92) 0.19 (0.02)  Fathers 0.12 (0.12) 0.14 (0.20) 0.12 (0.28) 0.05 (0.56) −0.10 (0.37) 0.15 (0.18)  Mothers 0.40 (< 0.001) 0.47 (< 0.001) 0.33 (0.007) 0.16 (0.07)a 0.08 (0.50) 0.24 (0.06)a Child impulsivity  Total sample 0.05 (0.41) −0.04 (0.65) 0.15 (0.08)a  Boy 0.05 (0.56) −0.05 (0.65) 0.17 (0.12)  Girl 0.6 (0.52) −0.00 (0.97) 0.12 (0.33) All values are r (significance level). In italics p < 0.05 aMarginally significant findings p < 0.10; analyses controlled for parental BMI Finally, the results of the mediation analyses to assess the potential mediating role of child impulsivity on the relationship between parental impulsivity scores and child BMI z-scores after controlling for parental BMI are presented in Table 3 (see Fig. 1 for a description of the different pathways). Child impulsivity was not found to be a mediator in any of the groups (i.e., total sample, fathers and mothers, or sons and daughters). For four out of the five groups (all except fathers), the A path linking parental impulsivity to child impulsivity was significant, indicating that parental impulsivity was positively associated with child impulsivity. Analyses further revealed that parental impulsivity scores were significantly associated with daughters’ BMI z-scores after controlling for child impulsivity (ß = 0.74, SE = 0.36, t = 2.07, p = 0.04), indicating a direct effect of parental impulsivity on daughters’ BMI z-scores. Indirect effects of parental impulsivity on child BMI z-scores were non-significant in all groups. Parental BMI did not act a as covariate, only on the relationship between parental impulsivity and children’s BMI in the total group (see Additional file 1).Table 3 Mediation analyses: a mediating role of child impulsivity on parent impulsivity scores and child body mass index z-scores Sample N Total sample 300 A path B path β = 0.54, SE = 0.13, t = 4.06, p < 0.001 β = 0.05, SE = 0.11, t = 0.47, p = 0.64 Direct effect (C’ path) β = 0.40, SE = 0.27, t = 1.50, p = 0.13 Indirect effect (a x b) β = 0.02, boot SE = 0.06, CI SE = −0.09 to 0.15 Fathers 167 A path β = 0.27, SE = 0.17, t = 1.55, p = 0.12 B path β = 0.08, SE = 0.16, t = 0.51, p = 0.61 Direct effect (C’ path) β = 0.19, SE = 0.36, t = 0.52, p = 0.61 Indirect effect (a × b) β = 0.02, boot SE = 0.05, CI SE = −0.04 to 0.20 Mothers 133 A path β = 0.99, SE = 0.20, t = 4.96, p < 0.001 B path β = −0.01, SE = 0.17, t = −0.09, p = 0.93 Direct effect (C’ path) β = 0.71, SE = 0.42, t = 1.70, p = 0.09* Indirect effect (a × b) β = −0.01, boot SE = 0.16, CI SE = −0.37 to 0.28 Son 151 A path β = 0.64, SE = 0.18, t = 3.63, p < 0.001 B path β = −0.08, SE = 0.17, t = −0.49, p = 0.62 Direct effect (C’ path) β = 0.09, SE = 0.39, t = 0.24, p = 0.81 Indirect effect (a × b) β = −0.05, boot SE = 0.10, CI SE = −0.30 to 0.12 Daughter 149 A path β = 0.44, SE = 0.19, t = 2.26, p = 0.03 B path β = 0.21, SE = 0.15, t = 1.38, p = 0.17 Direct effect (C’ path) β = 0.74, SE = 0.36, t = 2.07, p = 0.04 Indirect effect (a × b) β = 0.09, boot SE = 0.08, CI SE = −0.02 to 0.32 In bolditalics p < 0.05; * marginally significant findings p < 0.10 and p > 0.05 A path parental impulsivity on child impulsivity; B path child impulsivity on child BMI; direct effect (C’ path): parental impulsivity on child BMI after controlling for child impulsivity; indirect effect (a × b): parental impulsivity on child BMI; all analyses controlled for parental BMI Discussion This study examined associations between parental impulsivity and child BMI z-score, and assessed parent-reported child impulsivity as a potential mediator of this association. Analyses revealed that parental impulsivity scores were associated with daughters’ BMI z-scores. Evidence for a mediating role of parent-reported child impulsivity was not found. The possible mediating role of parenting should be examined in future studies. Although we did not assess aspects of parenting in our study, evidence is available suggesting that parental impulsivity contributes to child weight status through parenting processes. Parents can play a pivotal role in influencing their child’s BMI through their personality characteristics, and the way in which their personality is translated into parenting. Findings from a meta-analytic review on associations between the major personality factors and parenting showed that effect sizes were significant and robust (Prinzie et al. 2009), in line with the notion that personality affects parenting. This finding has been confirmed in more recent studies (e.g., De Haan et al. 2012; Sleddens et al. 2014). For example, De Haan et al. (2012) reported that the temperamental traits of agreeableness and extraversion were important predictors of parental over-reactivity and warmth. Sleddens et al. (2014) showed that parenting constructs such as nurturance, structure and behavioral control were positively correlated with the temperamental traits of extraversion, agreeableness, conscientiousness and openness to experience. Parenting constructs such as coercively controlling and overprotecting home environments were positively related to parental neuroticism. Since multiple studies have shown that an authoritative parenting style is associated with healthier child outcomes, including lower BMI values (Sleddens et al. 2011; Pinquart 2014), it is likely that parental impulsivity contributes to child weight status through parenting processes. In the present study, differences were found among fathers and mothers, and sons and daughters. For the relationship between parental impulsivity and child BMI z-score, there was only a significant positive relationship for daughters. It could be that impulsive girls with impulsive parents are more prone to disruptive eating behaviors which may in turn lead to increases in BMI z-scores. The current study did not confirm findings from previous studies linking child impulsivity to child BMI z-score (e.g., Bodell et al. 2012; Thamotharan et al. 2013). It could be that the effect of a child’s temperament on the development of his/her BMI z-score depends somewhat on parenting styles and practices, as reported by Zeller et al. (2008); Wu et al. (2011). Children with a difficult temperament and with mothers who scored low on nurturing were found to have a significantly higher risk of becoming overweight or obese (Zeller et al. 2008; Wu et al. 2011). As well as the influence of child temperament, future studies should consider the potential role of additional factors on child BMI z-score. For instance, the link between child impulsivity and BMI could be indirect. Braungart-Rieker et al. (2014) found that child impulsivity was only linked to child BMI indirectly through children’s food approach eating styles. However, secondary analyses on additional data of our study did not show evidence for the mediating mechanism of either parental BMI or child physical activity levels on the link between child impulsivity and child BMI z-scores, also not for the gender-related sub-groups. With regard to future research initiatives, it may also be beneficial to measure the child’s impulsivity directly. To specify, Thamotharan et al. (2013) conducted a meta-analytic review on the relation between child impulsivity and weight status and findings revealed a moderate effect size, such that impulsivity was greater among overweight and obese children, relative to healthy weight children. They conducted a sub-group analysis to see whether the effect sizes differed depending on type of measure used (i.e., behavioral or self-report). A large effect size was found for behavioral measures of impulsivity compared to a small effect size for self-report measures. The findings of the current study should be seen in light of these findings. Usually self-report measures appear to assess broad domains of behaviors over a larger time interval and therefore also more prone to several forms of biases. This is also consistent with results that behavioral and self-report measures of impulsivity are usually not highly correlated (Dougherty et al. 2005). One advantage of our study was that males and females were equally represented. Several limitations of the present study should, however, also be acknowledged. All of the variables we measured were reported by parents. This could at least in part explain why we did not find a relationship between child impulsivity and child BMI z-score. Our study may have been prone to social desirability effects, which may have diminished any association between impulsivity and child BMI z-score. Highly educated parents were somewhat overrepresented in our sample. Therefore, we must be cautious in generalizing our study findings. Additionally, our cross-sectional design limits the causal conclusions that can be drawn. Future studies are warranted to explore the influence of dynamic parenting processes as well as parent and child temperament on a child’s risk of becoming overweight or obese. As our results demonstrate, such studies should also take gender differences into account. The association between parental impulsivity and child BMI z-score is stronger than the association between parent-reported child impulsivity and child BMI z-score. Additional file 10.1186/s40064-016-3048-x Full disclosure of data, analyses and output. Authors’ contributions ES and GtH conceived of, designed and coordinated the study, contributed to the acquisition, performed the statistical analyses, and drafted the manuscript. SK and GK participated in the design, analysis of the data, and revision of the manuscript. All authors read and approved the final manuscript. Acknowledgements This research was funded by the Netherlands Heart Foundation (project number 2014T037) to ES, and by the Netherlands Organization for Health Research and Development (ZonMw; project number 525001004) to GtH. Competing interests The authors declare that they have no competing interests. ==== Refs References Anzman-Frasca S Stifter CA Birch LL Temperament and childhood obesity risk: a review of the literature J Dev Behav Pediatr 2012 33 732 745 10.1097/DBP.0b013e31826a119f 23095495 Barlow SE (2007). Expert committee recommendations regarding the prevention, assessment, and treatment of child and adolescent overweight and obesity: summary report. Pediatrics 120(S4):S164–S192. Aappublications.org/content/120/Supplement_4 Berge JM Wall M Bauer KW Neumark-Sztainer D Parenting characteristics in the home environment and adolescent overweight: a latent class analysis Obesity 2010 18 818 825 10.1038/oby.2009.324 19816417 Bergmeier H Skouteris H Horwood S Hooley M Richardson B Associations between child temperament, maternal feeding practices and child body mass index during the preschool years: a systematic review of the literature Obes Rev 2014 15 9 18 10.1111/obr.12066 23957249 Bodell LP Joiner TE Ialongo NS Longitudinal association between childhood impulsivity and bulimic symptoms in African American adolescent girls J Consult Clin Psychol 2012 80 313 316 10.1037/a0027093 22289129 Braet C Claus L Verbeken S Van Vlierberghe L Impulsivity in overweight children Eur Child Adolesc Psychiatry 2007 16 473 483 10.1007/s00787-007-0623-2 17876511 Braungart-Rieker JM Moore ES Planalp EM Lefever JB Psychosocial pathways to childhood obesity: a pilot study involving a high risk preschool sample Eat Behav 2014 15 528 531 10.1016/j.eatbeh.2014.07.004 25098723 Cislak A Safron M Pratt M Gaspar T Luszczynska A Family-related predictors of body weight and weight-related behaviours among children and adolescents: a systematic umbrella review Child Care Health Dev 2012 38 321 331 10.1111/j.1365-2214.2011.01285.x 21752064 Cortese S Angriman M Maffeis C Isnard P Konofal E Lecendreux M Purper-Ouakil D Vincenzi B Dalla Bernardina B Mouren M-C Attention-deficit/hyperactivity disorder (ADHD) and obesity: a systematic review of the literature Crit Rev Food Sci Nutr 2008 48 524 537 10.1080/10408390701540124 18568858 Davis C Attention-deficit/hyperactivity disorder: associations with overeating and obesity Curr Psychiatry Rep 2010 12 389 395 10.1007/s11920-010-0133-7 20632134 De Haan AD Deković M Prinzie P Longitudinal impact of parental and adolescent personality on parenting J Pers Soc Psychol 2012 102 189 199 10.1037/a0025254 21875227 De Onis M Onyango AW Borghi E Sivam A Nishida C Siekmann J Development of a WHO growth reference for school-aged children and adolescents Bull World Health Org 2007 85 660 667 10.2471/BLT.07.043497 18026621 Dougherty DM Mathias CW Marsh DM Jagar AA Laboratory behavioral measures of impulsivity Behav Res Methods 2005 37 82 90 10.3758/BF03206401 16097347 Fliers EA Buitelaar JK Maras A Bul K Höhle E Faraone SV Franke B Rommelse NNJ ADHD is a risk factor for overweight and obesity in children J Dev Behav Pediatr 2013 34 566 574 10.1097/DBP.0b013e3182a50a67 24131879 Fredriks AM Van Buuren S Wit JM Verloove-Vanhorick SP Body index measurements in 1996-7 compared with 1980 Arch Dis Child 2000 82 107 112 10.1136/adc.82.2.107 10648362 Gevers DWM Van Assema P Sleddens EFC De Vries NK Kremers SPJ Associations between general parenting, restrictive snacking rules, and adolescent’s snack intake. The roles of fathers and mothers and interparental congruence Appetite 2015 87 184 191 10.1016/j.appet.2014.12.220 25555538 Jansen PW Giallo R Westrupp EM Wake M Nicholson JM Bidirectional associations between mothers’ and fathers’ parenting consistency and child BMI Pediatrics 2013 132 e1513 e1520 10.1542/peds.2013-1428 24276845 MacKinnon DP Lockwood CM Williams J Confidence limits for the indirect effect: distribution of the product and resampling methods Multivariate Behav Res 2004 39 99 10.1207/s15327906mbr3901_4 20157642 Monasta L Batty GD Cattaneo A Lutje V Ronfani L Van Lenthe FJ Early-life determinants of overweight and obesity: a review of systematic reviews Obes Rev 2010 11 695 708 10.1111/j.1467-789X.2010.00735.x 20331509 Nunnally JC Bernstein IH Psychometric theory 1994 New York McGraw-Hill Patton JH Stanford MS Barratt ES Factor structure of the Barratt impulsiveness scale J Clin Psychol 1995 51 768 774 10.1002/1097-4679(199511)51:6<768::AID-JCLP2270510607>3.0.CO;2-1 8778124 Peters G-JY Abraham C Crutzen R Full disclosure: doing behavioural science necessitates sharing Eur Health Psychol 2012 14 77 84 Pinquart M Associations of general parenting and parent-child relationship with pediatric obesity: a meta-analysis J Pediatr Psychol 2014 39 381 393 10.1093/jpepsy/jst144 24756228 Preacher KJ Hayes AF SPSS and SAS procedures for estimating indirect effects in simple mediation models Behav Res Methods Instrum Comput 2004 36 717 731 10.3758/BF03206553 15641418 Prinzie P Stams G-JJM Deković M Reijntjes AHA Belsky J The relations between parents’ big five personality factors and parenting: a meta-analytic review J Pers Social Psychol 2009 2 351 362 10.1037/a0015823 Rothbart MK Ahadi SA Hershey KL Fisher P Investigations of temperament at three to 7 years: the children’s behavior questionnaire Child Dev 2001 72 1394 1408 10.1111/1467-8624.00355 11699677 Schag K Schönleber J Teufel M Zipfel S Giel KE Food-related impulsivity in obesity and binge eating disorder—a systematic review Obes Rev 2013 14 477 495 10.1111/obr.12017 23331770 Scholten EWM Schrijvers CTM Nederkoorn C Kremers SPJ Rodenburg G Relationship between impulsivity, snack consumption and children’s weight PLoS One 2014 9 e88851 10.1371/journal.pone.0088851 24586413 Simonds J, Rothbart MK (2004) The temperament in middle childhood questionnaire (TMCQ): a computerized self-report measure of temperament for ages 7–10. In: Paper presented at: occasional temperament conference. Athens Sleddens EFC Gerards SMPL Thijs C De Vries NK Kremers SPJ General parenting, childhood overweight and obesity-inducing behaviors: a review Int J Pediatr Obes 2011 6 e12 e27 10.3109/17477166.2011.566339 21657834 Sleddens EFC Kremers SPJ De Vries NK Thijs C Measuring child temperament: validation of a 3-item temperament measure and 13-item impulsivity scale Eur J Dev Psychol 2013 10 392 401 10.1080/17405629.2012.688102 Sleddens EFC, O’Connor TM, Watson KB, Hughes SO, Power TG, Thijs C, De Vries NK, Kremers SPJ (2014) Development of the comprehensive general parenting questionnaire for caregivers of 5–13 year olds. Int J Behav Nutr Phys Act 11:15. http://www.ijbnpa.org/content/11/1/15 Thamotharan S Lange K Zale EL Huffhines L Fields S The role of impulsivity in pediatric obesity and weight status: a meta-analytic review Clin Psychol Rev 2013 33 253 262 10.1016/j.cpr.2012.12.001 23313762 Van den Berg L Pieterse K Malike JA Luman M Willems van Dijk K Oosterlaan J Dekenarre-van de Waal HA Association between impulsivity, reward responsiveness and body mass index in children Int J Obes 2011 35 1301 1307 10.1038/ijo.2011.116 Wake M Nicholson JM Hardy P Smith K Preschooler obesity and parenting styles of mothers and fathers: Australian national population study Pediatrics 2007 120 e1520 e1527 10.1542/peds.2006-3707 18055667 Wu T Dixon WE Dalton WT Tudiver F Liu X Joint effects of child temperament and maternal sensitivity on the development of childhood obesity Matern Child Health 2011 15 469 477 10.1007/s10995-010-0601-z Zeller MH Boles RE Reiter-Purtill J The additive and interactive effects of parenting style and temperament in obese youth seeking treatment Int J Obes 2008 32 1474 1480 10.1038/ijo.2008.125

PMC005xxxxxx/PMC5001966.txt

==== Front SpringerplusSpringerplusSpringerPlus2193-1801Springer International Publishing Cham 311210.1186/s40064-016-3112-6ResearchRemoval of fluoride from water using a novel sorbent lanthanum-impregnated bauxite http://orcid.org/0000-0002-1834-1343Vivek Vardhan C. M. vivekvardhan2@gmail.com Srimurali M. msrimurali@gmail.com Department of Civil Engineering, Sri Venkateswara University, Tirupati, Andhra Pradesh 517501 India 26 8 2016 26 8 2016 2016 5 1 14262 10 2015 19 8 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.A novel sorbent, Lanthanum-Impregnated Bauxite (LIB), was prepared to remove fluoride from water. To understand the surface chemical composition and morphology, LIB was characterized using X-ray diffraction and scanning electron microscopy techniques. Experiments were performed to evaluate the sorption potential, dose of sorbent, kinetics, equilibrium sorption capacity, pH and influence of anions for defluoridation by LIB. Equilibrium isothermal studies were conducted to model the sorption and regeneration studies were carried out to evaluate the reusability of LIB. The results showed that LIB, at a dose of 2 g/L could remove 99 % of fluoride from an initial concentration of 20 mgF/L. Kinetic studies revealed the best fit of pseudo second order model. The sorption followed Langmuir isotherm model and the maximum sorption capacity of LIB for removal of fluoride was found to be 18.18 mg/g. Naturally occurring pH of water was found to be favorable for sorption. Usually occurring anions in water except nitrates influenced sorption of fluoride by LIB. Keywords FluorideWaterRemovalAdsorptionLanthanumBauxiteissue-copyright-statement© The Author(s) 2016 ==== Body Background Excessive fluoride in drinking water causes serious health problems such as brittleness of bones, dwarfishness, fluorosis and cancers (Chinoy 1991). The maximum contaminant level (MCL) of fluoride in drinking water is 1.5 mg/L, according to the World Health Organization (2004). Groundwater with fluoride concentration >1.5 mg/L is prevalent in several regions of the world, warranting treatment (Yeşilnacar et al. 2016; Atasoy et al. 2013; Vijaya Kumar et al. 1991; Gaciri and Davies 1993; Czarnowski et al. 1996). Several technologies such as adsorption (Vivek Vardhan and Karthikeyan 2011), coagulation and flocculation (Emamjomeh and Sivakumar 2006), electrodialysis (Adhikary et al. 1989), electrocoagulation (Khatibikamala et al. 2010) and reverse osmosis (Simons 1993) have been tried to remove fluoride from water with varying degrees of success. Chemical precipitation of fluoride using alum and lime, known as Nalgonda Technique (Nawlakhe et al. 1978) can be used for fluoride removal. However, it poses some problems such as generation of large volumes of sludge, which is difficult to deal with. Adsorption is considered to be a feasible technique especially for household applications or for small communities (Srimurali et al. 1998). Various sorbents such as activated alumina (Boruff 1934; Fink and Lindsay 1936; Swope and Hess 1937), bone char (Nemade et al. 2002), bauxite (Sujana and Anand 2011), magnesium amended activated alumina (Maliyekkal et al. 2008) and rice husk (Vivek Vardhan and Karthikeyan 2011) have been tried (Bhatnagar et al. 2011; Ayoob et al. 2008). Among various adsorbents used activated alumina is deemed to be the selective sorbent for removal of fluoride from water (Boruff 1934; Fink and Lindsay 1936; Swope and Hess 1937). However, due to some drawbacks such as optimum removal at a low pH value of 5.5, its practical scope of applicability is limited. Recently various rare earth materials such as lanthanum (Na and Park 2010), lanthanum modified activated alumina (Cheng et al. 2014), lanthanum oxide (Nagendra Rao and Karthikeyan 2012), lanthanum impregnated green sand (Vivek Vardhan and Srimurali 2016), cerium (Xu et al. 2001) and yttrium (Raichur and Basu 2001) have been used as sorbents for removal of fluoride from water. Though lanthanum has got good affinity for fluoride, there are some difficulties related to its use as an adsorbent. Compounds of lanthanum are present in fine powder form. Application of lanthanum compounds in powder form for adsorption is associated with practical limitations such as difficulty in separation from liquid, impeded hydraulic flow and leachate of metal with treated water (Maliyekkal et al. 2008). To overcome these problems, lanthanum had to be fixed onto a suitable substrate. Bauxite is an ore of aluminum and is abundantly available at low cost. In the present investigation, an attempt has been made to impregnate lanthanum onto bauxite, in order to develop a low-cost adsorbent and also to study the synergetic effect of lanthanum and bauxite on fluoride removal as well as to overcome the drawbacks associated with the use of lanthanum powder. Lanthanum Impregnated Bauxite (LIB) was prepared using La2CO3. La2CO3 is the base material for synthesis of other forms of Lanthanum and is available at low-cost. Also the quantity of La2CO3 that goes into impregnation for synthesis of LIB is very less. So, when used on a massive scale, LIB turns out to be a very low-cost adsorbent. However, the exact cost analysis will be done in future studies. LIB was characterized using X-ray diffraction (XRD) studies and Scanning Election Microscopy (SEM). Adsorption experiments were conducted in batch mode. Experiments involving Kinetics, isothermal equilibrium, pH and regeneration studies were carried out to evaluate the practical feasibility of application of LIB as an adsorbent for removal of fluoride from water. Methods Chemicals All reagents used in the present investigation were of analytical grade and procured from E. Merck Ltd, India. Water used in all batch sorption studies was laboratory distilled water prepared with a glass distillation unit (pH 6.7 ± 0.1 and specific conductivity 2.0 to 4.3 µS/cm). Stock solution of fluoride of 100 mg/L was prepared with distilled water using sodium fluoride. Aqueous fluoride solution was prepared by adding appropriate quantity of stock fluoride solution into distilled water and used in all adsorption experiments unless otherwise specified. LIB was prepared by thermal impregnation method as described below in adsorbent preparation. Raw bauxite was collected from mines at Mahboobabad, India. Lanthanum carbonate was purchased from Indian Rare Earths Limited, Aluva, Kerala, India. Adsorbent preparation Raw bauxite was crushed and sieved to get <75 micron particle size. Bauxite so obtained was heated in a muffle furnace at 400 °C for 4 h. This heated bauxite was cooled to room temperature in a desiccator and is called calcined bauxite. Calcined bauxite was stored in an air tight plastic container for further use. In a separate conical flask, La2(CO3)3 of 0.5 g was mixed with 50 mL of distilled water and dilute HCl was added to it drop wise till the La2(CO3)3 got completely dissolved. To this solution 20 g of prepared calcined bauxite was added and mixed using a magnetic stirrer for 3 h. The liquids were strained off and the solid material obtained was washed with distilled water. It was dried in a water bath at 110 °C for 6 h and subsequently heated in a muffle furnace at 950 °C for 4 h and then cooled. The material thus obtained was called LIB and used as an adsorbent in all further investigations. LIB was processed at high temperatures for lanthanum impregnation, whereas bauxite was calcined to improve its surface properties. So, this paper has the limitation of not studying bauxite and LIB, subject to same thermal treatment to bring out the exact differences due to lanthanum impregnation. Calcined bauxite is hereafter referred to as simply bauxite in this paper. Characterization of adsorbent LIB samples were analyzed by X-ray powder diffraction (XRD) technique before and after adsorption for studying its mineralogy. XRD analysis was carried out using a X-ray diffractometer, Philips: PW1830 with CuKα radiation. To study the surface morphology, scanning electron microscope (SEM) was used. SEM and EDAX images were obtained from a Carl Zeiss, EVO MA15 instrument. Particle size distribution was analyzed using Ankersmid particle size analyzer. Pore size analysis of bauxite and LIB were done by using a micropore analyzer (ASAP 2020, Micromeritics, USA) by Nitrogen chemisorption isotherm technique (Carabineiro et al. 2011). Batch adsorption experiments Sorption experiments were conducted in batch mode using 250 mL Teflon flasks with a 100 mL of 20 mg/L of aqueous fluoride solution. A known quantity of adsorbent was added to the prepared fluoride solution in Teflon flasks. It was agitated using a rotary shaker of make Kaizen Imperial at 160 rpm and at room temperature for specific contact periods ranging from 0 to 360 ± 1 min. The solutions contained in the flasks were then withdrawn at specified contact periods, filtered with 42 Whatman filter paper of pore size 2.5 µm and analysed for residual fluoride using SPADNS method (APHA) (APHA et al. 1996) at 570 Nm. A spectrophotometer, Evolution 201, of Thermo Scientific make was used to analyze fluoride. The contact period, at which there was no further reduction of fluoride, is considered the equilibrium contact time. Similarly the optimum usage of adsorbent was studied by varying the sorbent dose ranging from 0 to 8 ± 0.01 g/L for a constant equilibrium contact time. To understand the influence of pH, sorption experiments were conducted at different pH values ranging from 2 to 12. Starting pH adjustments were made using diluted NaOH and H2SO4. pH was measured using a Hanna make, pH analyzer. Optimum values obtained during preliminary investigations for various parameters were used in all further detailed experimentation. Fluoride ion concentrations varying from 5 to 70 mg/L were used in sorption equilibrium investigations, to arrive at the best fitting isothermal model. The reporting fluoride concentration range by SPADNS method is from 0 to 1.4 ± 0.1 mg/L. Appropriate dilutions of samples were made when fluoride exceeded the above mentioned concentration range. Concentrations of lanthanum and aluminum were measured using atomic absorption spectrometer with a graphite furnace (AAS, GBC 932 Plus). Kinetics of sorption In the present investigation pseudo first order, pseudo second order and intraparticle diffusion models were studied to understand the kinetics of adsorption of fluoride using bauxite and LIB. Pseudo first order equation and pseudo second order equation The mathematical equation of pseudo first order equation is as given in Eq. (1) (Lagergren 1898). 1 Logqe-qt=Logqe-K1t/2.303 where qe represents adsorbed fluoride at equilibrium and qt represents adsorbed fluoride at time t·k1 (L/min) represents rate constant of adsorption. A plot was drawn between (t) versus Log (qe − qt). K1 and qe were obtained from its slope and intercept. The linear form of mathematical equation for pseudo second order model is given in Eq. (2) (Ho and McKay 1999). 2 t/qt=1/K2qe2+t/qe where, K2 is a rate constant. Intraparticle diffusion analysis To design and control an adsorption system the mechanism involved and the rate limiting step are to be determined. In a well agitated system, migration of sorbate from a bulk solution to surround the adsorbent is not difficult (Weber and Morris 1963). Therefore bulk transport is rapid and it cannot be rate limiting. Similarly sorption of fluoride ions onto the active sites of sorbent is rapid and so this too cannot be a rate limiting step (Na and Park 2010). So, either film diffusion or intraparticle diffusion acts as rate slowing step or eventually becomes the rate controlling step (Yousef et al. 2011). To identify the rate controlling step and also to understand the mechanism involved in sorption, intraparticle diffusion equation, derived from unsteady state diffusion in flat plate is employed, which is given in Eq. (3) (Oliveira et al. 2005) 3 qt=kid·t1/2+C where, C is a constant, proportional to boundary layer thickness and kid is rate constant. If the plot of qt versus t1/2 is linear, it indicates the involvement of intraparticle diffusion. If the obtained straight line passes through, origin then it indicates that intraparticle diffusion alone is the rate limiting step. Conversely if the obtained straight line does not pass through origin, it indicates involvement of some other mechanism, in addition to intraparticle diffusion (Oliveira et al. 2005). Effect of competing ions The influence of anions on the efficiency of fluoride sorption by LIB was investigated. In order to find this, various individual ions of Cl−, SO42−, PO43−, HCO3− and NO3− of concentrations up to 100 mg/L were added each separately into 20 mg/L of aqueous fluoride solution and adsorption experiments were carried out using 2 g/L of LIB as well as 6 g/L of bauxite. The liquid samples were then withdrawn after reaching equilibrium time and analyzed for residual fluoride concentrations. Wherever the influence of phosphates was analyzed, for every 16 mg/L of PO43−, an error correction of −0.1 mg/L was made to rectify its interference with SPADNS method (Hach company 1989–2014). Determination of pH zero point charge pH of zero point charge (pHzpc) was found by a batch equilibrium method (Rivera-Utrilla et al. 2001). Typically, NaCl of 0.01 M concentration and 50 mL in quantity was taken into six conical flasks. pH of these solutions were varied between 2 and 12 using H2SO4 or NaOH. One gram of LIB was added to each flask and a plot was drawn between pH before addition of LIB and pH after addition of LIB. This plot yielded a straight line. The flasks were then agitated at room temperature for 48 h and then the pH values were noted. Now a plot was drawn between pH value of solution before agitation and pH value after 48 h of agitation, which eventually yielded a curve. The intersection point of the straight line and the curve is the value of pHzpc of LIB. Similar experiments were conducted with bauxite. Results obtained by this method are in close agreement with the results obtained by Carabineiro et al. (2011). Regeneration experiments After the agitated batch sorption experiments were conducted under optimal conditions with 20 mg/L aqueous fluoride solution, the liquids were strained off and the sorbent which got loaded to capacity was air dried for 48 h. Further it was desorbed by agitation with various eluents such as distilled water, NaOH and HCl, for a period of 180 min. The best desorbent was considered as the regeneration agent. Cycles of regeneration Regular batch adsorption experiments were conducted with a 20 mg/L of aqueous fluoride solution using adsorbent under optimal experimental conditions. After sorption, the spent sorbent, which got loaded to capacity was separated by filtration using a 42 Whatman filter paper and air dried for 48 h. Subsequently it was desorbed using the most appropriate eluent found through experimentation. Such regenerated sorbent was again separated and air dried to be used as a fresh sorbent for removal of fluoride from a 20 mg/L of aqueous fluoride solution. After agitation, the residual fluoride in solution was measured. This process was repeated several times until the residual fluoride exceeded the permissible limits. The number of cycles until fluoride in solution reached the permissible limit was considered the optimum cycles of reusability of sorbent. Results and discussion Sorbent characterization SEM image and EDAX spectra of LIB are presented in Fig. 1 and Table 1 respectively. The white colored dense precipitates observed could be attributed to impregnated lanthanum on the background granules of bauxite. In the EDAX spectrum, the elements, Al, La, Ti and Fe can be noticed, which give indirect evidence of presence of lanthanum on bauxite. From the particle size distribution analysis, it was observed that more than 90 % of LIB and bauxite particles/aggregates fell in the range of 40–55 µm.Fig. 1 SEM image of LIB Table 1 EDAX of LIB Element Atomic number Series Unn.C (Wt%) Norm.c (Wt%) Atom.c (at.%) Error (1 sigma) (Wt%) Ti 22 L-series 68.8 64.47 58.48 16.60 Al 13 K-series 21.13 19.8 31.88 1.28 Fe 26 L-series 10.85 10.17 7.91 3.87 La 57 M-series 5.93 5.56 1.74 4.38 Unn.C (Wt%): The unnormalised concentration in weight percent of the element Norm.C (Wt%): The normalised concentration in weight percent of the element Atom.c (at.%): The atomic weight percent Error (1 sigma) (Wt%): The error in the weight percent concentration at the 1 sigma level Influence of sorbent dose Experiments were conducted on LIB and bauxite separately to find their dose required for removal of fluoride from water. The corresponding experimental results are presented in Fig. 2. It can be observed from the figure that LIB at a dose of 2 g/L could remove 99 % of fluoride from an initial fluoride concentration of 20 mg/L, whereas bauxite at 6 g/L could remove 94 % of fluoride from an initial fluoride concentration of 20 mg/L. Removal of fluoride by bauxite was low, compared to LIB, possibly due to high affinity of lanthanum for fluoride. The concentration of lanthanum and aluminum in treated water were found to be 0.05 ± 0.01 and 0.02 ± 0.01 mg/L respectively, which are not harmful (Feng et al. 2006; Bureau of Indian Standards 2012).Fig. 2 Comparison of influence of doses of LIB and bauxite on fluoride removal. Initial fluoride = 20 mg/L Kinetic studies Influence of contact time Attaining equilibrium of adsorption during an adsorption process involves various diffusion mechanisms before the sorbate finally adsorbs onto the active adsorption sites on the sorbent (Biswas et al. 2009). Adsorption kinetics explains the rates at which different stages involving various mechanisms proceed. In the present study the time taken for adsorption of fluoride onto bauxite and LIB was investigated. It was observed that it took 120 min for removal of fluoride to below 1.5 mg/L using bauxite (Figure not shown). Figure 3 shows the time taken for sorption of various concentrations of fluoride onto LIB. Sorption was rapid in the initial 30 min. Later the rate of adsorption got stabilized. The plot of pseudo second order model for sorption of fluoride onto LIB is given in Fig. 4. The calculated parameters of the above two models for both bauxite and LIB are presented in Table 2. It can be observed from the obtained R2 values, that pseudo second order model fits best to both bauxite and LIB. Figure 5, depicts the plot between qt versus t1/2 for LIB. It can be observed from the figure that the plot yielded almost a straight line tending to pass through origin. This suggests that intraparticle diffusion alone is the rate limiting step. In general, in a well agitated system, film diffusion cannot be a rate limiting step (Weber and Morris 1963). It can be observed from Table 2 that Pseudo-second order model fits better to both LIB as well as to bauxite, based on regression analysis. This suggests a predominance of involvement of active chemical sites that aid in the process of sorption. The pore size characteristics of bauxite, LIB and activated alumina (Maliyekkal et al. 2008) are presented in Table 3. It can be observed that the pore size diameter has got reduced from 63 (Bauxite) to 54 nm (LIB), possibly due to lanthanum impregnation. The pores in activated alumina fall in mesoporous range and that of bauxite and LIB fall in macroporous range according to IUPAC classification (Everett 1973, 1976). This explains the high rate of sorption of fluoride onto LIB.Fig. 3 Kinetics of fluoride removal by LIB at various initial concentrations of fluoride (adsorbent dose = 2 g/L) Fig. 4 Plot of pseudo-second-order equation for sorption of fluoride onto LIB Adsorbent dose = 2 g/L. Initial fluoride = 20 mg/L Table 2 Comparison of parameters of kinetic models for adsorption of fluoride onto LIB and bauxite Pseudo-first-order Pseudo-second-order Intraparticle diffusion LIB qe (exp) = 9.8 (mg/g) qe (cal) = 5.462 (mg/g) K1 = 0.0152 (min−1) R2 = 0.8921 qe (cal) = 10.75 (mg/g) K2 = .0037 (min−1) R2 = 0.9965 Kid = 0.438 (g mg−1 min−1) C = 2.9041 R2 = 0.6986 Bauxite qe (exp) = 3.1666 (mg/g) qe (cal) = 0.9156 (mg/g) K1 = 0.0154 (min−1) R2 = 0.8956 qe (cal) = 1.781(mg/g) K2 = 0.0225 (min−1) R2 = 0.9964 Kid = 0.073 (g mg−1 min−1) C = 0.4837 R2 = 0.699 Fig. 5 Intraparticle diffusion plot for sorption of fluoride onto LIB Table 3 Pore size characteristics of bauxite, LIB and activated alumina Pore characteristics Bauxite LIB Activated alumina BET surface area 7 m2/g 14 m2/g 242 m2/g BJH adsorption cumulative volume of pores between 17.000 and 3000.000 Å diameter 0.09 cm3/g 0.17 cm3/g 0.29 cm3/g BJH Desorption cumulative volume of pores between 17.000 and 3000.000 Å diameter 0.13 cm3/g 0.17 cm3/g 0.30 cm3/g Adsorption average pore width (4 V/A by BET) 79 nm 49 nm 5 nm BJH Adsorption average pore diameter (4 V/A) 63 nm 54 nm 5 nm BJH Desorption average pore diameter (4 V/A) 48 nm 43 nm 5 nm Equilibrium isothermal studies Equlibrium isothermal studies are conducted to determine the capacity for adsorption of the given sorbent. A standard isothermal graph was plotted between equilibrium fluoride concentration in solution (Ce) and fluoride sorbed onto sorbent at equilibrium (qe) for initial fluoride concentrations ranging from 5 to 70 mg/L. Figure 6, shows the standard isothermal plot for fluoride sorbed onto LIB and Fig. 7 shows the standard isotherm plot for adsorption of fluoride onto bauxite. Both the sorbents showed almost a similar trend of high fluoride uptake at lower concentrations and with progressive increase in concentration of fluoride, the rate of fluoride uptake gradually decreased probably due to exhaustion of active sorption sites on the sorbents. Results of the experimental data were modeled using Langmuir and Freundlich isothermal models, to arrive at the best fitting model.Fig. 6 Standard isotherm plot of sorption of fluoride onto LIB (Fluoride concentration from 5 to 70 mg/L) Fig. 7 Standard isotherm plot of sorption of fluoride onto bauxite (Fluoride concentration from 5 to 70 mg/L) Langmuir and freundlich isotherm models Langmuir isotherm assumes monolayer coverage of adsorbate on sorbent. The linear form of Langmuir isotherm model is given in Eq. (4) (Langmuir 1916). 4 Ce/qe=1/qmax·b+Ce/qmax where Ce (mg/L) is concentration of fluoride in solution at equilibrium, qe (mg/g) is amount of fluoride sorbed onto the sorbent at equilibrium, qmax (mg/g) is maximum adsorption capacity and b (L/mg) is a constant related to energy. Freundlich isotherm model is based on the assumption that the surface of the sorbent is heterogeneous, with different sorption sites possessing different energies of sorption (Freundlich 1906). The linear form of Freundlich equation is given by Eq. (5). 5 logqe=logkf+1/nlogCe where kf represents relative adsorption capacity and n represents the intensity of sorption. The value of (1/n) suggests the nature of sorption. Value of 1/n greater than one suggests physisorption, whereas its value <1 suggests chemisorption (Freundlich 1906). Table 4 shows the values of all parameters of Freundlich model. It can be observed that the value of (1/n) is <1 suggesting chemisorption. From Table 4, it can be inferred that the best fitting isothermal model is Langmuir model, based on closeness of R2 value to 1. Also the predicted adsorption capacity of LIB is 18.18 mg/g, which is close to experimental value (17.2 mg/g). Similarly, in case of bauxite, the predicted and experimental values of maximum adsorption capacity are 7.72 and 3.2 mg/g respectively. The maximum adsorption capacities of various sorbents such as bauxite (Sujana and Anand 2011), lanthanum modified activated alumina (Cheng et al. 2014), lanthanum impregnated chitosan flakes (Jagtap et al. 2011), lanthanum incorporated chitosan beads (Bansiwal et al. 2009), metallurgical grade alumina (Pietrelli 2005),mixed rare earth oxides (Raichur and Basu 2001) and titanium rich bauxite (Das et al. 2005) along with the presently investigated sorbents are given in Table 5, for comparison. It can be observed from this table that the adsorption capacity of LIB is higher than all other sorbents except lanthanum hydroxide, which has an exceptionally high value of 242.2 mg/g. Similarly sorption capacity of bauxite used in this study has higher value of 7.7 mg/g compared to sorption capacity of bauxite obtained by Sujana and Anand (2011), probably due to calcination.Table 4 Comparison of isothermal constants for adsorption of fluoride onto LIB and bauxite Isotherm model Langmuir Freundlich LIB Isotherm parameters qmax = 18.18 mg/g b = 0.541 R2 = 0.997 Kf = 5.794 mg/g n = 2.753 R2 = 0.805 Bauxite Isotherm parameters qmax = 7.722 mg/g b = 0.379 R2 = 0.992 Kf = 1.902 mg/g n = 2.085 R2 = 0.861 Table 5 Comparison of isothermal constants for adsorption of fluoride onto LIB and bauxite Adsorbent qmax(mg/g) References Bauxite 5.16 Sujana and Anand (2011) Titanium rich bauxite 4.13 Das et al. (2005) Lanthanum hydroxide 242.2 Na and Park (2010) Lanthanum modified activated alumina 6.7 Cheng et al. (2014) Lanthanum impregnated chitosan flakes 1.27 Jagtap et al. (2011) Lanthanum incorporated chitosan beads 4.7 Bansiwal et al. (2009) Alumina metallurgical grade 12.57 Pietrelli (2005) Magnesia amended activated alumina 10.12 Maliyekkal et al. (2008) Mixed rare earth oxides 12.5 Raichur and Basu (2001) Lanthanum impregnated bauxite 18.18 Present study Bauxite 7.722 Present study Influence of pH pH plays a significant role in adsorption. pH studies were conducted for removal of fluoride from water using LIB and bauxite by adjusting the starting pH of solution in the pH range 3–12 and the results are presented in Fig. 8. It can be observed that for LIB the optimum fluoride removal was from pH 6.5 to 8.5, whereas for bauxite the optimum removal of fluoride was between 5.0 and 6.5. Beyond this range either at lower or higher pH values fluoride removal was observed to be significantly less. This pH range coincides with the pH range observed for removal of fluoride using titanium rich bauxite (Pietrelli 2005). It was observed that the equilibrium pH of solution increased by about 0.2–0.4 U than the initially adjusted pH, for both sorbents after adsorption at pH < 8. From pH above 8.0 the pH drift was not observed. As the drift in pH was observed only in the optimal fluoride removal pH range for both sorbents, which is eventually the natural occurring pH of water, the impact of drift was neglected. At low pH values H+ ions predominate and so could form a bond with fluoride forming HF (Maliyekkal et al. 2008). pKa of HF is 3.16 (Shen and Schafer 2014). When pH value of solution is lesser than 3.16, weakly charged HF ions predominate, which may not get readily adsorbed onto sorbent. However, when the value of pH is increased higher than 3.16, F− ions predominate, which get readily adsorbed. So, a gradual increase in removal of fluoride is observed after this pH value. Also, below pH value of 5, metallic salts present in bauxite, dissociate into their respective cations such as Al3+ and dissolve into solution (Adams 1999), thereby decreasing the active sorption sites on the surface of bauxite. So, removal of fluoride by bauxite is less below pH 5. Similarly below pH 7, lanthanum exists as La3+ ion (Cetiner and Xiong 2008), which dissociates itself from LIB causing a reduction in the active sites of sorption on LIB. However, when pH of solution containing bauxite is raised to more than 5 and that of LIB to more than 7, the cations such as Al3+ and La3+, get converted to their respective salts and stay on the sorbents, thereby aiding in adsorption. So, removal of fluoride, using bauxite increases from pH 5 onwards and removal of fluoride using LIB increases from pH 7 onwards. However further investigation is warranted to know the combined influence of dissolution of all the metal ions which influence sorption of fluoride at various pH values in lower pH range. At higher pH values OH− ions predominate which compete for active sorption sites on the surface of sorbent with F− ions. Thus due to the phenomena of competitive adsorption, sorption of fluoride ions onto the surface of sorbent could decrease. Also the pHzpc of LIB and bauxite were found to be 8.2 ± 0.1 and 6.0 ± 0.1 respectively. This further justifies the optimal sorption of fluoride in the observed ranges. The charge on the surface of a sorbent remains positive from lower values of pH to the value of pHzpc. From this point onwards any further increase in pH of solution changes the surface charge of sorbent from positive to negative.Fig. 8 Influence of pH on defluoridation by LIB and bauxite Influence of anions Groundwater may consists of several anions such as Cl−, SO42−, PO43−, HCO3− and NO3− in addition to fluoride, which might compete with fluoride for sorption onto the active sites on the sorbent (Fink and Lindsay 1936). This might reduce the sorption of fluoride onto LIB. So, the above mentioned ions were added each individually in concentrations up to 100 mg/L along with fluoride of concentration 20 mg/L in distilled water and sorption experiments were conducted. The impact of the anions tested here on sorption of fluoride onto LIB is depicted in Fig. 9. It can be observed that except nitrates, addition of other ions increased the final fluoride concentrations after adsorption to more than 1.5 mg/L. As lanthanum is predominant in LIB, it could possibly have less affinity for nitrates compared to other ions.Fig. 9 Influence of competing anions on sorption of fluoride onto LIB (Initial fluoride = 20 mg/L, Sorbent dose = 2 g/L) Regeneration studies To determine the potential of re-applicability of spent sorbent, regeneration studies were conducted. Initially adsorption experiments were conducted with 20 mg/L, aqueous fluoride solution using LIB, under optimum conditions. Loaded LIB was agitated with distilled water and the desorption was found to be very low (36 %). Therefore desorption studies were carried out using various eluents such as NaOH and H2SO4. It was observed that NaOH could elute fluoride from the loaded sorbent successfully. The influence of various concentrations of NaOH in eluting the sorbent loaded to capacity with LIB is presented in Fig. 10. A 4 % NaOH solution could elute nearly 95 % of fluoride. This could be due to exchange of OH ion in NaOH with F− ion, as given in Eq. 6. 6 MF+NaOH→MOH+NaF Fig. 10 Effect of NaOH concentration on the fluoride desorption The regenerated sorbent thus obtained was again used as a sorbent, and subsequently eluted with NaOH. Such cycles were repeated and the maximum removal of fluoride after each cycle is presented in Fig. 11. It can be observed that as against predicted 0.7 mg/L of residual fluoride, the residual fluoride after first cycle was found to be 1 mg/L and it rose to 1.4 mg/L after 3 cycles. Further, after 4th cycle, the residual fluoride was found to be 1.6 mg/L which exceeds the permissible limit. So, 3 cycles of regeneration can be considered safe for LIB. However, this is not quite in agreement with 5 % of fluoride residing on sorbent after elution after each cycle. Further detailed experimentation is warranted to understand the exact reason underlying this. As only 90 % of fluoride is eluted after each cycle, the defluoridation capacity decreases after each cycle. Also with each cycle of elution with NaOH, OH ions accumulation on adsorbent increases which reduces its defluoridation capacity. It can be observed that the regenerated sorbent can be used successfully up to 4 cycles without significant loss, in the sorption potential.Fig. 11 Number of cycles of defluoridation by LIB after regeneration Conclusions In this study a novel sorbent, LIB was prepared and investigated for removal of fluoride from water. LIB exhibited good sorption potential of 18.18 mg/g, compared to several other sorbents, collected from literature. However, further investigation is warranted to find the exact sorption capacity of LIB under natural groundwater conditions. The main conclusions are LIB was prepared by thermal impregnation method, and it reduced fluoride from distilled water from 20 to 0.7 mg/L. LIB at a dose of 2 g/L removed fluoride up to 99 % from an initial concentration of 20 mg/L of aqueous fluoride solution, whereas bauxite at a dose of 6 g/L removed up to 94 % of fluoride from an initial concentration of 20 mg/L of aqueous fluoride solution. The time taken for defluoridation by LIB was 120 min and it followed pseudo second order reaction, which indicates that the mechanism involved could be chemisorption. Pore diffusion seems to be the rate limiting step. The time taken for defluoridation by bauxite was 150 min. The sorption process by LIB conformed to Langmuir isotherm model. The maximum sorption capacity according to this model was 18.18 mg/g, which was close to observed experimental values. Bauxite followed a similar trend with a best fit to Langmuir isotherm model. The maximum sorption capacity of bauxite was found to be 7.722 mg/g. A pH range of 6.5–8.5 was found to be optimum for LIB, for removal of fluoride from water, which is a naturally occurring pH for waters. Bauxite exhibited optimum fluoride removal from pH 5.0 to 6.5. Addition of NO3− to aqueous fluoride solution water brought the residual fluoride concentration after sorption to 1.4 mg/L, whereas other individual ions added such as Cl−, SO42−, PO43− and HCO3− caused the final fluoride concentration after sorption to be more than 1.5 mg/L, probably due to competition of ions. 4 % NaOH regenerated LIB by 95 % and the effective number of cycles after regenerations were found to be 3 for removal of fluoride up to permissible limit. Authors’ contributions VVCM and SM conceived and designed this study. VVCM mainly and SM partly performed the experiments. Both authors read and approved the final manuscript. Acknowledgements The authors profusely thank and acknowledge the financial assistance in the form of waiver of Article Processing Charges, offered by Springer Open Waivers and Biomedcentral Waivers, towards publishing this paper. Competing interests The authors declare that they have no competing interests. ==== Refs References Adams ML (1999) Speciation and measurement of aluminium in environmental systems. Ph.D Thesis, University of Canterbury Adhikary SK Tipnis UK Harkare WP Govindum KP Defluoridation duringdesalination of brakish water by electrodialysis Desalination 1989 71 301 312 10.1016/0011-9164(89)85031-3 APHA, AWWA, WEF (1996) Standard methods for the examination of water and wastewater, 19th edn. Washington, DC, p 657 Atasoy AD Yesilnacar MI Sahin MO Removal of fluoride from contaminated ground water using raw and modified bauxite Bull Environ Contam Toxicol 2013 91 5 595 599 10.1007/s00128-013-1099-z 24072259 Ayoob S Gupta AK Bhat VT A conceptual overview on sustainable technologies for the defluoridation of drinking water Crit Rev Environ Sci Technol 2008 38 6 401 470 10.1080/10643380701413310 Bansiwal A Thakre D Labhshetwar N Meshram S Rayalu S Fluoride removal using lanthanum incorporated chitosan beads Colloids Surf B Biointerfaces 2009 74 216 224 10.1016/j.colsurfb.2009.07.021 19699619 Bhatnagar A Kumar E Sillanpää M Fluoride removal from water by adsorption: a review Chem Eng J 2011 171 3 811 840 10.1016/j.cej.2011.05.028 Biswas K Gupta K Ghosh UC Adsorption of fluoride by hydrous iron(III)–tin(IV) bimetal mixed oxide from the aqueous solutions Chem Eng J 2009 149 196 206 10.1016/j.cej.2008.09.047 Boruff CS Removal of fluorides from drinking water Ind Eng Chem 1934 26 69 71 10.1021/ie50289a016 Bureau of Indian Standards: IS 10500-2012 Drinking water specifications 2012 New Delhi Bureau of Indian Standards Carabineiro SAC Thavorn-Amornsri T Pereira MFR Figueiredo JL Adsorption of ciprofloxacin on surface-modified carbon materials Water Res 2011 45 4583 4591 10.1016/j.watres.2011.06.008 21733541 Cetiner ZS Xiong Y Chemical controls on the solubility, speciation and mobility of lanthanum at near surface conditions: a geochemical modeling study Appl Geochem 2008 23 2301 2315 10.1016/j.apgeochem.2008.04.014 Cheng J Meng X Jing C Hao J La3+ -modified activated alumina for fluoride removal from water J Hazard Mater 2014 278 343 349 10.1016/j.jhazmat.2014.06.008 24996152 Chinoy NJ Effects of fluoride on physiology of animals and human beings Indian J Environ Toxicol 1991 1 17 32 Czarnowski W Wrzesniowska K Krechniak J Fluoride in drinking water and human urine in Northern and Central Poland Sci Total Environ 1996 191 177 184 10.1016/0048-9697(96)05259-X 8885430 Das N Pattanaik P Das R Defluoridation of drinking water using activated titanium rich bauxite J Colloid Interface Sci 2005 292 1 10 10.1016/j.jcis.2005.06.045 16126217 Emamjomeh MM Sivakumar M An empirical model for defluoridation by batchmonopolar electrocoagulation/flotation (ECF) process J Hazard Mater 2006 131 118 125 10.1016/j.jhazmat.2005.09.030 16298054 Everett DH IUPAC manual appendix II, Part I Pure Appl Chem 1973 31 579 Everett DH IUPAC manual appendix II, Part II: heterogeneous catalysis Pure Appl Chem 1976 31 71 Feng L Xiao H He X Li Z Li F Liu N Zhao Y Huang Y Zhang Z Chai Z Neurotoxicological consequence of long-term exposure to lanthanum Toxicol Lett 2006 165 112 120 10.1016/j.toxlet.2006.02.003 16542800 Fink GJ Lindsay FK Activated alumina for removing fluorides from drinking water Ind Eng Chem 1936 80 947 948 10.1021/ie50320a017 Freundlich HMF Uber die adsorption in losungen Z Phys Chem 1906 57 385 470 Gaciri SJ Davies TC The occurrence and geochemistry of fluoride in somenatural waters of Kenya J Hydrol 1993 143 395 412 10.1016/0022-1694(93)90201-J Hach company/HachLange GmbH, manual on fluoride (1989–2014) Ho YS McKay G Pseudo-second order model for sorption processes Process Biochem 1999 34 451 465 10.1016/S0032-9592(98)00112-5 Jagtap S Yenkie MK Das S Rayalu S Synthesis and characterization of lanthanum impregnated chitosan flakes for fluoride removal in water Desalination 2011 273 267 275 10.1016/j.desal.2010.12.032 Khatibikamala V Torabian A Janpoor F Hoshyaripour G Fluoride removal from industrial wastewater using electrocoagulation and its adsorption kinetics J Hazard Mater 2010 179 276 280 10.1016/j.jhazmat.2010.02.089 20347217 Lagergren S About the theory of so-called adsorption of soluble substances Kungliga Svenska Vetenskapsakademiens Handlingar 1898 24 1 39 Langmuir I The adsorption of gases on plane surface of glass, mica andplatinum J Am Chem Soc 1916 40 1361 1403 10.1021/ja02242a004 Maliyekkal SM Shukla S Philip L Nambi IM Enhanced fluoride removal from drinking water by magnesia-amended activated alumina granules Chem Eng J 2008 140 183 192 10.1016/j.cej.2007.09.049 Na CK Park HJ Defluoridation from aqueous solution by lanthanum hydroxide J Hazard Mater 2010 183 512 520 10.1016/j.jhazmat.2010.07.054 20702038 Nagendra Rao CR Karthikeyan J Removal of fluoride from water by adsorption onto lanthanum oxide Water Air Soil Pollut 2012 223 1101 1114 10.1007/s11270-011-0928-0 Nawlakhe WG Kulkarni DN Pathak BN Bulusu KR Defluoridation of water by Nalgonda technique Indian J Environ Health 1978 17 1 26 Nemade PD Rao AV Alappat BJ Removal of fluorides from water using low cost adsorbents Water Sci Technol Water Supply 2002 2 311 317 Oliveira EA Montanher SF Andrade AD Nobrega JA Rollemberg MC Equilubrium studies for the sorption of chromium and nickel from aqueous solutions using raw rice bran Process Biochem 2005 40 3485 3490 10.1016/j.procbio.2005.02.026 Pietrelli L Fluoride wastewater treatment by adsorption onto metallurgical grade alumina Anal Chim 2005 95 303 312 10.1002/adic.200590035 Raichur AM Basu MJ Adsorption of fluoride onto mixed rare earth oxides Sep Purif Technol 2001 24 121 127 10.1016/S1383-5866(00)00219-7 Rivera-Utrilla J Bautista-Toledo I Ferro-García MA Moreno-Castilla C Activated carbon surface modifications by adsorption of bacteria and their effect on aqueous lead adsorption J Chem Technol Biotechnol 2001 76 1209 1215 10.1002/jctb.506 Shen J Schafer A Removal of fluoride and uranium by nanofiltration and reverse osmosis: a review Chemosphere 2014 117 679 691 10.1016/j.chemosphere.2014.09.090 25461935 Simons R Trace element removal from ash dam waters by nanofiltration anddiffusion dialysis Desalination 1993 89 325 341 10.1016/0011-9164(93)80145-D Srimurali M Pragathi A Karthikeyan J A study on removal of fluorides from drinking water by adsorption onto low-cost materials Environ Pollut 1998 99 285 289 10.1016/S0269-7491(97)00129-2 15093323 Sujana MG Anand S Fluoride removal studies from contaminated groundwater by using bauxite Desalination 2011 267 222 227 10.1016/j.desal.2010.09.030 Swope GH Hess RH Removal of fluoride from natural water by defluoridation Ind Eng Chem 1937 29 424 426 10.1021/ie50328a015 Vijaya Kumar V Sai CST Rao PLKM Rao CS Studies on the distribution of fluoride in drinking water sources in Medchal Block, Ranga Reddy District, Andhra Pradesh, India J Fluor Chem 1991 55 3 229 236 10.1016/S0022-1139(00)82350-7 Vivek Vardhan CM Karthikeyan J Removal of fluoride from water using low-cost materials Int Water Technol J 2011 2 1 12 Vivek Vardhan CM Srimurali M Defluoridation of drinking water using a novel sorbent: lanthanum-impregnated green sand Desalination Water Treat 2016 Weber WJ Morris JC Intraparticle diffusion during the sorption of surfactants onto activated carbon J Sanit Eng Div AESC 1963 89 53 61 WHO (2004) Guidelines for drinking water quality. Geneva Xu Y-M Ning A-R Zhao J Preparation and defluorination performance ofactivated cerium(IV) oxide/SiMCM-41 adsorbent in water J Colloid Interface Sci 2001 235 66 69 10.1006/jcis.2000.7344 11237443 Yeşilnacar Mİ Yetiş AD Dülgergil ÇT Kumral M Atasoy AD Dogan TR Tekiner Sİ Bayhan İ Aydoğdu M Geomedical assessment of an area having high-fluoride groundwater in southeastern Turkey Environ Earth Sci 2016 75 162 10.1007/s12665-015-5002-6 Yousef RI El-Eswed B Al-Muhtaseb A Adsorption characteristics of natural xeolites as solid adsorbents for phenol removal from aqueous solution: kinetics, mechanism, and thermodynamics studies Chem Eng J 2011 171 1143 1149 10.1016/j.cej.2011.05.012

PMC005xxxxxx/PMC5001967.txt

==== Front SpringerplusSpringerplusSpringerPlus2193-1801Springer International Publishing Cham 311010.1186/s40064-016-3110-8ResearchStructural connectivity profile of scans without evidence of dopaminergic deficit (SWEDD) patients compared to normal controls and Parkinson’s disease patients Kim Mansu mansooru@skku.edu 1http://orcid.org/0000-0001-5681-8918Park Hyunjin +82-31-299-4956hyunjinp@skku.edu 231 Department of Electronic Electrical and Computer Engineering, Sungkyunkwan University, Suwon, Korea 2 School of Electronic and Electrical Engineering, Sungkyunkwan University, Suwon, Korea 3 Center for Neuroscience Imaging Research (CNIR), Institute for Basic Science, Suwon, Korea 26 8 2016 26 8 2016 2016 5 1 142131 3 2016 19 8 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Background In this study, we investigated the structural connectivity profile of patients with scans without evidence of dopaminergic deficit (SWEDD) compared with normal controls (NC) and patients with Parkinson’s disease (PD). An accurate understanding of SWEDD is important so that appropriate therapeutic options can be presented to patients. Methods Diffusion magnetic resonance imaging of NC (n = 40), SWEDD (n = 40) and PD patients (n = 40) was obtained from a research database. Tractography, the process of obtaining fiber information was performed. Connectivity analysis was performed on 16 connections in the cortico-basal ganglia-thalamo-cortical circuit. Group-wise differences among NC, PD and SWEDD patients were quantified in terms of structural connectivity based on fiber density. Then, we investigated correlations with the clinical score using the Movement Disorder Society-Sponsored Unified Parkinson’s Disease Rating Scale (MDS-UPDRS). A support vector machine classifier and leave-one-out cross-validation were applied to separate the NC, SWEDD and PD groups. Results Pallidum–putamen and sensorimotor cortex–putamen connections showed significant group-wise differences among NC, PD and SWEDD patients and correlated with the MDS-UPDRS score. Conclusions Pallidum–putamen and sensorimotor cortex–putamen connections might form a structural connectivity profile unique to SWEDD and could be a potential imaging biomarker for future movement disorder research. Keywords Scans without evidence of dopaminergic deficitDiffusion tensor imagingTractographyCorrelation with clinical scoreInstitute for Basic ScienceIBS-R015-D1Park Hyunjin http://dx.doi.org/10.13039/501100003725National Research Foundation of KoreaNRF-2016R1A2B4008545Park Hyunjin issue-copyright-statement© The Author(s) 2016 ==== Body Background Parkinson’s disease (PD) is a common neurodegenerative disorder characterized by motor symptoms and cognitive impairment (Stoessl 2011). Diagnosis of PD is typically performed according to the criteria from the United Kingdom’s Parkinson’s Disease Society Brain Bank. PD is associated with a loss of dopaminergic neurons in the substantia nigra (SN) that project to the striatum (Obeso et al. 2008). This dopamine imbalance causes inhibition of basal ganglia output and dysfunction within cortico-basal ganglia-thalamo-cortical circuits (CBGT) (Obeso et al. 2008; Sharman et al. 2013). Functional neuroimaging techniques such as 18F dopa positron emission tomography or dopamine transporter single-photon emission computed tomography (DaT-SPECT) are adopted to assess dopaminergic dysfunction in PD patients. PD patients showed significantly reduced striatal uptake compared with normal controls (NC) using SPECT (Tolosa et al. 2006). However, approximately 10 % of clinically diagnosed PD patients have normal dopaminergic functional imaging and classified as having scans without evidence of dopaminergic deficit (SWEDD) (Schneider et al. 2007). To date, no consensus regarding the etiology of SWEDD exists. Some researchers consider it an early phase of PD, while others argue that it is very different from PD (Batla et al. 2014). Recent studies showed that abnormalities in cortical plasticity differed between PD and SWEDD patients (Schwingenschuh et al. 2010). Therapeutic options differ between SWEDD and PD because SWEDD patients are relatively insensitive to levodopa therapy (Fahn and Group 2005). Correctly understanding SWEDD is important so that appropriate therapeutic options can be presented to patients. Diffusion tensor imaging (DTI) is a tensor based model of diffusion weighted magnetic resonance imaging (MRI) technique that can provide in vivo information on the microstructural integrity of brain tissue using anisotropic water diffusion. DTI data are processed with an algorithm known as tractography to perform the reconstruction of large white matter tracts. The processed fiber information is analyzed using connectivity analysis, which considers the brain as a complex network. Connectivity derived from DTI is known as structural connectivity. Various MRI techniques including resting state functional MRI (rs-fMRI) and DTI were applied to compare PD patients and NC (Kim et al. 2013; Sharman et al. 2013; Shu et al. 2011; Wu et al. 2009; Yu et al. 2013; Zhang et al. 2015; Ziegler et al. 2014). These studies reported PD related brain alterations compared with NC using rs-fMRI, DTI and track-based spatial statistics (Sharman et al. 2013; Zhang et al. 2015; Kim et al. 2013). A recent study investigated PD and SWEDD patients using structural connectivity and found four structural connections to explain a clinical score (Kim and Park 2016). That study explored the whole brain regions of interest (ROIs), while our study focused on regions of CBGT. The previous study adopted number of fibers but this study adopted a more refined measure of fiber density. The prior study considered PD and SWEDD, while this study considered three groups (i.e., PD, SWEDD and NC) to better characterize group differences. In this study, the main objective was to investigate structural connectivity profile of SWEDD compared with NC and PD. A secondary objective was to investigate correlation between structural connectivity results with clinical scores of Movement Disorder Society-Sponsored Unified Parkinson’s Disease Rating Scale (MDS-UPDRS). We obtained diffusion MRI data from a research database, the Parkinson’s Progression Markers Initiative (PPMI) (Marek et al. 2011). Connectivity analysis was applied to brain regions of CBGT circuitry (Obeso et al. 2000, 2008). Group-wise differences among SWEDD, NC and PD patients were assessed in terms of structural connectivity based on fiber density. Significant connectivity differences were identified and further investigated with the MDS-UPDRS scores. Briefly, we identified structural connectivity profiles unique to SWEDD patients and found the identified connections were significantly correlated with MDS-UPDRS scores. Methods Subjects This study was a retrospective analysis of anonymized imaging data and was approved by the Institutional Review Board (IRB) of Sungkyunkwan University. Our study did not require participant’s consent as we analyzed anonymized data. Participant data were anonymized and de-identified prior to analysis. The study included 120 participants classified into NC (n = 40), SWEDD (n = 40) and PD (n = 40) categories. The sub-groups were classified based on the criteria established by the PPMI consortium (Marek et al. 2011). Detailed criteria for the SWEDD group follows. First, the patients must have at least two of the following symptoms: (1) resting tremor, (2) bradykinesia, (3) rigidity (must have either resting tremor or bradykinesia). Patients were also included if they had asymmetric tremor or asymmetric bradykinesia. Second, patients were diagnosed of PD for 2 years or less at the time of screening with confirmation for no evidence of dopamine transporter deficit using DaT-SPECT imaging. Third, patients were not expected to require PD medication within at least 6 months from baseline. Fourth, patients were 30 years or older at the time of PD diagnosis. The age and sex ratios of each group were matched as shown in the Table 1. Details regarding the subjects, including MDS-UPDRS scores, are also shown in Table 1.Table 1 Participant information NC SWEDD PD p value (NC-SWEDD/SWEDD-PD/NC-PD) Number of subjects n = 40 n = 40 n = 40 – Age (years) (mean ± SD) 60.92 ± 10.59 60.73 ± 10.72 61.86 ± 8.81 0.94/0.61/0.67 Sex (male/female) 24/16 26/14 26 /14 0.65/1/0.65 Disease duration (month) (mean ± SD) – 6.41 ± 8.03 7.37 ± 7.79 <0.05/0.48/<0.05 MDS-UPDRS scores (mean ± SD) 0.57 ± 1.37 13.15 ± 9.17 22.67 ± 9.17 <0.05/<0.05/<0.05 Values are reported as the mean ± standard deviation (SD) Imaging data We obtained diffusion-weighted and T1- and T2-weighted MRI data from the PPMI database (Marek et al. 2011). Diffusion images were acquired with a Siemens 3T scanner using the following parameters: 3T scanner, b = 1000 s/mm2, 64 diffusion gradient directions with 1 b0 image, image matrix = 116 × 116 × 72 and voxel resolution = 1.98 × 1.98 × 2 mm3. The subjects underwent T1- and T2-weighted MRI as well as image pre-processing steps required in addition to DTI data acquisition. Acquisition parameters for the T1-weighted images were as follows: TR = 2300 ms, TE = 2.98 ms, TI = 900 ms, image matrix = 240 × 256 × 176 and voxel resolution = 1 × 1 × 1 mm3. The parameters for the T2-weighted images were: TR = 3000 ms, TE = 101 ms, image matrix = 228 × 256 × 48 and voxel resolution = 0.9375 × 0.9375 × 3 mm3. Image pre-processing Pre-processing steps were required to extract fiber information from the DTI data. Excellent review articles on this procedure exist, and thus, only a brief summary is given below (Daducci et al. 2012). Image pre-processing was performed using the Connectome Mapping Toolkit (CMTK), a Python-based open-source software (www.cmtk.org) (Daducci et al. 2012). For each subject, distortions caused by eddy currents and simple head motion during scans were corrected using FSL’s Eddy current tool and MCFLIRT (Smith et al. 2004). Then, the T1-, T2- and diffusion-weighted images were aligned to the Montreal Neurological Institute space with non-linear registration using FSL (Smith et al. 2004). The registered T1-weighted image was segmented into white matter, grey matter and cerebrospinal fluid using Freesurfer (Fischl 2012). The segmented white matter was later used to guide the tractography algorithm. The complete image pre-processing procedures are provided in Fig. 1.Fig. 1 Image pre-processing procedures ROI specifications Connectivity analysis requires that ROIs are specified so that correlations among them can be investigated. Our analysis focused on the CBGT circuit, which consisted of eight regions: caudate, putamen, pallidum, thalamus, SN, sensorimotor cortex, associative cortex and limbic cortex, as shown in Fig. 2. The sensorimotor circuit, which includes the pre-central, post-central and para-central gyrus (Brodmann areas 1–5), is related to motor symptoms in PD (Albin et al. 1989). The associative circuit, which includes the dorsolateral prefrontal, middle and superior frontal cortex (Brodmann areas 8, 9, 44–47), is concerned with executive function and is affected by age and PD (Leh et al. 2010). The limbic circuit, which includes the medial temporal, orbitofrontal, posterior and anterior cingulate cortex, insula, entorhinal, hippocampus and amygdala, is related to stuttering and other movement disorders (Purves et al. 2008). All eight regions are structurally connected to the striatum (Parent and Hazrati 1995). Subcortical structures, caudate, putamen, pallidum and thalamus were specified by segmentation results from the registered T1-weighted images using Freesurfer (Fischl 2012). The SN region was specified by transferring ROI information from a pre-defined atlas via image co-registration (Keuken et al. 2014). The co-registration mapped the atlas information onto the subject’s image space so that both images reside on the same spatial framework. Three cortical structures, the sensorimotor, associative and limbic regions, were specified by the Desikan–Killiany anatomical atlas (Desikan et al. 2006). Fig. 2 ROI specifications. A total of eight ROIs were defined from T1 and T2-weighted images, including the associative, limbic, sensorimotor cortex, caudate, putamen, pallidum, thalamus and substantia nigra regions White matter tractography Fibers were constructed using tractography implemented using the Diffusion Toolkit (Wang et al. 2007). Fractional anisotropy (FA) of each voxel was computed using diffusion tensor data. Tractography was performed using the fiber assignment by continuous tracking (FACT) algorithm implemented in the Diffusion Toolkit to reconstruct all of the brain fibers (Wang et al. 2007; Mori and van Zijl 2002). The FACT algorithm propagated a line from the center of a seed voxel along the direction of the dominant vector, which was determined by the largest eigenvector of the tensor until the streamline exited to the next voxel. The starting point of the next voxel was the intercept of the previous voxel. Seed voxels were all voxels in the eight ROIs and connections were retained only if the seeds reached the one of the remaining ROIs. The tractography terminated when the algorithm entered a region with an abrupt change in fiber direction angle more than 60° and was limited to white matter regions and their neighbors, as fibers mainly exist in white matter. Illustrations of tracked fibers of the three significant connections for representative PD patients and NC are shown in Fig. 3.Fig. 3 Illustration of tractography. The tracked fibers of three significant connections for representative PD patients and NC. PD cases are listed in the first column and NC cases are listed in the second column. Colored regions denote different ROIs. Fibers are displayed in streamlines. Top sub-figures associative cortex–thalamus connection. Middle sub-figures Sensorimotor cortex–putamen connection. Bottom sub-figures Pallidum–putamen connection Structural network construction Structural connectivity was assessed using 8 regions as nodes and 16 connections of interest as edges of a graph. There were 28 (=8 choose 2) possible connections between 8 regions and only 16 connections of the CBGT circuits were considered, as shown in the first column of Table 2. We applied a threshold of five fibers so that only structural connections with more than five tracked fibers connecting two regions were considered. This approach was chosen to reduce the chance of falsely identifying connections and thus leading to a more robust analysis (Shu et al. 2009, 2011; Im et al. 2014). The nodes were brain ROIs in the CBGT circuit, as described previously. Edge values were structural connectivity values between nodes. Structural connectivity values were defined as the fiber density, defined as the product of the connection density (CD) and the connection efficacy (CE) of the fibers (Hagmann et al. 2010). Our measure of CD incorporated the number of fibers connecting the regions and then normalized for the area of the brain regions and length of the fiber connecting two regions (Fischi-Gómez et al. 2015; Hagmann et al. 2008). The CD between two regions was defined as follows: CDi,j=2Si+Sj∑f∈fibers1lf, where f is the fiber connection between regions i and j, lf is the length of the fiber connection and Si is the surface area of region i. Mean FA value along the fiber connections was used as the CE measure and FA is related to the fiber integrity. FA values reflect the degree of anisotropic water diffusion and is influenced by axonal myelination and the diameter and has a high correlation with conductivity (Tuch et al. 2001; Hagmann et al. 2010; Fischi-Gómez et al. 2015; Feldman et al. 2010). Many other studies also adopted FA to assess CE (Shu et al. 2011; Fischi-Gómez et al. 2015; Zhang et al. 2011; Schwingenschuh et al. 2010; Wen et al. 2011). The edge values were entered into a matrix, whose elements were fiber density values. We adopted a simple network model that considered undirected and weighted edges. The matrix was referred to as the connectivity matrix in this study.Table 2 Structural connectivity values for the three groups Connections Structural connectivity (mean ± SD) Corrected p value NC SWEDD PD PD-NC SWEDD-NC SWEDD-PD Associative cortex–caudate 0.671 ± 1.381 0.299 ± 0.453 0.232 ± 0.319 0.017 0.054 0.214 Associative cortex–putamen 3.452 ± 1.561 3.471 ± 1.517 3.734 ± 1.771 0.228 0.487 0.232 Associative cortex–thalamus 0.833 ± 0.488 0.650 ± 0.367 1.173 ± 0.750 0.009 0.032 <0.001 Limbic cortex–caudate 2.197 ± 1.463 1.902 ± 1.310 0.710 ± 1.215 <0.001 0.162 <0.001 Limbic cortex–putamen 3.175 ± 1.272 3.271 ± 1.474 2.557 ± 1.147 0.013 0.387 0.008 Limbic cortex–thalamus 1.535 ± 0.810 1.379 ± 0.866 1.054 ± 0.889 0.005 0.194 0.048 Sensorimotor cortex–caudate 0.430 ± 0.993 0.170 ± 0.471 0.053 ± 0.076 <0.001 0.062 0.016 Sensorimotor cortex–putamen 1.250 ± 0.742 1.485 ± 0.930 1.728 ± 1.108 0.011 0.010 0.027 Sensorimotor cortex–thalamus 0.964 ± 0.574 0.947 ± 0.537 1.019 ± 0.703 0.325 0.511 0.292 Pallidum–caudate 0.676 ± 0.644 0.838 ± 0.744 0.631 ± 0.825 0.374 0.145 0.112 Pallidum–putamen 2.166 ± 0.898 1.840 ± 0.917 1.283 ± 0.522 <0.001 0.045 <0.001 Pallidum–thalamus 0.799 ± 0.519 0.926 ± 0.523 0.565 ± 0.452 0.007 0.143 <0.001 Putamen–thalamus 0.439 ± 0.437 0.477 ± 0.388 0.346 ± 0.298 0.132 0.295 0.044 Substantia nigra–putamen 0.004 ± 0.017 0.001 ± 0.004 0.004 ± 0.015 0.496 0.179 0.192 Substantia nigra–thalamus 0.001 ± 0.005 0.004 ± 0.012 0.009 ± 0.003 0.193 0.045 0.223 Substantia nigra–pallidum 0.096 ± 0.243 0.095 ± 0.363 0.160 ± 0.129 0.210 0.485 0.350 Structural connectivity values based on fiber density were reported for 16 connections within the CBGT circuit. The values were reported as the mean ± standard deviation (SD). Group-wise differences in NC-PD, SWEDD-NC and SWEDD-PD comparisons were reported using corrected p values. The italicised text shows identified connections with significant differences among the three comparisons Statistical tests Group-wise differences among the SWEDD, NC and PD groups were assessed by exploring 16 connections, which was equivalent to investigating 16 elements in the connectivity matrix. For each group, the connectivity matrices of participants were stacked into three-dimensional matrices. Each group had a single three-dimensional matrix. Each element in the stacked connectivity matrix contained 40 observations. We performed non-parametric permutation tests for 16 connections of interests to identify group-wise differences among the NC, SWEDD and PD groups. A permutation test is a non-parametric approach that does not require the estimated parameter to follow a normal distribution and has been widely adopted in neuroimaging research (Smith et al. 2013). We performed the permutation tests by randomly assigning NC, SWEDD and PD patients 10,000 times. One permutation involved randomly assigning the first 40 cases to the NC group, the next 40 cases to the PD group and the remaining 40 cases to the SWEDD group. Differences in structural connectivity were considered significant if they did not belong to the 95 % of the null distribution derived from the permutation tests (p < 0.05, corrected) (Nichols and Holmes 2002; Bullmore et al. 1999). Correlation with clinical scores Correlation analysis was performed to detect possible links between structural connectivity and clinical scores. We pooled connectivity matrices between the groups (i.e., NC, PD and SWEDD) into a single matrix and then calculated the Spearman correlations using the MDS-UPDRS scores for each element in the matrix. Multiple comparison issues were adjusted using Holm–Bonferroni correction which accounted for 16 pair-wise correlation analyses. Classification using identified connections The three significant connections were fed into a support vector machine (SVM) classifier framework with a quadratic kernel to separate the NC, SWEDD and PD groups. The technical details of SVM are found in review articles (Vapnik 1999). We applied the leave-one-out cross-validation method to distinguish training and test data, due to the limited number of subjects available. For example, given 40 NC and 40 PD cases, we assigned 1 case as the test case and used the remaining 79 cases as the training data for the SVM classifier. The process was repeated 80 times, choosing a different test case each time. The SVM classifier seeks a decision boundary that can effectively separate samples near the decision boundary. Classifier accuracy, sensitivity and specificity were computed by comparing the classifier outcomes with the known ground truth using MATLAB. The entire procedure was performed for the NC-PD, NC-SWEDD and PD-SWEDD classifications. Results Structural connectivity differences Structural connectivity results (i.e., fiber density values) for the three groups are reported in Table 2, which shows the mean and standard deviation (SD) for the fiber density values of 16 connections. The values were computed from both hemispheres. Overall, nine connections were significantly different between PD and NC. Of these connections, seven connections were smaller in PD compared with NC and two connections were smaller in NC compared with PD (p < 0.05, corrected). Four connections were significantly different between SWEDD and NC. Among these connections, two connections were smaller in NC and two connections were smaller in SWEDD (p < 0.05, corrected). Nine connections showed significant differences between PD and SWEDD. Among these connections, seven connections were smaller in PD and two connections were smaller in SWEDD (p < 0.05, corrected). PD patients showed significant connection differences compared with NC in associative cortex–caudate, associative cortex–thalamus, limbic cortex–caudate, limbic cortex–putamen, limbic cortex-thalamus, sensorimotor cortex–caudate, sensorimotor cortex–putamen, pallidum–putamen and pallidum–thalamus connections. SWEDD patients showed significant differences compared with NC in associative cortex–caudate, associative cortex–thalamus, sensorimotor cortex–putamen, pallidum–putamen and SN–thalamus connections. PD patients showed significant connection differences compared with SWEDD in associative cortex–thalamus, limbic cortex–caudate, limbic cortex–putamen, limbic cortex–thalamus, sensorimotor cortex–caudate, sensorimotor cortex–putamen, pallidum–putamen, pallidum–thalamus and putamen–thalamus connections. Associative cortex–thalamus, sensorimotor cortex–putamen and pallidum–putamen connections were commonly identified as significant in NC-PD, SWEDD-NC and SWEDD-PD comparisons, as shown in italic font in Table 2. Correlation between identified connections and clinical scores Correlation analysis was performed to identify possible links between structural connectivity and clinical score (i.e., MDS-UPDRS) for all 16 connections, as shown in Table 3. The correlation analysis results of the three previously identified connections are shown below. The pallidum–putamen connection showed a significant negative correlation between structural connectivity and clinical score (r = −0.352, corrected p = 0.001). We observed significant positive correlations between structural connectivity and clinical score (r = 0.280, corrected p = 0.014) for the sensorimotor cortex–putamen connection. No significant correlation for the associative cortex–thalamus connection was found (r = 0.088, corrected p = 1). In summary, two connections out of three showed a significant correlation with MDS-UPDRS score.Table 3 Correlation between the structural connectivity and MDS-UPDRS III score Connections Structural connectivity (mean ± SD) Corr coef. (corrected p value) NC SWEDD PD Total Associative cortex–caudate 0.671 ± 1.381 0.299 ± 0.453 0.232 ± 0.319 0.401 ± 0.874 −0.171 (0.61) Associative cortex–putamen 3.452 ± 1.561 3.471 ± 1.517 3.734 ± 1.771 3.552 ± 1.612 0.034 (1) Associative cortex–thalamus 0.833 ± 0.488 0.650 ± 0.367 1.173 ± 0.750 0.886 ± 0.595 0.088 (1) Limbic cortex–caudate 2.197 ± 1.463 1.902 ± 1.310 0.710 ± 1.215 1.603 ± 1.471 −0.404 (<0.001) Limbic cortex–putamen 3.175 ± 1.272 3.271 ± 1.474 2.557 ± 1.147 3.001 ± 0.132 −0.070 (1) Limbic cortex–thalamus 1.535 ± 0.810 1.379 ± 0.866 1.054 ± 0.889 1.323 ± 0.872 −0.239 (0.104) Sensorimotor cortex–caudate 0.430 ± 0.993 0.170 ± 0.471 0.053 ± 0.076 0.217 ± 0.638 −0.161 (0.711) Sensorimotor cortex–putamen 1.250 ± 0.742 1.485 ± 0.930 1.728 ± 1.108 1.488 ± 0.951 0.280 (0.014) Sensorimotor cortex–thalamus 0.964 ± 0.574 0.947 ± 0.537 1.019 ± 0.703 0.977 ± 0.605 0.076 (1) Pallidum–caudate 0.676 ± 0.644 0.838 ± 0.744 0.631 ± 0.825 0.715 ± 0.741 −0.099 (1) Pallidum–putamen 2.166 ± 0.898 1.840 ± 0.917 1.283 ± 0.522 1.766 ± 0.873 −0.352 (0.001) Pallidum–thalamus 0.799 ± 0.519 0.926 ± 0.523 0.565 ± 0.452 0.763 ± 0.517 −0.183 (0.54) Putamen–thalamus 0.439 ± 0.437 0.477 ± 0.388 0.346 ± 0.298 0.421 ± 0.380 −0.177 (0.572) Substantia nigra–putamen 0.004 ± 0.017 0.001 ± 0.004 0.004 ± 0.015 0.003 ± 0.013 0.037 (1) Substantia nigra–thalamus 0.001 ± 0.005 0.004 ± 0.012 0.009 ± 0.003 0.003 ± 0.009 0.103 (1) Substantia nigra–pallidum 0.096 ± 0.243 0.095 ± 0.363 0.160 ± 0.129 0.117 ± 0.261 −0.017 (0.851) Structural connectivity values based on fiber density were reported for 16 connections within the CBGT circuit. Correlation between the structural connectivity and MDS-UPDSR scores is reported as the Spearman correlation coefficient with the corrected p value in the rightmost column. The italicised text shows identified connections with significant group-wise differences in NC-PD, SWEDD-NC and SWEDD-PD comparisons as reported in Table 2 Classifier performance The SVM classifier using a quadratic kernel was applied to separate the NC, SWEDD and PD groups. Classifier performance in terms of accuracy, sensitivity and specificity are reported in Table 4, classifying the NC-PD, NC-SWEDD and PD-SWEDD cases. Overall, the classification results were generally good (i.e., mean sensitivity, specificity and accuracy were 70.83, 79.17 and 75.00 %, respectively).Table 4 Classifier performance to separate the NC-PD, NC-SWEDD and PD-SWEDD classifications Group Sensitivity (%) Specificity (%) Accuracy (%) NC versus PD 62.5 85 73.75 NC versus SWEDD 62.5 82.5 72.5 SWEDD versus PD 87.5 70 78.75 Discussion In this study, we identified pallidum–putamen, sensorimotor cortex–putamen and associative cortex–thalamus connections as connectivity profile unique to SWEDD using structural connectivity analyses. Moreover, pallidum–putamen and sensorimotor cortex–putamen connections were correlated with the MDS-UPDRS score. First, our results revealed decreased structural connectivity in pallidum–putamen connection in PD patients compared with NC and SWEDD. Neuroimaging studies have reported altered functional or structural connectivity in PD compared with NC subjects (Kim et al. 2013; Sharman et al. 2013; Wu et al. 2009; Yu et al. 2013; Zhang et al. 2015). One functional connectivity study showed a decreased connection in pallidum–putamen using rs-fMRI (Sharman et al. 2013). Others reported decreased levels of degree centrality, a graph network measure of local importance in the supplementary motor area and putamen using rs-fMRI in PD compared to NC subjects (Wu et al. 2009). Structural connectivity studies of PD patients showed marked reduction in connectivity in the nigrostriatal tract (connections among SN, STN, putamen and pallidum) using DTI (Zhang et al. 2015). Thus, our results were consistent with previous findings. Second, our results revealed decreased structural connectivity in sensorimotor–putamen connection in PD patients compared with NC and SWEDD. One functional connectivity study using rs-fMRI showed increased connection in putamen and supplementary motor area, a sub-region of sensorimotor cortex in PD patients (Yu et al. 2013). Another rs-fMRI study reported an increase in degree centrality in the parietal cortex of PD patients compared with NC subjects (Wu et al. 2009). One study using track-based spatial statistics analysis showed that bilateral motor-related tracts, such as the cortico-fugal pathway that connects the motor cortex and cerebral peduncle via the internal capsule, had higher mean diffusivity values in PD patients than in NC (Kim et al. 2013). Another study using track density reported that primary somatosensory cortices showed a significantly increased track density (Ziegler et al. 2014). Our results were consistent with previous findings. Third, our results revealed decreased structural connectivity in associative cortex–thalamus connection in PD patients compared with NC and SWEDD. One functional connectivity study using rs-fMRI reported an increase in degree centrality in the dorsolateral prefrontal cortex, a sub-region of the association cortex, in PD patients (Wu et al. 2009). A comparable study investigating associative cortex–thalamus connection using structural connectivity for PD was largely lacking. Thus, our results were partially consistent with previous findings. The pallidum–putamen and sensorimotor cortex–putamen connections were significantly correlated with clinical scores, while associative cortex–thalamus connection was not correlated with clinical scores. The pallidum–putamen connection was reported as a key pathway for motor control within the CBGT circuit (Obeso et al. 2000). The sensorimotor cortex–putamen connection was related to motor control via glutamatergic projections in the cortico-striatal pathway. Pallidum–putamen and sensorimotor cortex–putamen connections were correlated with MDS-UPDRS scores, which might corroborate the existing research. The associative cortex–thalamus connection was reported as playing a secondary role in processing motor information and thus, might be less linked to MDS-UPDRS scores than the other two connections (Purves et al. 2008). Thus, we believe that pallidum–putamen and sensorimotor cortex–putamen connections might form a structural connectivity profile unique to SWEDD that could be a potential imaging biomarker for future movement disorder research. Our study had several limitations. Our study was limited by small sample size. Only 40 SWEDD patients had both DTI, T1 and T2-weighted MRI data available in the database and thus we were limited to 40 SWEDD cases. The PD patients in our study have greater motor impairments then SWEDD patients and thus the connectivity difference between two groups could have come from either dopamine differences or degree of motor symptom severity. Further research controlling for effects of motor symptoms are needed. Another confounding factor is the lateralization of PD symptom onset (Stewart et al. 2009; Weintraub et al. 2005). The PPMI data did not consider unilateral PD symptom onset and thus the differences in connectivity could have come from variations in lateralized symptom onsets. We limited connectivity analysis to 16 known connections within the CBGT circuit, which did not cover the entire brain. We intended to focus on known connections first, thereby establishing a baseline for further research. We adopted DTI to assess fiber information, but DTI cannot distinguish between efferent and afferent connections and model complex fiber orientations; in addition, its limited voxel resolution only allows DTI to account for major fiber tracts. Use of high angular resolution diffusion imaging (HARDI) allows for complex modeling within a voxel, but HARDI requires longer scan times than DTI. In many cases, DTI is the practical option for assessing in vivo fiber information. Brain networks can be assessed not only using DTI, but also by other imaging modalities including fMRI. Multi-modal analysis of the brain network will allow incorporation of complementary information derived from different modalities to better quantify SWEDD characteristics. To date, ground truth regarding diagnosis of SWEDD is difficult to achieve. One study reported that some SWEDD patients converted to PD while others did not (Batla et al. 2014). A longitudinal follow-up exam would allow us to better assess validity of SWEDD diagnosis. Our study retrieved data from a research database, which lacked such follow-up data. We believe future research should consider such longitudinal data. There has been an update to the criteria of SWEDD and PD, which might need to be applied to PPMI data so that classification of SWEDD cases could be validated (Postuma et al. 2015). This is also left for future work. Conclusions In this study, we adopted connectivity analysis based on fiber density to characterize SWEDD patients compared to NC and PD patients. Connectivity analysis within the GBCT circuit was applied to NC (n = 40), SWEDD (n = 40) and PD (n = 40) participants. Pallidum–putamen, sensorimotor cortex–putamen and associative cortex–thalamus connections were significant (corrected p < 0.05) and could separate SWEDD from NC and PD patients in terms of structural connectivity based on fiber density. In addition, two of those connections, pallidum–putamen and sensorimotor cortex–putamen, were correlated with the MDS-UPDRS score (r = −0.352, corrected p = 0.001 and r = 0.280, corrected p = 0.014, respectively). Significant connectivity results were fed into a SVM classifier. The mean performance of NC-PD, NC-SWEDD and PD-SWEDD classifications were 70.83 % (sensitivity), 79.17 % (specificity) and 75.00 % (accuracy). These results confirmed the connections separating SWEDD from PD patients and NC were important features and well correlated with well-established clinical scores. We believe these connections could potentially serve as a unique structural connectivity profile that distinguishes SWEDD from NC and PD patients. Abbreviations PDParkinson’s disease SWEDDscans without evidence of dopaminergic deficit SNsubstantia nigra CBGTcortico-basal ganglia-thalamo-cortical NCnormal control DTIdiffusion tensor imaging rs-fMRIresting state functional MRI STNsubthalamic nucleus ROIregion of interest PPMIParkinson’s Progression Markers Initiative MDS-UPDRSMovement Disorder Society-Sponsored Unified Parkinson’s Disease Rating Scale FAfractional anisotropy FACTthe fiber assignment by continuous tracking CDconnection density CEconnection efficacy SVMsupport vector machine HARDIhigh angular resolution diffusion imaging Authors’ contributions MK and HP wrote the manuscript and researched data. MK performed the experiments. HP is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Both authors read and approved the final manuscript. Acknowledgements This study (H.P.) was supported by the Institute for Basic Science (IBS-R015-D1, http://www.ibs.re.kr/). This work was also supported by NRF (National Research Foundation of Korea) (Grant Number NRF-2016R1A2B4008545). Imaging data were obtained from the PPMI (http://www.ppmi-info.org/) funded by The Michael J. Fox Foundation for Parkinson’s Research and others. There were no other sources of funding than those listed above. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests The authors declared that they have no competing interests. ==== Refs References Albin RL Young AB Penney JB The functional anatomy of basal ganglia disorders Trends Neurosci 1989 12 10 366 375 10.1016/0166-2236(89)90074-X 2479133 Batla A Erro R Stamelou M Schneider SA Schwingenschuh P Ganos C Bhatia KP Patients with scans without evidence of dopaminergic deficit: a long-term follow-up study Mov Disord 2014 29 14 1820 1825 10.1002/mds.26018 25350529 Bullmore ET Suckling J Overmeyer S Rabe-Hesketh S Taylor E Brammer MJ Global, voxel, and cluster tests, by theory and permutation, for a difference between two groups of structural MR images of the brain IEEE Trans Med Imaging 1999 18 1 32 42 10.1109/42.750253 10193695 Daducci A Gerhard S Griffa A Lemkaddem A Cammoun L Gigandet X Meuli R Hagmann P Thiran J-P The connectome mapper: an open-source processing pipeline to map connectomes with MRI PLoS One 2012 7 12 e48121 10.1371/journal.pone.0048121 23272041 Desikan RS Ségonne F Fischl B Quinn BT Dickerson BC Blacker D Buckner RL Dale AM Maguire RP Hyman BT An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest Neuroimage 2006 31 3 968 980 10.1016/j.neuroimage.2006.01.021 16530430 Fahn S Group PS Does levodopa slow or hasten the rate of progression of Parkinson’s disease? J Neurol 2005 252 4 iv37 iv42 16222436 Feldman HM Yeatman JD Lee ES Barde LH Gaman-Bean S Diffusion tensor imaging: a review for pediatric researchers and clinicians JDBP 2010 31 4 346 20453582 Fischi-Gómez E Vasung L Meskaldji D-E Lazeyras F Borradori-Tolsa C Hagmann P Barisnikov K Thiran J-P Hüppi PS Structural brain connectivity in school-age preterm infants provides evidence for impaired networks relevant for higher order cognitive skills and social cognition Cereb Cortex 2015 25 9 2793 2805 10.1093/cercor/bhu073 24794920 Fischl B FreeSurfer Neuroimage 2012 62 2 774 781 10.1016/j.neuroimage.2012.01.021 22248573 Hagmann P Cammoun L Gigandet X Meuli R Honey CJ Wedeen VJ Sporns O Mapping the structural core of human cerebral cortex PLoS Biol 2008 6 7 e159 10.1371/journal.pbio.0060159 18597554 Hagmann P Sporns O Madan N Cammoun L Pienaar R Wedeen VJ Meuli R Thiran J-P Grant P White matter maturation reshapes structural connectivity in the late developing human brain Proc Natl Acad Sci 2010 107 44 19067 19072 10.1073/pnas.1009073107 20956328 Im K Paldino MJ Poduri A Sporns O Grant PE Altered white matter connectivity and network organization in polymicrogyria revealed by individual gyral topology-based analysis Neuroimage 2014 86 182 193 10.1016/j.neuroimage.2013.08.011 23954485 Keuken MC Bazin P-L Crown L Hootsmans J Laufer A Müller-Axt C Sier R van der Putten E Schäfer A Turner R Quantifying inter-individual anatomical variability in the subcortex using 7T structural MRI NeuroImage 2014 94 40 46 10.1016/j.neuroimage.2014.03.032 24650599 Kim M Park H Using Tractography to Distinguish SWEDD from Parkinson’s Disease Patients Based on Connectivity Parkinsons Dis 2016 Kim HJ Kim SJ Kim HS Choi CG Kim N Han S Jang EH Chung SJ Lee CS Alterations of mean diffusivity in brain white matter and deep gray matter in Parkinson’s disease Neurosci Lett 2013 550 64 68 10.1016/j.neulet.2013.06.050 23831353 Leh SE Petrides M Strafella AP The neural circuitry of executive functions in healthy subjects and Parkinson’s disease Neuropsychopharmacology 2010 35 1 70 85 10.1038/npp.2009.88 19657332 Marek K Jennings D Lasch S Siderowf A Tanner C Simuni T Coffey C Kieburtz K Flagg E Chowdhury S The parkinson progression marker initiative (PPMI) Prog Neurobiol 2011 95 4 629 635 10.1016/j.pneurobio.2011.09.005 21930184 Mori S van Zijl P Fiber tracking: principles and strategies–a technical review NMR Biomed 2002 15 7–8 468 480 10.1002/nbm.781 12489096 Nichols TE Holmes AP Nonparametric permutation tests for functional neuroimaging: a primer with examples Hum Brain Mapp 2002 15 1 1 25 10.1002/hbm.1058 11747097 Obeso JA Rodriguez-Oroz MC Rodriguez M Lanciego JL Artieda J Gonzalo N Olanow CW Pathophysiology of the basal ganglia in Parkinson’s disease Trends Neurosci 2000 23 S8 S19 10.1016/S1471-1931(00)00028-8 11052215 Obeso JA Rodríguez-Oroz MC Benitez-Temino B Blesa FJ Guridi J Marin C Rodriguez M Functional organization of the basal ganglia: therapeutic implications for Parkinson’s disease Mov Disord 2008 23 S3 S548 S559 10.1002/mds.22062 18781672 Parent A Hazrati L-N Functional anatomy of the basal ganglia. I. The cortico-basal ganglia-thalamo-cortical loop Brain Res Rev 1995 20 1 91 127 10.1016/0165-0173(94)00007-C 7711769 Postuma RB Berg D Stern M Poewe W Olanow CW Oertel W Obeso J Marek K Litvan I Lang AE MDS clinical diagnostic criteria for Parkinson’s disease Mov Disord 2015 30 12 1591 1601 10.1002/mds.26424 26474316 Purves D Augustine G Fitzpatrick D Hall W LaMantia A McNamara J White L Neuroscience 2008 Sunderland De Boeck, Sinauer Schneider SA Edwards MJ Mir P Cordivari C Hooker J Dickson J Quinn N Bhatia KP Patients with adult-onset dystonic tremor resembling parkinsonian tremor have scans without evidence of dopaminergic deficit (SWEDDs) Mov Disord 2007 22 15 2210 2215 10.1002/mds.21685 17712858 Schwingenschuh P Ruge D Edwards MJ Terranova C Katschnig P Carrillo F Silveira-Moriyama L Schneider SA Kägi G Palomar FJ Distinguishing SWEDDs patients with asymmetric resting tremor from Parkinson’s disease: a clinical and electrophysiological study Mov Disord 2010 25 5 560 569 10.1002/mds.23019 20131394 Sharman M Valabregue R Perlbarg V Marrakchi-Kacem L Vidailhet M Benali H Brice A Lehéricy S Parkinson’s disease patients show reduced cortical–subcortical sensorimotor connectivity Mov Disord 2013 28 4 447 454 10.1002/mds.25255 23144002 Shu N Liu Y Li J Li Y Yu C Jiang T Altered anatomical network in early blindness revealed by diffusion tensor tractography PLoS One 2009 4 9 e7228 10.1371/journal.pone.0007228 19784379 Shu N Liu Y Li K Duan Y Wang J Yu C Dong H Ye J He Y Diffusion tensor tractography reveals disrupted topological efficiency in white matter structural networks in multiple sclerosis Cereb Cortex 2011 21 11 2565 2577 10.1093/cercor/bhr039 21467209 Smith SM Jenkinson M Woolrich MW Beckmann CF Behrens TE Johansen-Berg H Bannister PR De Luca M Drobnjak I Flitney DE Advances in functional and structural MR image analysis and implementation as FSL Neuroimage 2004 23 S208 S219 10.1016/j.neuroimage.2004.07.051 15501092 Smith SM Vidaurre D Beckmann CF Glasser MF Jenkinson M Miller KL Nichols TE Robinson EC Salimi-Khorshidi G Woolrich MW Functional connectomics from resting-state fMRI Trends Cogn Sci 2013 17 12 666 682 10.1016/j.tics.2013.09.016 24238796 Stewart KC Fernandez HH Okun MS Rodriguez RL Jacobson CE Hass CJ Side onset influences motor impairments in Parkinson disease Parkinsonism Relat Disord 2009 15 10 781 783 10.1016/j.parkreldis.2009.02.001 19447063 Stoessl AJ Neuroimaging in Parkinson’s disease Neurotherapeutics 2011 8 1 72 81 10.1007/s13311-010-0007-z 21274687 Tolosa E Wenning G Poewe W The diagnosis of Parkinson’s disease Lancet Neurol 2006 5 1 75 86 10.1016/S1474-4422(05)70285-4 16361025 Tuch DS Wedeen VJ Dale AM George JS Belliveau JW Conductivity tensor mapping of the human brain using diffusion tensor MRI Proc Natl Acad Sci 2001 98 20 11697 11701 10.1073/pnas.171473898 11573005 Vapnik VN An overview of statistical learning theory IEEE Trans Neural Netw 1999 10 5 988 999 10.1109/72.788640 18252602 Wang R, Benner T, Sorensen A, Wedeen V (2007) Diffusion toolkit: a software package for diffusion imaging data processing and tractography. In: Proceedings of the international society for magnetic resonance in medicine, vol 3720 Weintraub D Newberg AB Cary MS Siderowf AD Moberg PJ Kleiner-Fisman G Duda JE Stern MB Mozley D Katz IR Striatal dopamine transporter imaging correlates with anxiety and depression symptoms in Parkinson’s disease J Nucl Med 2005 46 2 227 232 15695780 Wen W Zhu W He Y Kochan NA Reppermund S Slavin MJ Brodaty H Crawford J Xia A Sachdev P Discrete neuroanatomical networks are associated with specific cognitive abilities in old age J Neurosci 2011 31 4 1204 1212 10.1523/JNEUROSCI.4085-10.2011 21273405 Wu T Wang L Chen Y Zhao C Li K Chan P Changes of functional connectivity of the motor network in the resting state in Parkinson’s disease Neurosci Lett 2009 460 1 6 10 10.1016/j.neulet.2009.05.046 19463891 Yu R Liu B Wang L Chen J Liu X Enhanced functional connectivity between putamen and supplementary motor area in Parkinson’s disease patients PLoS One 2013 8 3 e59717 10.1371/journal.pone.0059717 23555758 Zhang Z Liao W Chen H Mantini D Ding J-R Xu Q Wang Z Yuan C Chen G Jiao Q Altered functional–structural coupling of large-scale brain networks in idiopathic generalized epilepsy Brain 2011 134 10 2912 2928 10.1093/brain/awr223 21975588 Zhang Y Wu IW Buckley S Coffey CS Foster E Mendick S Seibyl J Schuff N Diffusion tensor imaging of the nigrostriatal fibers in Parkinson’s disease Mov Disord 2015 30 9 1229 1236 10.1002/mds.26251 25920732 Ziegler E Rouillard M André E Coolen T Stender J Balteau E Phillips C Garraux G Mapping track density changes in nigrostriatal and extranigral pathways in Parkinson’s disease Neuroimage 2014 99 498 508 10.1016/j.neuroimage.2014.06.033 24956065

PMC005xxxxxx/PMC5001968.txt

==== Front SpringerplusSpringerplusSpringerPlus2193-1801Springer International Publishing Cham 311310.1186/s40064-016-3113-5ResearchRandom rotation survival forest for high dimensional censored data Zhou Lifeng lfzhou@csu.edu.cn Wang Hong wh@csu.edu.cn Xu Qingsong qsxu@csu.edu.cn School of Mathematics and Statistics, Central South University, South Shaoshan Road, Changsha, 410075 Hunan China 26 8 2016 26 8 2016 2016 5 1 142511 6 2016 19 8 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Recently, rotation forest has been extended to regression and survival analysis problems. However, due to intensive computation incurred by principal component analysis, rotation forest often fails when high-dimensional or big data are confronted. In this study, we extend rotation forest to high dimensional censored time-to-event data analysis by combing random subspace, bagging and rotation forest. Supported by proper statistical analysis, we show that the proposed method random rotation survival forest outperforms state-of-the-art survival ensembles such as random survival forest and popular regularized Cox models. Keywords Survival ensembleRotation forestTime-to-event dataCensored dataHigh-dimensional dataSocial Science Foundation for Young Scholars of Ministry of Education of China15YJCZH166Wang Hong issue-copyright-statement© The Author(s) 2016 ==== Body Background Survival analysis of censored data plays a vital role in statistics with abundant applications in various fields such as biostatistics, engineering, finance and economics. As an example, regression analysis of time-to-event data finds wide applications in reliability studies in industrial engineering and inter-child birth times research in demography and sociology. To estimate the probability that a subject (patient or equipment) will survive past a certain time, various parametric, semi-parametric and no-parametric models such as Cox proportional hazard (Cox PH) model (Cox and Oakes 1984; David 1972), survival neural network (Faraggi and Simon 1995), survival tree (Bou-Hamad et al. 2011; LeBlanc and Crowley 1995), regularized Cox PH model (Fan and Li 2002), regularized accelerated failure time (AFT) model (Huang et al. 2006), supervised principal components based survival models (Li and Li 2004) have been proposed. The past two decades have seen various survival ensembles with parametric and/or non-parametric base models and combining techniques. These techniques include bagging (Hothorn et al. 2004, 2006), boosting (Binder and Schumacher 2008; Binder et al. 2009; Hothorn and Bühlmann 2006; Li and Luan 2005; Ma and Huang 2007; Ridgeway 1999; Wang and Wang 2010), random survival forest (RSF) (Ishwaran et al. 2010, 2011) and the recently proposed rotation survival forest (RotSF) (Zhou et al. 2015). Bagging stochastically changes the distribution of the training data by constructing a base survival model based on different bootstrap samples (Hothorn et al. 2004). Boosting based approaches adaptively change the distribution of the training data according to the performance of previously trained base models and usually either use all covariates to fit the gradients in each step for low-dimensional data (Ridgeway 1999) or update only one estimate of parameters corresponding to only one covariate in a componentwise manner in case of high-dimensional data (Binder et al. 2009). Random survival forest (RSF) (Ishwaran et al. 2008) extends random forest (RF) (Breiman 2001) to right-censored time-to-event data using the same principles underlying RF and enjoys all RF’s important properties. In RSF, tree node splits are designed to maximizing survival differences between subtree nodes. A so-called ensemble cumulative hazard function (CHF) can be estimated by aggregating Nelson–Aalen estimators for all “in-bag” data samples. All these survival ensembles have demonstrated their usefulness compared to previous single algorithms. Rotation survival forest (RotSF) (Zhou et al. 2015) is newly proposed survival ensemble based on rotation forest (RotF) (Rodriguez et al. 2006), in which the training data for each base model is formed by applying PCA transformation to rotate the original covariates axes. In RotSF and other RotF based approaches, ensemble diversity is achieved by covariates transformation for each base model and prediction accuracy is promoted by keeping all principal components in the training data set. However, due to intensive computations during eigenvalue decomposition of data covariance matrix, such approaches often fails when dealing with high-dimensional data. In view of the fact that dimensionality reduction can be achieved by random subspace (Ho 1998) method which randomly selects a small number of dimensions from a given covariate set in building a base model, we propose a new survival ensemble called random rotation survival forest (RRotSF) for analyzing high-dimensional survival data. The proposed methodology can be viewed as a combination of rotation forest, random subspace and bagging (Breiman 1996). And it extends the RotSF approach from low dimensional to high dimensional time-to-event censored data. In this study, the decision tree algorithm is chosen as the base survival model for our survival ensemble as it is the most popular non-parametric method in analyzing survival data (Bou-Hamad et al. 2011). Methods Given a training dataset: D=(τq,δq,Xq),q=1,…,n, where τq is the survival time for the q-th sample, δq is the censored status indicator, Xq is a variable set V of p covariates and n is the number of observed samples, a high-level description of how the proposed RRotSF algorithm train a base survival model Si in the ensemble is presented in the following:Randomly select r<p covariates from the p-dimensional data set D and the newly obtained training set Dr=(τq,δq,Xqi) consists of r-dimensional training samples. Here we set r=p for simplicity. Generate a bootstrap sample D′=(τq′,δq′,X′qi) of size n from Dr to enhance diversity and for calculating covariate importance. Randomly split variables V into k=r/M equal size subsets Vj,j=1,…,k and denote the not used covariates (remaining variables) as RV. Apply PCA to each bagged training subset with Vj covariates. Retain all derived principal component rotations Mjs and set rotations of RV to 0 to inject more randomness. Arrange all PCA rotations to match variable order in V and obtain rotation matrix Ria. Use the newly transformed data Dt=(τq′,δq′,Xqi′Ria) as the training set to train a base survival model Si. The major difference between RotF (also RotSF) and the proposed RRotSF lies in that the former transforms the whole training set via PCA while the latter transforms only a random subspace of the whole training set which in turn greatly reduces the computational complexity caused by eigenvalue decomposition of high-dimensional covariance matrix. The pseudo-code of the proposed RRotSF algorithm is presented in Algorithm 1: Some parameters should be specified before applying RRotSF. Similar to other ensemble methods, ensemble size which specifies the number of base survival models can be tuned by the users. Parameters M which controls the number within a feature subset is set to 2 as is done in RotSF. Results and discussion In the experiments, we perform five replications of two fold cross-validation as suggested by Dietterich (1998). In fivefold to twofold cross-validation, the dataset is randomly divided into two halves, the first half is used for training and the other half for testing and vice versa. This process is repeated five times for each dataset. Datasets In order to carry out empirical comparisons, we want to test the proposed algorithm on five real high-dimensional benchmark datasets. In the following datasets, when distant metastasis-free survival (DMFS) time values are available, DMFS values are used as the primary survival end-points, otherwise relapse-free or overall survival time values are applied. A short introduction of the benchmark datasets are given below. UPP dataset The UPP dataset contains transcript profiles of 251 p53-sequenced primary breast tumors published by Miller et al. (2005). In each patient sample, 44,928 gene features and 21 clinical covariates are provided. The data can be obtained from the R package “breastCancerUPP” of “Bioconductor”. MAINZ dataset The MAINZ breast cancer dataset provided by Schmidt et al. (2008) contains the gene expression patterns of 200 tumors of patients who were not treated by systemic therapy after surgery using a discovery approach. Each patient sample contains 22,283 gene features and 21 clinical covariates. The dataset is available from the R package “breastCancerMAINZ” of “Bioconductor”. TransBig dataset This breast cancer dataset contains gene expression and clinical data published in Desmedt et al. (2007). The data contains 198 samples to independently validate a 76-gene prognostic breast cancer signature as part of the TransBig project. In the data, 22,283 gene features and 21 clinical covariates are provided for each sample. The dataset can be obtained through the R package “breastCancerTRANSBIG” of “Bioconductor”. VDX dataset The Veridex (VDX) dataset which contains 344 patients with primary breast cancer was published in Wang et al. (2005). In the data, 22,283 gene features and 21 clinical covariates are provided for each sample. The dataset can be obtained through the R package “breastCancerVDX” of “Bioconductor”. TCGA dataset This dataset is provided by The Cancer Genome Atlas (TCGA) and presented in Fang and Gough (2014). It contains both clinical covariates and gene expression information of 3096 cancer patients covering 12 major types of cancers. In each sample, 19,420 gene state information and 5 clinical covariates are provided. The data is available from the R package “dnet” of “CRAN”. Summary information including gene features, clinical covariates and the number of samples of all datasets can be found in the following Table 1.Table 1 Summary of five benchmark datasets used Gene features Clinical covariates Samples UPP 44,928 21 251 MAINZ 22,283 21 200 TransBig 22,283 21 198 VDX 22,283 21 344 TCGA 19,420 5 3096 Performance metrics and statistical tests In survival analysis, we are much concerned with the relative risks between patients with different covariates information. Hence, as suggested by Ishwaran et al. (2008), we adopt Harrell’s concordance index (C-index, CI) (Harrell et al. 1996) to evaluate the accuracy of such relative risks in our later experiments and rank their performance on all datasets. CI can be calculated in the following steps:Create all pairs of observed survival times. For all valid survival time pairs, namely, pairs where one survival time Tj1 is greater than the other Tj2, test whether the corresponding predictions are concordant, i.e, ηj1>ηj2. If so, add 1 to the running sum s; If ηj1=ηj2, add 0.5 to the sum s; If ηj1<ηj2, add 0 to the sum s. Count the number n of valid survival time pairs. Divide the total sum s by the number of valid survival time pairs n and we obtain CI=s/n. Similar to AUC used in classification, CI usually lies between 0.5 and 1. When CI=1, it means that the model has a perfect prediction accuracy and when CI=0.5, it implies that the model is just like random guessing. The results obtained in experiments are further validated by some proper statistical tests. As suggested by Demšar (2006) and was done in Zhou et al. (2015), we use the non-parametric Friedman rank sum test (Demšar 2006) to test the statistical significance of various survival models. If the value Friedman test is large enough, the null hypothesis that there is no significant difference among the different survival models can be rejected and some post-hoc such as Nemenyi test can be applied to find where the differences lie. If the differences are not significant according to the Nemenyi statistics, we use a two-sample Wilcoxon test to check whether the difference between pairs is significant. Comparison results Here, we compare RRotSF with five popular survival models. The first two methods are random survival forest (RSF) with different splitting rules, namely RSF-Logrank (RSFl) and RSF-Logrankscore (RSFs); the third and forth methods are regularized Cox proportional hazard models, i.e. Cox-Lasso and Cox-Ridge; the fifth method is fast cocktail Cox method (CockTail). For the ease of notation, RRotSF, RSFl, RSFs, Cox-Lasso, Cox-Ridge and CockTail are denoted by A, B, C, D, E and F respectively when necessary. Comparisons with five models are conducted with corresponding “glmnet” (Simon et al. 2011), “randomForestSRC” (Ishwaran et al. 2008), and “fastcox” (Yang and Zou 2012) packages in R. Default settings are adopted for all models. For ensemble methods, i.e. RRotSF, RSFl and RSFs, 500 trees are built. The corresponding experiment results are listed in the following Table 2. In this table, the numerics in each entry are the average of CI values on fivefold to twofold cross-validation. The best performance in each column (on each dataset) is highlighted by the italic font.Table 2 Performance in terms of averaged CI UPP MAINZ TransBig VDX TCGA RRotSF 0.6210 0.6997 0.5540 0.6248 0.6287 RSFl 0.6461 0.7069 0.5177 0.5630 0.5740 RSFls 0.5813 0.6234 0.5375 0.5950 0.6569 Cox-Lasso 0.5763 0.6375 0.5482 0.5327 0.7032 Cox-Ridge 0.6149 0.6802 0.5702 0.6234 0.5516 CockTail 0.5906 0.6298 0.5383 0.5227 0.7051 According to Table 2, the proposed RRotSF takes the first place once, takes the second place three times and also takes a fourth place. Though RSFl takes the first place twice, its performance on other tree datasets are rather poor: it takes one fourth, one fifth and one last place respectively. From Table 2, whether RRotSF beats RSFl is not clear at this time but we can safely say that RRotSF outperforms other models, namely RSFls, Cox-Lasso, Cox-Ridge and CockTail in most cases in terms of averaged CI. To further evaluate the performance of all compared models, we have ranked each model on every run on these benchmark datasets. This allows us to compare performance of all models in a consistent and nonparametric way. Figure 1 presents the boxplot of ranks of six models in all runs of the experiments.Fig. 1 Ranks of performance in terms of CI From the above, we can observe that RRotSF excels, followed by Cox-Ridge and RSFl models. The worst performer on these datasets is CockTail. In spite of the ranks, we also want to contrast these statements with some statistical tests. The Friedman rank sum test outputs a p-value of 0.0001136 which reject the null hypothesis that there is no significant difference among these models and a post-hoc Nemenyi test is applied. Using RRotSF (A) as the control, we obtain the p-values of Nemenyi test for different pairs: pBA=0.35512,pCA=0.04505,pDA=0.00184,pEA=0.61393 and pFA=0.00022. It can be seen that there exist significant differences between RRotSF and RSFs, Cox-Lasso or CockTail. The differences between RRotSF and RSFl or Cox-Ridge are not significantly different according to the Nemenyi test. However, a pairwise comparison using the Wilcoxon test rejects the hypothesis of equivalence with low p-values (pBA=0.02189 and pEA=0.02129). This also indicates RRotSF is also superior to RSFl and Cox-Ridge on these benchmark datasets. Therefore, in terms of C-index metric, RRotSF outperforms state-of-the-art survival models such as Random Survival Forest, regularized Cox proportional hazard models on these benchmark datasets. It is clear that other methods (ensembles and not) are available but the goal here is to illustrate some key features of RRotSF and not to provide an exhaustive comparison across methods. Parameter sensitivity analysis In addition to the above experiments, we also want to examine the sensitivity of RRotSF to the choice of parameters in the underlying survival models. First, we want to test the performance of RRotSF with different subspace values (the number of variable with each variable subset) r. In view of the fact that r<p/5 may result in a less accurate base survival tree and r>5p may cause RRotSF cease to work due to memory overflow problems as all the datasets here are high-dimensional ones, we only test RRotSF with r values ranging from p/5 and 5p in the experiments. Figure 2 shows the performance of RRotSF with different values of r on all five benchmark datasets. Performance of the default value( r=p ) of r on each datasets is indicated by a purple circle.Fig. 2 Performance with different values of r From Fig. 2, one may observe that except for the values at the very beginning on TransBig and TCGA datasets, RRotSF seems insensitive to changes of r values. This is very encouraging result, as it demonstrates that RRotSF is robust with respect to r, even if non-default values are chosen. Next, we want to test the performance of RRotSF with number of variable with a subset M. If M=1, then any projection reduces to rescaling of the variable axes. If M=p, there is only one variable set, i.e. all the variables are used for PCA transformation. In both cases, the ensemble diversity are degraded and prediction errors are larger than those with values in between (Kuncheva and Rodríguez 2007). To see how the choice of M may influence RRotSF’s performance, we test M with values between 2 and 100. Figure 3 shows the performance of RRotSF with different values of M on all five benchmark datasets.Fig. 3 Performance with different values of M The results shown in Fig. 3 agree with the results obtained for RotF in the classification context (Kuncheva and Rodríguez 2007), i.e., there is no consistent pattern or regularity for M with small values. We also consider the time efficiency of RRotSF for different values of M. From Algorithm 1, one may observe that the major time complexity lies with PCA operations in transforming k group of variable subsets. If M is small, each PCA operations will be faster but as the number of variable subsets could be greater, we have to do more PCA transformations. If M is large, each PCA operation will take a longer time but the the number of variable subsets and hence the number of PCA operations could be less. Figure 4 shows the running times (in seconds) of RRotSF on benchmark datasets with different combinations of M and k.Fig. 4 Time complexity with different values of M From Fig. 4, we notice a sharp decrease in RRotSF’s running time when M increases if 2≤M≤5, and a very slow decrease when M increases if 5<M≤20. When M>20, the values of M have no direct influence on RRotSF’s time efficiency as RRotSF’s running time remain almost steady on all five benchmark datasets. If we focus only on RRotSF’s time efficiency, we should choose a larger M (M<20). However, to make RRotSF also work for some low-dimensional datasets, M should be set to a small value to ensure that there is enough diversity among the survival ensemble. Hence, to make a tradeoff between time efficiency and prediction accuracy, the M value can be set to 2 or 3 for simplicity, though it is not a optimal value in most cases. From the above, both the choices r=p and M=2 in the default setting for RRotSF are not the best choice in terms of C-index and are just serendipitous guesses in this study. As we have shown in the above comparison results, RRotSF has outperformed other popular survival models for these rather unfavourable values, we may conclude that RRotSF is not sensitive to the choice of r and M. As both values work well in the experiments, we propose to use these values as default values in the future. Of course, one can use cross-validation techniques to tune these parameters on some particular cases for a better performance. Conclusion In this study, we have developed a new ensemble learning algorithm, random rotation survival forest, for high-dimensional survival analysis. By studying the famous benchmark datasets, we have found that the proposed method generally outperforms state-of-the-art survival models such as random survival forest, regularized Cox proportional hazard models in terms of C-Index metric. As a non-parametric approach, RRotSF does not impose parametric assumptions on hazard functions, and it extends the well-known rotation forest methodology to high-dimensional data analysis. The R code and and the supplementary material are available at url: https://github.com/whcsu/RRotSF and we are working hard to provide an R package for the proposed RRotSF algorithm as soon as possible. Authors' contributions The work presented here was carried out in collaboration between all authors. LFZ,HW and QSX defined the research theme; HW and QSX designed the algorithm; HW carried out the experiments and analyzed the data; LFZ and HW interpreted the results and wrote the paper. All authors read and approved the final manuscript. Acknowlegements This work was supported in part by Social Science Foundation for Young Scholars of Ministry of Education of China Under Grant No. 15YJCZH166. The authors want to thank all reviewers and the editor for their valuable and constructive comments, which greatly improves the quality of this paper. Competing interests The authors declare that they have no competing interests. ==== Refs References Binder H Schumacher M Allowing for mandatory covariates in boosting estimation of sparse high-dimensional survival models BMC Bioinform 2008 9 1 14 10.1186/1471-2105-9-14 Binder H Allignol A Schumacher M Beyersmann J Boosting for high-dimensional time-to-event data with competing risks Bioinformatics 2009 25 7 890 896 10.1093/bioinformatics/btp088 19244389 Bou-Hamad I Larocque D Ben-Ameur H A review of survival trees Stat Surv 2011 5 44 71 10.1214/09-SS047 Breiman L Bagging predictors Mach Learn 1996 24 2 123 140 Breiman L Random forests Mach Learn 2001 45 1 5 32 10.1023/A:1010933404324 Cox DR Oakes D Analysis of survival data 1984 Boca Raton CRC Press David CR Regression models and life tables (with discussion) J R Stat Soc 1972 34 187 220 Demšar J Statistical comparisons of classifiers over multiple data sets J Mach Learn Res 2006 7 1 30 Desmedt C Piette F Loi S Wang Y Lallemand F Haibe-Kains B Viale G Delorenzi M Zhang Y d’Assignies MS Strong time dependence of the 76-gene prognostic signature for node-negative breast cancer patients in the transbig multicenter independent validation series Clin Cancer Res 2007 13 11 3207 3214 10.1158/1078-0432.CCR-06-2765 17545524 Dietterich TG Approximate statistical tests for comparing supervised classification learning algorithms Neural Comput 1998 10 7 1895 1923 10.1162/089976698300017197 9744903 Fan J Li R Variable selection for cox proportional hazards model and frailty model Ann Stat 2002 30 1 74 99 10.1214/aos/1015362185 Fang H Gough J The ’dnet’ approach promotes emerging research on cancer patient survival Genome Med 2014 6 64 25246945 Faraggi D Simon R A neural network model for survival data Stat Med 1995 14 1 73 82 10.1002/sim.4780140108 7701159 Harrell FE Lee KL Mark DB Tutorial in biostatistics multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors Stat Med 1996 15 361 387 10.1002/(SICI)1097-0258(19960229)15:4<361::AID-SIM168>3.0.CO;2-4 8668867 Ho TK The random subspace method for constructing decision forests IEEE Trans Pattern Anal Mach Intell 1998 20 8 832 844 10.1109/34.709601 Hothorn T Bühlmann P Model-based boosting in high dimensions Bioinformatics 2006 22 22 2828 2829 10.1093/bioinformatics/btl462 16940323 Hothorn T Lausen B Benner A Bagging survival trees Stat Med 2004 23 1 77 91 10.1002/sim.1593 14695641 Hothorn T Bühlmann P Dudoit S Molinaro A Van Der Laan MJ Survival ensembles Biostatistics 2006 7 3 355 373 10.1093/biostatistics/kxj011 16344280 Huang J Ma S Xie H Regularized estimation in the accelerated failure time model with high-dimensional covariates Biometrics 2006 62 3 813 820 10.1111/j.1541-0420.2006.00562.x 16984324 Ishwaran H Kogalur UB Blackstone EH Lauer MS Random survival forests Ann Appl Stat 2008 2 3 841 860 10.1214/08-AOAS169 Ishwaran H Kogalur UB Gorodeski EZ Minn AJ Lauer MS High-dimensional variable selection for survival data J Am Stat Assoc 2010 105 489 205 217 10.1198/jasa.2009.tm08622 Ishwaran H Kogalur UB Chen X Minn AJ Random survival forests for high-dimensional data Stat Anal Data Min 2011 4 1 115 132 10.1002/sam.10103 Kuncheva LI Rodríguez JJ Haindl M Kittler J Roli F An experimental study on rotation forest ensembles Multiple classifier systems 2007 New York Springer 459 468 LeBlanc M Crowley J Thall PF A review of tree-based prognostic models Recent advances in clinical trial design and analysis 1995 New York Springer 113 124 Li L Li H Dimension reduction methods for microarrays with application to censored survival data Bioinformatics 2004 20 18 3406 3412 10.1093/bioinformatics/bth415 15256406 Li H Luan Y Boosting proportional hazards models using smoothing splines, with applications to high-dimensional microarray data Bioinformatics 2005 21 10 2403 2409 10.1093/bioinformatics/bti324 15713732 Ma S Huang J Clustering threshold gradient descent regularization: with applications to microarray studies Bioinformatics 2007 23 4 466 472 10.1093/bioinformatics/btl632 17182700 Miller LD Smeds J George J Vega VB Vergara L Ploner A Pawitan Y Hall P Klaar S Liu ET An expression signature for p53 status in human breast cancer predicts mutation status, transcriptional effects, and patient survival Proc Natl Acad Sci USA 2005 102 38 13550 13555 10.1073/pnas.0506230102 16141321 Ridgeway G The state of boosting Comput Sci Stat 1999 31 172 181 Rodriguez JJ Kuncheva LI Alonso CJ Rotation forest: a new classifier ensemble method IEEE Trans Pattern Anal Mach Intell 2006 28 10 1619 1630 10.1109/TPAMI.2006.211 16986543 Schmidt M Böhm D von Törne C Steiner E Puhl A Pilch H Lehr H-A Hengstler JG Kölbl H Gehrmann M The humoral immune system has a key prognostic impact in node-negative breast cancer Cancer Res 2008 68 13 5405 5413 10.1158/0008-5472.CAN-07-5206 18593943 Simon N Friedman JH Hastie T Tibshirani R Regularization paths for Cox’s proportional hazards model via coordinate descent J Stat Softw 2011 39 5 1 13 10.18637/jss.v039.i05 27065756 Wang Y Klijn JG Zhang Y Sieuwerts AM Look MP Yang F Talantov D Timmermans M van Meijer-Gelder ME Yu J Gene-expression profiles to predict distant metastasis of lymph-node-negative primary breast cancer Lancet 2005 365 9460 671 679 10.1016/S0140-6736(05)70933-8 15721472 Wang Z Wang C Buckley-James boosting for survival analysis with high-dimensional biomarker data Stat Appl Genet Mol Biol 2010 9 1 24 Yang Y Zou H A cocktail algorithm for solving the elastic net penalized Cox’s regression in high dimensions Stat Interface 2012 6 2 167 173 10.4310/SII.2013.v6.n2.a1 Zhou L Xu Q Wang H Rotation survival forest for right censored data PeerJ 2015 3 1009 10.7717/peerj.1009

PMC005xxxxxx/PMC5001969.txt

==== Front SpringerplusSpringerplusSpringerPlus2193-1801Springer International Publishing Cham 306310.1186/s40064-016-3063-yResearchSchwarz alternating methods for anisotropic problems with prolate spheroid boundaries Dai Zhenlong 140901006@stu.njnu.edu.cn 1Du Qikui duqikui@njnu.edu.cn 1Liu Baoqing lyberal@163.com 21 Jiangsu Key Laboratory for NSLSCS, School of Mathematics Sciences, Nanjing Normal University, No. 1 Wenyuan Road, Nanjing, 210023 People’s Republic of China 2 School of Applied Mathematics, Nanjing University of Finance and Economics, No. 3 Wenyuan Road, Nanjing, 210023 People’s Republic of China 26 8 2016 26 8 2016 2016 5 1 14233 5 2016 12 8 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Schwarz alternating algorithm, which is based on natural boundary element method, is constructed for solving the exterior anisotropic problem in the three-dimension domain. The anisotropic problem is transformed into harmonic problem by using the coordinate transformation. Correspondingly, the algorithm is also changed. Continually, we analysis the convergence and the error estimate of the algorithm. Meanwhile, we give the contraction factor for the convergence. Finally, some numerical examples are computed to show the efficiency of this algorithm. Keywords Schwarz alternating algorithmExterior anisotropic problemProlate ellipsoidalArtificial boundaryIteration methodhttp://dx.doi.org/10.13039/501100001809National Natural Science Foundation of China1137119811401296Liu Baoqing Jiangsu Provincial Natural Science Foundation of ChinaBK20141008Liu Baoqing Natural science fund for colleges and universities in Jiangsu Province14KJB110007Liu Baoqing issue-copyright-statement© The Author(s) 2016 ==== Body Background How to deal with boundary value problems has always been a essential part of partial differential equation. Finite difference method (FDM) (Evans 1977) and finite element method (FEM) (Brenner and Scott 1996) are the most widely used method to solve PDE numerically. These two methods become in vain when it comes to the problem over unbounded domain. To overcome this, boundary element method (BEM), which can reduce the dimension of the computational domain and is suitable for solving problems in the unbounded domains, is proposed in Feng (1980). Although, it is better to handle BEM with infinite regions, it doesn’t work so well as FEM in bounded ones. Hence, the coupling of BEM and FEM becomes the interest of researchers. Papers MacCamy and Marin (1980), Hsiao and Porter (1986), Wendland (1986), Costabel (1987), Han (1990) had focused on this method. In 1983, Feng firstly proposed a direct and natural coupling method. Later in the same year, Feng and Yu (1983) formally named the method as natural boundary element method (NBEM). Meanwhile, the DtN method, which has the similar principle with NBEM, is proposed in Keller and Givoli (1989), Grote and Keller (1995). Du and Yu (2001), Hu and Yu (2001), Gatica et al. (2003), Koyama (2007), Koyama (2009), Das and Mehrmann (2016), Das and Natesan (2014), Das (2015) and references therein present the applications of this methods. Among the reasons that effects the NBEM, the shape of artificial boundary is the essential one. Classically, circle (Givoli and Keller 1989) and spherical (Grote and Keller 1995; Wu and Yu 1998, 2000a) are chosen as the artificial boundaries. Few papers Grote and Keller (1995), Wu and Yu (2000b), Huang and Yu (2006) focus on the special artificial boundaries. These papers also proved the classic artificial boundaries were not suitable for the problem with irregular shape. On the other hand, the coupling of FEM and BEM are not enough as the performance of computer developed. The domain decomposition method (DDM) (Brenner and Scott 1996), which separates the infinite region as sum of bounded one and unbounded one with an artificial boundary on which an iteration method is constructed in, is applied on the NBEM (Yu 1994). Wu and Yu (2000b) applied this method over an infinite region. Continually, Huang et al. (2009) and Luo et al. (2013) applied this method in different problems. In this paper, we consider the anisotropic harmonic problem over an exterior three-dimensional domain. A Schwartz alternating method is designed for the numerical solution with prolate artificial boundaries. The outline of the paper is as follows. In “Schwarz alternating algorithm based on NBR” section, we divide the original domain Ω into two overlapping subdomains Ω1 and Ω2 by choosing two artificial boundaries Γ1 and Γ2, then we construct the Schwarz alternating algorithm. We prove the convergence of the algorithm in “Convergence of the algorithm” section. The convergence rate of the algorithm is analysed in the “Analysis of the convergence rate” section. In “The error estimates of the algorithm” section, we deduce the error estimates of the discrete algorithm. In “Numerical results” section, numerical examples are computed to express the advantages of this method. Finally, we give some conclusions in “Conclusions” section. Schwarz alternating algorithm based on NBR Let Ω⊂R3 be a cuboid Lipschitz unbounded domain and Γ0=∂Ω is its boundary. We consider the following exterior Dirichlet problem 1 -K1∂2∂x2+K1∂2∂y2+K2∂2∂z2u=0,inΩ,u=g,onΓ0,u→0asr→∞, where K1 and K2 are two different anisotropic parameters, g is a given function that satisfies g∈H1/2(Γ0), and r=x2+y2+z2. The third item of Eq. (1) keeps the existence and uniqueness of the solution. Let Γ1={(x,y,z):x2+y2d2+z2c2=1,c>d>0} and Γ2={(x,y,z):x2+y2b2+z2a2=1,a>b>0} denote two artificial prolate spheroids. For clarity, we must mention that d>b and c>a. This means that Γ2 is totally inside Γ1. Define Ω2 as the unbounded domain outside the boundary Γ2 and Ω1 be a bounded domain between Γ0 and Γ1 (see Fig. 1).Fig. 1 Domain participation According to DDM (Brenner and Scott 1996), we construct the Schwarz alternating method as follows: 2 -K1∂2∂x2+K1∂2∂y2+K2∂2∂z2u1(2k+1)=0,inΩ1,u1(2k+1)=u2(2k),onΓ1,u1(2k+1)=g,onΓ0, and 3 -K1∂2∂x2+K1∂2∂y2+K2∂2∂z2u2(2k+2)=0,inΩ2,u2(2k+2)=u1(2k+1),onΓ2,u2(2k+2)→0,asr→∞, where k=0,1,… and u2(0)=u~. Setting the initial value of function u2(0) on boundary Γ1 as u2(0)|Γ1=0. Hence, we can solve the problem (2). Furthermore, with the limitation of u1(1) on Γ2, one solves the problem (3). Sequentially, we solve the problem in Ω1 again with substituting the value of solution u2(2) on Γ1. Then , we repeat the steps for k=1,2,… and so on. By the above description, obviously, we applied FEM in the problem over Ω1 and BEM (Feng and Yu 1983) in Ω2. Before using BEM to solve problem (3), the following transformation is introduced. 4 x=K1x1,y=K1y1,z=K2z1. For simplicity, the corresponding signals under the coordinate system (x1,y1,z1) can be defined by adding an apostrophe on the original ones, e.g. Ω→Ω′. Therefore, problem (3) can be expressed as the harmonic problem according to the new coordinate system. 5 -∂2∂x12+∂2∂y12+∂2∂z12u2(2k+2)=0,inΩ2′,u2(2k+2)=u1(2k+1),onΓ2′,u2(2k+2)→0,asr′→∞, We introduce the prolate spheroidal coordinates (μ,θ,φ), such that Γ2′ coincides with the prolate spheroid μ=μ2 and Ω2′={(μ,θ,φ)|μ>μ2>0,θ∈[0,π],φ∈[0,2π]}. 6 x1=fsinhμsinθcosφ,μ≥μ2>0,y1=fsinhμsinθsinφ,θ∈[0,π],z1=fcoshμcosθ,φ∈[0,2π], where f=a2K2-b2K1, a=fcoshμ2 and b=fsinhμ2. For simplicity, the problem (5) can be expressed as 7 -Δu=0,inΩ2′,u=u1,onΓ2′,u→0,asr′→∞. By the separation of variable (Zhang and Jin 1996), we have the solution of (7) as follows 8 u(μ,θ,φ)=∑n=0∞∑m=-nnQnm(coshμ)Qnm(coshμ2)UnmYnm(θ,φ)≡H(u2,μ,θ,φ),μ≥μ2>0, where Unm=∫02π∫0πu2(μ2,θ,φ)Ynm∗(θ,φ)sin(θ)dθdφ,Ynm∗=(-1)mYnm(θ,φ)=(-1)m2n+14π(n-m)!(n+m)!Pnm(cos(θ))eimφ. Pnm and Qnm are the first and second kind of the associated Legendre functions. Therefore, the solution u of (7) restricted on Γ1′ can be expressed as u(μ1,θ,φ)=H(u2,μ1,θ,φ). Similarly, we have the equivalent problem of (2). Thus, the Schwarz alternating algorithm can be expressed as follows: 9 -Δu1(2k+1)=0,inΩ1′,u1(2k+1)=g′,onΓ0′,u1(2k+1)=u2(2k),onΓ1′, and 10 -Δu2(2k+2)=0,inΩ2′,u2(2k+2)=u1(2k+1),onΓ2′,u2(2k+2)→0,asr′→∞. where k=0,1,…. The detail is similar to the original. Convergence of the algorithm We define the following spaces W01(Ω′)=vv1+x12+y12+z12∈L2(Ω′);∂v∂x1,∂v∂y1,∂v∂z1∈L2(Ω′),W˚01(Ω′)={v∈W01(Ω′)|v|Γ0′=0}. Solutions of (9) and (10) are in V1=H01(Ω1′) and V2=W˚01(Ω2′), respectively. Moreover, we denote the W˚01(Ω′) as V. Both functions of V1 and V2 can be extended into V. For example, we can extend u∈V1 by zero in Ω′\Ω1′ to a function in V. Hence, we have the equivalent variational form of (5): 11 Findw=u-u~∈W˚01(Ω′),such thatDΩ′(w,v)=-DΩ′(u~,v),∀v∈W˚01(Ω′), where DΩ′(u,v)=∫Ω′∇u·∇vdx1dy1dz1, u~∈W01(Ω′) has compact support and u~|Γ0′=g. |u|1=DΩ′(u,u) is an equivalent norm of W˚01(Ω′). If g∈H12(Γ0′), then there exists u~ such that the solution of (11) exists and is uniquely determined. Then (9) and (10) are equivalent to the following variational problems: 12 Findw1(2k+1)=u1(2k+1)-u(2k)|Ω1′∈V1,such thatDΩ1′(w1(2k+1),v)=-DΩ1′(u(2k),v),∀v∈V1, and 13 Findw2(2k+2)=u2(2k+2)-u(2k+1)|Ω2′∈V2,such thatDΩ2′(w2(2k+2),v)=-DΩ2′(u(2k+1),v),∀v∈V2. Let u(2k+1)=u1(2k+1),inΩ1′u2(2k),inΩ′\Ω1′,u(2k+2)=u1(2k+1),inΩ′\Ω2′u2(2k+2),inΩ2′, and u(0)=u~, then we have DΩ′(u-u(2k+1),v1)=0,∀v1∈V1,DΩ′(u-u(2k+2),v2)=0,∀v2∈V2. Noticing u(2k+1)-u(2k)∈V1,u(2k+2)-u(2k+1)∈V2 and u-u(2k+1)∈V,u-u(2k+2)∈V, Hence, 14 u(2k+1)-u(2k)=PV1(u-u(2k)),u(2k+2)-u(2k+1)=PV2u-u(2k+1) where PVi:V→Vi(i=1,2) means the projection operator under the inner product DΩ′(·,·) in V. Thus (14) is equivalent to 15 u-u(2k+1)=PV1⊥(u-u(2k)),u-u(2k+2)=PV2⊥(u-u(2k+1)). Denote the errors as ei(k)=u-u(k)(i=1,2). This leads to e1(2k+1)=PV1⊥PV2⊥e1(2k-1),e2(2k+2)=PV2⊥PV1⊥e2(2k), This implies that, if {e1(2k+1)} and {e2(2k)} are convergent, then their limits are in V1⊥∩V2⊥. Similar to the proofs given in Yu (1994, 2002); Luo et al. (2013) we can show the following result. Theorem 1 There exists a constantα, 0≤α<1, such that e1(2k+1)1≤α2ke1(1)1,e2(2k+2)1≤α2k+2e2(0)1. It is obvious to conclude α keeps the convergence of Schwarz alternating method. In the next section, we will prove the contraction factor α. Analysis of the convergence rate By Theorem 1, one may find the convergence rate of the above Schwarz alternating algorithm is closely related to the contraction factor α, i.e. the overlapping extent of Ω1′ and Ω2′. Although it can be deduced intuitively that the larger the overlapping part is, the faster convergence rate will be, yet we find it difficult to analyse the convergence rate for general unbounded domain Ω′. However, under certain assumptions, we can find out the relationship between contraction factor α and overlapping extent of Ω1′ and Ω2′. We define three prolate spheroids with the same semi-interfocal distance 16 Γi′={(μ,θ,φ):μ=μi,θ∈[0,π],φ∈[0,2π]},i=0,1,2, where μ1>μ2>μ0>0. We consider the following boundary value problem over domain Ω1′ 17 -Δu=0,inΩ1′,u=g0,onΓ0′,u=g1,onΓ1′. Suppose that 18 gi(θ,φ)=∑n=0+∞∑m=-nnGnm(i)Ynm(θ,φ),i=0,1, where Gnm(i)=∫0π∫02πgi(θ,φ)Ynm∗(θ,φ)sin(θ)dθdφ,i=0,1. Then by the separation of variables, we can obtain the solution of (17) 19 u(μ,θ,φ)=∑n=0+∞∑m=-nnS(μ,μ1)Gnm(0)+S(μ0,μ)Gnm(1)S(μ0,μ1)Ynm(θ,φ), where S(x,y)=Pnm(coshx)Qnm(coshy)-Pnm(coshy)Qnm(coshx). According to the property of the associated Legendre functions (Gradshteyn and Kyzhik 1980), we have the following lama. Lemma 1 Let Pnm(x)=dn+mdxn+m(x2-1)n, wheren, mare both nonnegative integers. If0≤m<n, thenPnm(x)hasn-mdifferent zeros-1=α1≤α2≤⋯≤αn-m=1withαi=-αn-m-(i-1),i=1,…,n-m-1. Lemma 2 Ifμ>μ0, then we conclude 20 Pnm(coshμ0)Pnm(coshμ)<coshμ0coshμn, and 21 Qnm(coshμ)Qnm(coshμ0)<coshμ0coshμn. Proof By the definition of Pnm(x) we have Pnm(coshμ0)Pnm(coshμ)=sinhμ0sinhμm-2∏i=1n-m(coshμ0-αi)∏i=1n-m(coshμ-αi). For monotonicity, the following holds for i=1,2,⋯,n-m, (coshμ0-αi)(coshμ0-αn-m-i+1)(coshμ-αi)(coshμ-αn-m-i+1)=(cosh2μ0-αi2)(cosh2μ-αi2)<cosh2μ0cosh2μ. Hence, Pnm(coshμ0)Pnm(coshμ)<coshμ0coshμn. □ On the other hand, (21) can be easily proved by the proposition of Huang and Yu (2006), Theorem 2 Supposeg0is continuous onΓ0and (16) holds. If we apply the Schwarz alternating algorithm given in “Schwarz alternating algorithm based on NBR”section, then 22 supΩ¯1|u-u(2k+1)|≤C1αk and 23 supΩ¯2|u-u(2k+2)|≤C2αk+1 hold true, the constantCi(i=1,2)depend only ong0andQnm(coshμi)Qnm(coshμ0)while 24 0<α=Qnm(coshμ1)S(μ0,μ2)Qnm(coshμ2)S(μ0,μ1)<1. Proof Similar to (8), so the solution of the unbounded problem outside of Γ0 can be expressed as u(μ,θ,φ)=∑n=0∞∑m=-nnQnm(coshμ)Qnm(coshμ0)Gnm(0)Ynm(θ,φ),μ≥μ0. Let u~=0. By using the algorithm, one has u(μ,θ,φ)-u(2k+1)(μ,θ,φ)=∑n=0+∞∑m=-nnQnm(coshμ1)Qnm(coshμ0)Qnm(coshμ1)S(μ0,μ2)Qnm(coshμ2)S(μ0,μ1)kS(μ0,μ)S(μ0,μ1)Gnm(0)Ynm(θ,φ), where μ0≤μ≤μ1.□ By defining α=Qnm(coshμ1)S(μ0,μ2)Qnm(coshμ2)S(μ0,μ1), we will show (24). From Lemma 2, we have T(μ)>coshμcoshμ02n>1,μ>μ0, and T(μ1)T(μ2)>coshμ1coshμ22n>1, where T(μ) is defined as T(μ)=Pnm(coshμ)Qnm(coshμ0)Pnm(coshμ0)Qnm(coshμ). Since α=T(μ2)-1T(μ1)-1=1+T(μ2)-T(μ1)T(μ1)-1, we obtain 0<α<1. Hence, (22) is accomplished. Obviously, (23) can be proved with similar process. Finally, the theorem is proved. Remark The convergence is related on the overlapping part of Ω1′ and Ω2′. From Theorem 2, we conclude the larger the overlapping part is, the smaller the contraction factor α will be, which identically means the faster the Schwarz alternating algorithm converging. The error estimates of the algorithm Denote Sh(Ω1′) as the linear finite element space over Ω1′, where the elements are partitioned as tetrahedrons. Let S˚h(Ω1′)=vh∈Sh(Ω1′)|vh|Γ0′∪Γ1′=0. S˚h(Ω1′) can be regarded as the subspace of V by zero extension. Therefore, we have the discrete Schwarz alternating algorithm as 25 Findw1h(2k+1)=u1h(2k+1)-uh(2k)|Ω1′∈S˚h(Ω1′)such thatDΩ1′(w1h(2k+1),vh)=-DΩ1′(uh(2k),vh),∀vh∈S˚h(Ω1′), and 26 Findw2h(2k+2)=u2h(2k+2)-uh(2k+1)|Ω2′∈V2such thatDΩ2′(w2h(2k+2),v)=-DΩ2′(uh(2k+1),v),∀vh∈V2, where uh(2k+1)=u1h(2k+1),inΩ1′uh(2k),inΩ′\Ω1′,uh(2k+2)=uh(2k+1),inΩ′\Ω2′u2h(2k+2),inΩ2′, and uh(0)=u~. By Yu (2002), the solution of (26) can be written as 27 u2h(2k+2)=Pγuh(2k+1), where P:H12(Γ2′)→W01(Ω2′) denotes Poisson integral operator and γ:H1(Ω1′)→H12(Γ2′) denotes trace operator. Combining with (27) and the discrete algorithm, one can easily have the following iteration value: uh(2k+1)=u~+∑i=0kw1h(2i+1),onΩ′¯\Ω2′∑i=0kw1h(2i+1)+∑j=0k-1Pγw1h(2j+1)-w1h(2j+1)+δk(Pγu~-u~),inΩ1′\(Ω′¯\Ω2′),∑j=0k-1Pγw1h(2j+1)+δk(Pγu~-u~),onΩ′\Ω1′, and uh(2k+2)=u~+∑i=0kw1h(2i+1),onΩ′¯\Ω2′∑i=0kw1h(2i+1)+∑j=0k[Pγw1h(2j+1)-w1h(2j+1)]+(Pγu~-u~),inΩ1′\(Ω′¯\Ω2′),∑j=0k-1Pγw1h(2j+1)+(Pγu~-u~),onΩ′\Ω1′, where δk=0,ifk=0,1,ifk>0. The term ∑j=0k-1 vanishes at k=0. Set Ah(Ω2′)=Pγ(vh+αu~+βw)-(vh+αu~+βw)|Ω′¯2|vh∈S˚h(Ω1′),α,β∈R,w=u-u~. Similarly, we have the Ah(Ω2′) as the subspace of V. Hence, Ah(Ω2′)⊂V2⊂V. We have the following variational problem on the discrete space 28 Findvh∗∈S˚h(Ω1′)+Ah(Ω2′)such thatDΩ′(vh∗,vh)=-DΩ′(u~,vh),∀vh∈S˚h(Ω1′)+Ah(Ω2′). Obviously, the solution of (28) exists uniquely . Set uh∗=vh∗+u~. Similarly in Wu and Yu (2000b), we have the following error estimates. Theorem 3 For the discrete Schwarz alternating algorithm and the constantαinTheorem 1, the following error estimates hold |u-uh(2k+1)|1≤Ch+α2k|uh∗-uh(1)|1,|u-uh(2k+2)|1≤Ch+α2k+2|uh∗-uh(0)|1. Numerical results Some numerical examples are computed to show the efficiency of our algorithm in this section. Using the method developed in “Schwarz alternating algorithm based on NBR” section. The linear elements is used in the computation of FEM. Computationally, we consider on three meshes: Mesh I, Mesh II and Mesh III. Each mesh is a refinement of its former one, especially as Mesh I is the primary. The refinement is defined as each of elements of the former mesh is divided into eight similar shape equally. e and eh denote the maximal error of all node functions on Γ1h, respectively, i.e., e(k)=supPi∈Ω1hu(Pi)-u1h(2k+1)(Pi),eh(k)=supPi∈Ω1hu1h(2k-1)(Pi)-u1h(2k+1)(Pi). qh(k) is the rate of convergence, i.e. qh(k)=eh(k-1)eh(k). Moreover, we use the relative maximum norm (‖Eu‖∞) of the errors between numerical solutions and the exact solutions: ‖Eu‖∞=|u-uh|∞,Ω1|u|∞,Ω1. Example 1 Set the cubic Ω={(x,y,z)||x|≤1,|y|≤1,|z|≤3} and Γ0 be its surface of Ω. The exact solution of problem (5) be u=x/K1((x2+y2)/K1+z2/K2)3/2. Also g=u|Γ0. By the theoretical analysis, we take two confocal prolate ellipsoidal surfaces as artificial boundaries, which can be expressed as Γ1={(μ,θ,φ)|μ1=1.5,θ∈[0,π],φ∈[0,2π]} and Γ2={(μ,θ,φ)|μ2=1.25,θ∈[0,π],φ∈[0,2π]}. And the semi-interfocal distance f1=f2=6. Moreover, we have K1=1 and K2=3. The efficient results are the case in Tables 1, 2 and Fig. 2.Table 1 The relation between convergence rate and mesh: μ1=1.5, μ2=1.25 Mesh k Number of iteration and corresponding values 0 1 2 3 4 5 I e 2.4726E−1 9.0403E−2 5.4826E−2 8.0814E−3 8.0782E−3 8.0774E−3 eh – 2.8013E−2 3.6179E−3 7.2392E−4 1.5669E−4 3.6362E−4 qh – – 77.4294 4.9977 4.6200 4.3092 II e 8.6794E−2 4.0215E−3 3.1259E−5 2.9243E−5 2.9104E−5 2.9100E−5 eh – 1.0366E−4 3.4624E−6 3.1645E−7 2.8591E−7 2.8503E−7 qh – – 29.9437 10.9409 1.1068 1.0031 III e 1.6827E−3 9.2546E−4 7.4972E−5 7.4802E−5 7.4792E−5 7.4753E−5 eh – 9.2858E−4 7.6389E−5 6.6424E−6 5.9675E−6 5.5203E−6 qh – – 12.1564 11.5004 1.1131 1.0817 Table 2 The relation between convergence rate and overlapping degree (Mesh II) μ1 μ2 k Number of iteration and corresponding values 0 1 2 3 4 5 1.5 1.2 e 6.4728E−2 4.6532E−3 3.4571E−5 2.6119E−5 2.6084E−5 2.6002E−5 eh – 2.0222E−3 1.2045E−4 4.5076E−5 9.0874E−6 9.0244E−6 qh – – 16.7890 3.8033 4.9290 1.0660 1.5 1.0 e 4.5186E−2 1.0521E−3 9.0705E−5 5.4413E−5 1.2218E−5 1.2103E−5 eh – 1.3736E−3 4.8967E−5 2.6640E−7 1.4184E−7 7.5349E−7 qh – – 28.0516 18.3810 2.7813 2.8248 1.5 0.8 e 1.4825E−3 6.7734E−4 9.2125E−5 1.8249E−5 5.6719E−6 5.5017E−6 eh – 6.4936E−4 2.1429E−5 1.2093E−6 8.2674E−8 1.0827E−8 qh – – 30.3022 17.7197 14.62807 7.6359 Fig. 2 Maximal errors in relative maximum norm From Table 1, we can see the convergence is really fast. Both e and eh are smaller than them on former mesh. And the Fig. 2 shows us the errors converge rapidly. Both of them reveal that the fine the mesh, the faster the convergence. The numbers of Table 2 testify the remark in “The error estimates of the algorithm” section. By taking different μ1 and μ2, we chose 3 couples of artificial boundaries. Geometrically, the bigger the |μ1-μ2|, the bigger the overlapping domain. Within the same triangular partition (Mesh II), we conclude that the bigger the overlapping domain, the faster the convergence. Example 2 Generally, the Ω is chosen as a prolate ellipsoidal. Set the semi-interfocal f0=4 and Γ0={(μ,θ,φ)|μ0=0.5,θ∈[0,π],φ∈[0,2π]}. Set K1=K2=1. Thus, the exact solution of problem (5) is u=1((x2+y2)/K1+z2/K2)1/2. and g=u|Γ0. Similarly, we choose two artificial boundaries Γ1 and Γ2, which are both confocal with Γ0=∂Ω as f1=f2=f0=6. Let Γ1={(μ,θ,φ)|μ1=2.5,θ∈[0,π],φ∈[0,2π]} and Γ2={(μ,θ,φ)|μ2=2.0,θ∈[0,π],φ∈[0,2π]}. The corresponding results are the case in Tables 3, 4 and Fig. 3.Table 3 The relation between convergence rate and mesh: μ1=2.5, μ2=2.0 Mesh k Number of iteration and corresponding values 0 1 2 3 4 5 I e 2.1078E−2 8.4562E−3 5.9623E−3 4.6782E−3 4.6511E−3 4.6407E−3 eh 9.0022E−4 3.0713E−5 2.1630E−6 1.5593E−6 1.1858E−6 qh 29.3106 14.1992 1.3871 1.3150 II e 8.3741E−3 7.6501E−3 4.6829E−3 9.4296E−4 8.6241E−4 8.5788E−4 eh – 7.7637E−4 1.4383E−6 3.7605E−8 9.6070E−9 2.4529E−9 qh – – 53.9787 38.2471 3.9143 3.9166 III e 1.8257E−3 5.4865E−4 4.2731E−5 3.5722E−5 3.5605E−5 3.5592E−5 eh – 1.0350E−6 5.2502E−9 1.2387E−10 3.6938E−11 5.0933E−11 qh – – 197.1280 51.8669 11.4751 6.2403 Table 4 The relation between convergence rate and overlapping degree (Mesh II) μ1 μ2 k Number of iteration and corresponding values 0 1 2 3 4 5 2.5 1.8 e 7.4537E−3 8.6547E−4 4.6829E−4 9.5781E−5 8.7710E−5 8.7058E−5 eh – 6.0775E−7 4.7353E−8 5.3837E−9 6.2859E−10 5.6858E−10 qh – – 12.8344 8.7955 8.5647 1.1055 2.5 1.6 e 2.4832E−3 7.6489E−4 5.4952E−5 3.6848E−5 2.6981E−5 2.6773E−5 eh – 2.9321E−7 1.1713E−8 5.8642E−10 2.8518E−10 2.1763E−10 qh – – 25.0324 19.9742 2.0563 1.3104 2.5 1.4 e 5.4377E−4 7.6811E−5 6.8129E−6 8.1056E−7 8.0859E−7 8.05378E−7 eh – 4.2367E−7 6.0310E−9 1.0814E−10 1.9075E−11 9.2494E−12 qh – – 70.2475 55.76912 5.6694 2.06226 Fig. 3 Maximal errors in relative maximum norm The data of Tables 3 and 4 show us a good convergence. And the analysis of the numbers can be similar to Example 1. Conclusions In this paper, we construct a Schwarz alternating algorithm for the anisotropic problem on the unbounded domain. The algorithm uses the DDM based on FEM and natural boundary element method. The theoretical analysis shows its convergence is first-order. Further, the rate of convergence is dependent on the overlapping domain. Some numerical examples testify the theoretical conclusions. We can investigate the Schwarz alternating algorithm for anisotropic problem with three different parameters over unbounded domain. Full details and results will be given in a future publication. Authors’ contributions All authors completed this paper together. All authors read and approved the final manuscript. Acknowledgements All authors are greatly indebted to the referees as the valuable suggestions and comments.This work was subsidized by the National Natural Science Foundation of China (11371198, 11401296), Jiangsu Provincial Natural Science Foundation of China (BK20141008), Natural science fund for colleges and universities in Jiangsu Province (14KJB110007). Competing interests The authors declare that they have no competing interests. ==== Refs References Brenner SC Scott LR The mathematical theory of finite element methods 1996 Berlin Springer Costabel M Brebbia CA Symmetric methods for the coupling of finite elements and boundary elements Boundary elements IX 1987 Berlin Springer 411 420 Das P Comparison of a priori and a posteriori meshes for singularly perturbed nonlinear parameterized problems J Comput Appl Math 2015 290 16 25 10.1016/j.cam.2015.04.034 Das P Mehrmann V Numerical solution of singularly perturbed convection–diffusion–reaction problems with two small parameters BIT Numer Math 2016 56 51 76 10.1007/s10543-015-0559-8 Das P Natesan S Optimal error estimate using mesh equidistribution technique for singularly perturbed system of reaction–diffusion boundary value problems Appl Math Comput 2014 249 265 277 Du Q Yu D On the natural integral equation for initial boundary value problems of two dimensional hyperbolic equation Acta Math Appl Sin 2001 24 17 26 Evans D Numerical solution of exterior problems by the peripheral block over-relaxation method IMA J Appl Math 1977 19 399 405 10.1093/imamat/19.4.399 Feng K Differential versus integral equations and finite versus infinite elements Math Numer Sin 1980 2 100 105 Feng K (1983) Finite element method and natural boundary reduction. In: Proceedings of the international congress of mathematicians, Warszawa, pp 1439–1453 Feng K, Yu DH (1983) Canonical integral equations of elliptic boundary value problems and their numerical solutions. In: Proceedings of China–France symposium on the finite element method. Science Press, Beijing, 1983, pp 211–252 Gatica G Gatica L Stephan E A FEM-DtN formulation for a non-linear exterior problem in incompressible elasticity Math Methods Appl Sci 2003 26 151 170 10.1002/mma.349 Givoli D Keller J A finite element method for large domains Comput Methods Appl Mech Eng 1989 76 41 66 10.1016/0045-7825(89)90140-0 Gradshteyn IS Kyzhik IM Table of integrals, series, and products 1980 8 London Academic Press Grote MJ Keller JB On non-reflecting boundary conditions J Comput Phys 1995 122 231 243 10.1006/jcph.1995.1210 Han H A new class of variational formulations for the coupling of finite and boundary element methods J Comput Math 1990 8 223 232 Hsiao GC Porter JF Brebbia CA A hybrid method for an exterior boundary value problem based on asympototic expansion, boundary integral equation and finite element approximations Innovative numerical methods in engineering 1986 Berlin Springer 83 88 Hu Q Yu D A solution method for a certain nonlinear interface problem in unbounded domains Computing 2001 67 119 140 10.1007/s006070170010 Huang H Liu D Yu D Solution of exterior problem using ellipsoidal artificial boundary J Comput App Math 2009 231 434 446 10.1016/j.cam.2009.03.009 Huang H Yu D Natural boundary element method for three dimensional exterior harmonic problem with an inner prolate spheroid boundary J Comput Math 2006 24 193 208 Keller JB Givoli D An exact non-reflection boundary conditions J Comput Phys 1989 82 172 192 10.1016/0021-9991(89)90041-7 Koyama D Error estimates of the DtN finite element method for the exterior Helmholtz problem J Comput Appl Math 2007 200 21 31 10.1016/j.cam.2005.12.004 Koyama D Error estimates of the finite element method for the exterior Helmholtz problem with a modified DtN boundary condition J Comput Appl Math 2009 232 109 121 10.1016/j.cam.2008.10.034 Luo X Du Q Huang H He T A Schwarz alternating algorithm for a three-dimensional exterior harmonic problem with prolate spheroid boundary Comput Math Appl 2013 65 1129 1139 10.1016/j.camwa.2013.02.004 MacCamy RC Marin SP A finite element method for exterior interface problems Int J Math Math Sci 1980 3 311 350 10.1155/S0161171280000233 Wendland WL Shaw RP On asymptotic error estimates for combined BEM and FEM Innovative numerical methods in engineering 1986 Berlin Springer 55 70 Wu JM Yu DH The natural integral equations of 3-D harmonic problems and their numerical solution Math Numer Sin 1998 4 419 430 Wu J Yu D The natural boundary element method for exterior elliptic domain Chin J Num Math Appl 2000 22 91 104 Wu J Yu D The overlapping domain decomposition method for harmonic equation over exterior three-dimensional domain J Comput Math 2000 18 83 94 Yu D A domain decomposition method based on the natural boundary reduction over unbounded domain Math Numer Sin 1994 16 448 459 Yu D Natural boundary integral method and its applications 2002 Dordrecht Kluwer Academic Zhang S Jin J Computation of special functions 1996 New York Wiley

PMC005xxxxxx/PMC5001970.txt

==== Front Struct DynStruct DynSDTYAEStructural Dynamics2329-7778American Crystallographic Association 10.1063/1.49616131.4961613025693SDYSD-FEMTO1216-00049Special Topic: The Hamburg Conference on Femtochemistry (Femto12)PrefacePreface to the Special Edition on Femtochemistry and “The Hamburg Conference on Femtochemistry 2015 (FEMTO12)” Mullins Terry 1Küpper Jochen 1,2,3a)1 Center for Free-Electron Laser Science, DESY, Notkestrasse 85, 22607 Hamburg, Germany2 Department of Physics, University of Hamburg, Luruper Chaussee 149, 22761 Hamburg, Germany3 The Hamburg Center for Ultrafast Imaging, University of Hamburg, Luruper Chaussee 149, 22761 Hamburg, Germanya) Email: jochen.kuepper@cfel.de. URL: https://www.femto12.org and https://www.controlled-molecule-imaging.org. 24 8 2016 7 2016 24 8 2016 3 4 04300113 7 2016 13 8 2016 © 2016 Author(s).2016Author(s)2329-7778/2016/3(4)/043001/3All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/). crossmark ==== Body This special issue is dedicated to the memory of Dr. Ahmed Zewail, the founder of the field and the public-keynote speaker of the Leuchtturm (Tower of Light) Lecture (https://www.femto12.org/leuchtturm) at FEMTO12. Femtochemistry addresses the very nature of the chemical bond. It aims at unraveling the ultrafast dynamics of bond forming and breaking. Molecules and atoms are the building blocks of nature, and their chemical dynamics define the evolution of our world. Chemical bonds are made by valence electrons and have energies of a few electron volts (eV). Chemical reactions are mostly initiated by heat. The huge driving forces of chemical reactions are the strong intramolecular Coulomb forces, mediated by valence electron and nuclear dynamics. Sometimes, light is used to trigger chemical reactions, often to mimick the sudden availability of thermal energy. Because light can be provided in ultrashort pulses with very high temporal precision, such pulses provide a well defined starting point of the chemical reaction and allow to clock the ongoing subsequent chemical dynamics.1 The resulting high temporal resolution makes ultrashort laser pulses a superb tool to investigate the ultrafast steps of chemical processes, i.e., femtochemistry. Through this approach to ultrafast chemical processes, femtochemistry crosses the traditional boundaries of physics, chemistry, and biology. Within its scope, a variety of questions arise: How do the building blocks of chemistry, such as atoms, molecules, and light, interact with one another at the most fundamental level and on the shortest time scales to drive both simple and complex systems from initial to final states? What is the role of the transition state2 and how can we directly observe it? How can we understand increasingly complex systems? How do the most rapid dynamics of the smallest sub-systems and particles initiate large scale changes in macroscopic systems on long timescales? How much correlation is contained in that extension? Can biological systems be understood in terms of their quantum properties? In how far can all these processes be controlled and driven to tailor desired processes and materials? New kids on the block for the investigation of these properties are the recently introduced short-wavelength free-electron lasers and ultrashort-pulse electron sources, which aim at recording so-called molecular movies with femtosecond temporal and atomic-resolution (100 pm) spatial resolution. Their application in femtochemistry and femtobiology was highlighted elsewhere.3–5 This special issue was initiated together with the Hamburg Conference on Femtochemistry (FEMTO12), which was the twelfth in a series of conferences that was started in Berlin in 1993. FEMTO12 was held in Hamburg-Bahrenfeld on 12–17 July 2015 and provided a platform for rich discussions on the ongoing experimental and theoretical work to disentangle and understand ultrafast molecular processes related to valence electron dynamics. This includes studies of these effects in molecules and clusters, liquids and solutions, biological systems, interfaces and surfaces, nanosystems, and polymers and condensed matter. The eleven papers in this issue highlight some of the wonderful research presented and the constructive discussions had at the conference. DNA is arguably one of the most important biological molecules. In nature, it is exclusively found in a hydrated environment and its interactions with water are, therefore, of great importance. The dynamics of hydrated DNA oligomers are investigated by Guchhait et al.6 Energy transport is an important process, which naturally lends itself to laser control techniques. Brüning et al.7 investigate the controlled localization of energy in photo-excited aggregates. Thallmair et al.8 discuss the role of conical intersections, which become more prominent as complexity increases, in the ultrafast branching in the photoinduced bond cleavage process of diphenylmethyl derivatives. Catalytic reactions are extremely important for many practical applications. Kunnis et al.9 investigate the spin cross-over and ligation of the Fe(CO)5 complex in solution upon photoexcitation. Bond breaking and bond rearrangement, the resulting transition states, as well as charge transfer processes belong to some of the most fundamental concepts in chemistry. Li et al.10 investigate the effect of the carrier envelope phase on dissociative photoionization of toluene. Stensitzki et al.11 characterize the photoreaction of hexacoordinated Al(tpfc-Br8)(py)2. Stensitzki et al.12 investigate photoisomerization products and timescales in channelrhodopsin-1. Boll et al.13 investigate ultrafast electron transfer in dissociating molecules. Kim et al.14 investigate the transition state structure of the [Au(CN)2−]3 trimer upon photoexcitation. Imanbaew et al.15 investigate the cation and monoanion of fluorescein in the gas phase and observe signatures of differing excited state structures. Spatial structure can enormously affect dynamics and interactions. Li and Vendrell16 investigate protonated water clusters after extreme UV photoionization and find site-dependent correlated proton-electron hole dynamics. Besides DESY, this work has been supported by the European Research Council through the Consolidator Grant COMOTION (ERC-614507), the Helmholtz Association “Initiative and Networking Fund,” and the Deutsche Forschungsgemeinschaft (DFG) through the excellence cluster “The Hamburg Center for Ultrafast Imaging—Structure, Dynamics and Control of Matter at the Atomic Scale” (CUI, EXC1074). ==== Refs 1. A. H. Zewail , “ Femtochemistry: Atomic-scale dynamics of the chemical bond ,” J. Phys. Chem. A , 5660 –5694 (2000 ).10.1021/jp001460h 2. H. Eyring and M. Polanyi , “ Über einfache Gasreaktionen ,” Z. Phys. Chem. B , 279 –311 (1931 ). 3. “Faraday discussion 171 on ‘Emerging photon technologies for chemical dynamics’ ,” Faraday Disc. , 1 –2 (2014 ).10.1039/C4FD90030H 4. G. Sciaini and R. J. Dwayne Miller , “ Femtosecond electron diffraction: Heralding the era of atomically resolved dynamics ,” Rep. Prog. Phys. , 096101 (2011 ).10.1088/0034-4885/74/9/096101 5. J. C. H. Spence and H. N. Chapman , “ The birth of a new field ,” Philos. Trans. R. Soc. B , 20130309 (2014 ).10.1098/rstb.2013.0309 6. B. Guchhait , Y. Liu , T. Siebert , and T. Elsaesser , “ Ultrafast vibrational dynamics of the DNA backbone at different hydration levels mapped by two-dimensional infrared spectroscopy ,” Struct. Dyn. , 043202 (2016 ).10.1063/1.4936567 26798841 7. C. Brüning , J. Wehner , J. Hausner , M. Wenzel , and V. Engel , “ Exciton dynamics in perturbed vibronic molecular aggregates ,” Struct. Dyn. , 043201 (2016 ).10.1063/1.4936127 26798840 8. S. Thallmair , M. K. Roos , and R. de Vivie-Riedle , “ Molecular features in complex environment: Cooperative team players during excited state bond cleavage ,” Struct. Dyn. , 043205 (2016 ).10.1063/1.4941600 26958588 9. K. Kunnus , I. Josefsson , I. Rajkovic , S. Schreck , M. Quevedo , W. Beye , C. Weniger , S. Grübel , M. Scholz , D. Nordlund , W. Zhang , R. W. Hartsock , K. J. Gaffney , W. F. Schlotter , J. J. Turner , B. Kennedy , F. Hennies , F. M. F. de Groot , S. Techert , M. Odelius , Ph. Wernet , and A. Föhlisch , “ Identification of the dominant photochemical pathways and mechanistic insights to the ultrafast ligand exchange of Fe(CO)5 to Fe(CO)4EtOH ,” Struct. Dyn. , 043204 (2016 ).10.1063/1.4941602 26958587 10. H. Li , N. G. Kling , B. Förg , J. Stierle , A. Kessel , S. A. Trushin , M. F. Kling , and S. Kaziannis , “ Carrier-envelope phase dependence of the directional fragmentation and hydrogen migration in toluene in few-cycle laser fields ,” Struct. Dyn. , 043206 (2016 ).10.1063/1.4941601 26958589 11. T. Stensitzki , Y. Yang , A. Berg , A. Mahammed , Z. Gross , and K. Heyne , “ Ultrafast electronic and vibrational dynamics in brominated aluminum corroles: Energy relaxation and triplet formation ,” Struct. Dyn. , 043210 (2016 ).10.1063/1.4949363 27226980 12. T. Stensitzki , Y. Yang , V. Muders , R. Schlesinger , J. Heberle , and K. Heyne , “ Femtosecond infrared spectroscopy of channelrhodopsin-1 chromophore isomerization ,” Struct. Dyn. , 043208 (2016 ).10.1063/1.4948338 27191011 13. R. Boll , B. Erk , R. Coffee , S. Trippel , T. Kierspel , C. Bomme , J. D. Bozek , M. Burkett , S. Carron , K. R. Ferguson , L. Foucar , J. Küpper , T. Marchenko , C. Miron , M. Patanen , T. Osipov , S. Schorb , M. Simon , M. Swiggers , S. Techert , K. Ueda , C. Bostedt , D. Rolles , and A. Rudenko , “ Charge transfer in dissociating iodomethane and fluoromethane molecules ionized by intense femtosecond X-ray pulses ,” Struct. Dyn. , 043207 (2016 ).10.1063/1.4944344 27051675 14. K. H. Kim , J. G. Kim , K. Y. Oang , T. W. Kim , H. Ki , J. Jo , J. Kim , T. Sato , S. Nozawa , S.-I. Adachi , and H. Ihee , “ Femtosecond X-ray solution scattering reveals that bond formation mechanism of a gold trimer complex is independent of excitation wavelength ,” Struct. Dyn. , 043209 (2016 ).10.1063/1.4948516 27191012 15. D. Imanbaew , M. F. Gelin , and C. Riehn , “ Rotational and vibrational dynamics in the excited electronic state of deprotonated and protonated fluorescein studied by time-resolved photofragmentation in an ion trap ,” Struct. Dyn. , 043211 (2016 ).10.1063/1.4953367 27376104 16. Z. Li and O. Vendrell , “ Correlated proton-electron hole dynamics in protonated water clusters upon extreme ultraviolet photoionization ,” Struct. Dyn. , 043203 (2016 ).10.1063/1.4939897 26798842

PMC005xxxxxx/PMC5002025.txt

==== Front Front PhysiolFront PhysiolFront. Physiol.Frontiers in Physiology1664-042XFrontiers Media S.A. 10.3389/fphys.2016.00372PhysiologyOriginal ResearchValidation of a Ramp Running Protocol for Determination of the True VO2max in Mice Ayachi Mohamed Niel Romain Momken Iman Billat Véronique L. Mille-Hamard Laurence *Unité de Biologie Intégrative des Adaptations à l'Exercice, Université d'Evry Val d'EssonneEvry, FranceEdited by: Gary Iwamoto, University of Illinois at Urbana–Champaign, USA Reviewed by: Amanda Nelson, University of Wisconsin–Green Bay, USA; Thomas Lowder, University of Central Arkansas, USA *Correspondence: Laurence Mille-Hamard laurence.hamard@univ-evry.frThis article was submitted to Exercise Physiology, a section of the journal Frontiers in Physiology 29 8 2016 2016 7 37224 5 2016 12 8 2016 Copyright © 2016 Ayachi, Niel, Momken, Billat and Mille-Hamard.2016Ayachi, Niel, Momken, Billat and Mille-HamardThis is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.In the field of comparative physiology, it remains to be established whether the concept of VO2max is valid in the mouse and, if so, how this value can be accurately determined. In humans, VO2max is generally considered to correspond to the plateau observed when VO2 no longer rises with an increase in workload. In contrast, the concept of VO2peak tends to be used in murine studies. The objectives of the present study were to determine whether (i) a continuous ramp protocol yielded a higher VO2peak than a stepwise, incremental protocol, and (ii) the VO2peak measured in the ramp protocol corresponded to VO2max. The three protocols (based on intensity-controlled treadmill running until exhaustion with eight female FVB/N mice) were performed in random order: (a) an incremental protocol that begins at 10 m.min−1 speed and increases by 3 m.min−1 every 3 min. (b) a ramp protocol with slow acceleration (3 m.min−2), and (c) a ramp protocol with fast acceleration (12 m.min−2). Each protocol was performed with two slopes (0 and 25°). Hence, each mouse performed six exercise tests. We found that the value of VO2peak was protocol-dependent (p < 0.05) and was highest (59.0 ml.kg 0.75.min−1) for the 3 m.min−2 0° ramp protocol. In the latter, the presence of a VO2max plateau was associated with the fulfillment of two secondary criteria (a blood lactate concentration >8 mmol.l−1 and a respiratory exchange ratio >1). The total duration of the 3 m.min−2 0° ramp protocol was shorter than that of the incremental protocol. Taken as a whole, our results suggest that VO2max in the mouse is best determined by applying a ramp exercise protocol with slow acceleration and no treadmill slope. VO2maxmiceexercise protocolcomparative physiologyperformance ==== Body Introduction Although rodents are often used as models in exercise physiology, there is no consensus on the use of a standardized exercise protocol for determining the maximum oxygen uptake (VO2max) in these species. In fact, the concept of peak oxygen consumption (VO2peak) is preferred in mice. Given that VO2max is the main determinant of performance in human exercise physiology (i.e., the greatest possible oxygen uptake during physical exercise involving a large proportion of the total muscle mass (Cohn, 1987), it remains to be established whether this concept is valid in the mouse and, if so, how VO2max can be accurately determined. It is widely acknowledged that VO2max in humans corresponds to both the cardiovascular system's functional limitation and the organism's aerobic capacity. Since the VO2max concept was introduced by Hill and Lupton (1923), the use of exercise protocols with progressive or stepped increments has been validated in human - although the optimal choice of exercise protocol is still subject to debate. In stepwise protocols, the height of the step (i.e., the magnitude of the increment) and the duration of each workload level are left to the investigator's discretion.Since the 1960s, a number of different incremental protocols (with variations in running speed, treadmill slope or both) have been tested for their reliability in determining VO2max (Balke and Ware, 1959; Bruce et al., 1963; Froelicher et al., 1974). In contrast, ramp protocol are characterized by a continuous, gradual increase in the workload (i.e., power, speed or slope) up to maximum values. Many researchers have compared incremental protocols with ramp protocols, in order to establish the most efficient method for determining VO2max (Whipp et al., 1981; Astorino et al., 2005; Yoon et al., 2007). These studies have shown that the ramp exercise protocol is well suited to the human's aerobic metabolism and thus enables VO2max to be accurately determined. However, ramp protocols take longer to complete, and incremental protocols are preferred for the routine measurement of VO2max because they allow other performance indicators (such as the ventilatory threshold and the lactate threshold) to be determined. In humans, VO2max is generally considered to correspond to the plateau observed when VO2 no longer increases with speed. However, about half of tested subjects do not reach a plateau before they abandon the protocol; secondary criteria then have to be used to establish when the last (peak) VO2 value indeed corresponds to VO2max. Three secondary criteria have been proposed: (i) the maximum heart rate at the end of the test, which corresponds to an estimate of the theoretical maximum (Åstrand, 1952; Astrand, 1960; Maritz et al., 1961); (ii) an end-of-exercise respiratory exchange ratio (RER) >1.15 (Issekutz et al., 1962); and (iii) an end-of-exercise blood lactate concentration >8 mmol.l−1. For the purposes of comparative physiology, VO2max has also been determined in rodents. This parameter can be used in studies of exercise training or in descriptive studies of genetically modified animals (Kemi et al., 2002; Hoydal et al., 2007; Mouisel et al., 2014). As in humans, the relationship between running intensity and oxygen uptake is linear in mice (as demonstrated during steady-state, fixed-intensity running (Fernando et al., 1993; Schefer and Talan, 1996; Wisløff et al., 2001); this enables the use of incremental protocols. However, various strains of mouse have been used, and an effect of strain on treadmill performance has been evidenced. FVB mice achieve high maximum and critical speeds during forced treadmill exercise (Lightfoot et al., 2001; Lerman et al., 2002; Billat et al., 2005). Furthermore, age (Schefer and Talan, 1996) gender (Hoydal et al., 2007) may affect VO2max. VO2peak decreases in old age, although female and male mice appear to have similar levels of performance (Kemi et al., 2002; Billat et al., 2005). Consequently, the disparities in the literature data on VO2peak can be explained (at least in part) by differences in age and strain. Although, the mouse has been widely used to study the biochemical and molecular adaptations to exercise, a number of different protocols have been applied; this may explain (at least in part) the broad range of values obtained for VO2peak. Furthermore, it has been reported that VO2peak in mice is slope-dependent (Kemi et al., 2002). The incremental protocols described in the literature differ in their duration, increment size and the criteria used to determine exhaustion (usually the animal's behavior or the shape of the VO2/time curve) (Dohm et al., 1994; Rezende et al., 2005; Hawkins et al., 2007). It is not known whether a ramp protocol is suitable for determining VO2peak in mice or whether this value is protocol-dependent. Kemi et al. (2002) were the first to estimate the animal's level of exhaustion by applying secondary criteria (i.e., the RER and blood lactate levels) (Kemi et al., 2002). However, the presence or absence of a VO2 plateau, the latter's characteristics and the relationship between VO2peak and VO2max have not previously been studied in the mouse. We hypothesized that VO2peak and VO2max in mice are protocol-dependent and that (as in humans) a ramp exercise protocol would be suitable for determining VO2max. Thus, the objective of the present study in mice was to determine whether (i) a continuous ramp protocol yielded a higher VO2peak than a stepwise, incremental protocol, and (ii) the VO2peak measured in the ramp protocol corresponded to VO2max. Methods Animal One-year-old male FVB mice (n = 8) were selected for use in this study by virtue of their high level of performance on a treadmill (Lerman et al., 2002). The mice were kept in a specific and opportunistic pathogen-free animal facility (CERFE, Genopole, Evry, France) at a temperature of 22°C and with light-dark cycles 12/12-h. The animals were fed a standard diet ad libitum. Our protocol was approved by our institutions Animal Care and Use Committee on Care and complied with the European Convention of the Council of Europe for the protection of vertebrate animals used for experimental and other scientific purposes. Familiarization Mice were familiarized with the single-lane, motorized treadmill (adjustable belt speed: 0–99.9 m.min−1; Columbus Instruments, Columbus, OH, USA) during four 10-min running sessions (at 0, 3, 6, and 9 m.min−1), with a 48-h interval between each session. All mice subsequently included in the study were able to run for the required time at 9 m.min−1. The running speed was not increased further, in order to avoid a training effect. The exercise protocol The treadmill was set up in a metabolic chamber. Three different protocols were applied: an incremental protocol (IP) with a starting speed of 10 m.min−1 and an increment of 3 m.min−1 every 3 min; a ramp protocol with a starting speed of 3 m.min−1 and an acceleration of 0.05 m.min−1.s−1 (corresponding to 3 m.min−2), hereafter referred to as “Ramp3”; and a ramp protocol with a starting speed of 3 m.min−1 and an acceleration of 0.2 m.min−1.s−1 (corresponding to 12 m.min−2), hereafter referred to as “Ramp12.” Each of the three protocols was performed with two different slopes (0 and 25°); hence, each mouse performed six sessions. To avoid conditioning bias, the test sequence was randomized and there was 24-h interval between each session. The exercise session lasted until exhaustion, which was defined as the mouse's inability to maintain running speed despite being in contact with the electrical grid for more than 5 consecutive seconds (Mille-Hamard et al., 2012). All mice were compliant in all tests. The resting blood lactate concentration was measured at the start of the test ([Lac]rest) and 2 min after the end of each run ([La]max). To this end, a blood drop was collected at the tail vein (using the tail snip method), placed on a test strip and inserted into a lactate analyzer (Lactate Pro, Arkray, Inc., Kyoto, Japan). Gas measurements Ambient air was fed through the metabolic chamber at a rate of 0.66 l.min−1; the flow was chosen such that the incoming vs. outgoing difference in O2 fraction was within the sensor's range of measurement (−0.3 to −0.8% O2). A fan was used to mix the incoming air with the air around the treadmill and blow it toward the animal. The air flowed from the front of the treadmill to the rear of the treadmill and then returned toward the front under the belt. This created a rapid, circular “loop” of mixed gases (i.e., incoming “fresh” air mixed with the accumulated, exhaled gases), from which a sample was drawn for analysis every 5 s. Samples were dried prior to measurement of the O2 and CO2 fractions. The gas analyzers were calibrated with standardized gas mixtures (Air Liquide Santé, Paris, France) before each test session, as recommended by the manufacturer. To allow rapid comparisons over a wide range of body weights (especially with human data), dimensional analyses and empirical studies have shown that VO2 should be divided by the body mass raised to the power of 0.75 (Taylor et al., 1981; Hoydal et al., 2007; Mille-Hamard et al., 2012). Data analysis VO2peak was defined as the highest observed value of VO2 when averaged over successive 15 s periods. VO2max was defined as in humans (i.e., the highest VO2peak value recorded during a set of different test protocols, and the occurrence of a VO2 plateau). The VO2 plateau was determined when the VO2 did not increase by more than 1% of the difference between the VO2 at rest and VO2peak over a 30 s period, despite an increase in running speed. The mouse's maximum speed (Vmax) was defined as the running speed at the end of the protocol. The RER was defined as the ratio between the amount of oxygen (O2) consumed and the amount of carbon dioxide (CO2) produced in the metabolic chamber. The maximum respiratory exchange ratio (RERmax) was defined as the highest observed value of the RER when averaged over successive 15 s periods. Statistics Data are expressed as the mean ± standard deviation (SD). Statistical analysis was carried out with a two-way repeated measures ANOVA, followed by a Holm-Sidak post-hoc test. The threshold for statistical significance was set to p < 0.05. All statistical analyses were performed using STATISTICA software (version 9.0, Statsoft, Berkeley, CA, USA). Results VO2peak in each exercise protocol The highest observed VO2peak (59.0 ± 0.61 ml.kg−0.75.min−1, Figure 1) was obtained during the Ramp3 0° protocol. This value was significantly greater than those obtained in the other protocols. The presence of a slope influenced the value of VO2peak, which was higher in IP 25° than in IP 0° but lower in Ramp3 25° and Ramp12 25° than in Ramp3 0° and Ramp12 0°. The minimum VO2 determined at the beginning of the protocol (referred to as the VO2 at rest) was essentially the same in all protocols (mean: 43.6 ± 3.9 ml.kg−0.75.min−1). Figure 1 VO2peak in each exercise protocol : one-year-old sedentary FVB/N mice (n = 8) performed six exhaustive exercise protocols with a treadmill slope of 25 or 0°. IP, an incremental protocol with a starting speed of 10 m.min−1 and an increment of 3 m.min−1 every 3 min; Ramp3, a ramp protocol with a starting speed of 3 m.min−1 and an acceleration of 3 m.min−2 (0.05 m.min−1.s−1); Ramp12, a ramp protocol with a starting speed of 3 m.min−1 and an acceleration of 12 m.min−2 (0.2 m.min−1.s−1); §, a significant difference between 25 and 0° for the same protocol (p < 0.05); *, differs significantly from all other protocols (p < 0.05). Observation of a VO2peak plateau as a function of the exercise protocol As shown in Table 1, all mice displayed a VO2 plateau for at least 30 s during the Ramp3 0° and IP 25° protocols (mean plateau duration: 57.5 s ± 11.3 and 75 ± 11.24 s, respectively). During other protocols, some (but not all) mice reached a VO2 plateau for at least 30 s (Table 1) Table 1 Percentages of mice reaching a VO2 plateau for least 30 s, as defined in the Methods section. IP (%) Ramp3 (%) Ramp12 (%) Slope of 0° 87.5 100 87.5 Slope of 25° 100 75 87.5 IP, an incremental protocol with a starting speed of 10 m.min−1 and an increment of 3 m.min−1 every 3 min; Ramp3, a ramp protocol with a starting speed of 3 m.min−1 and an acceleration of 3 m.min−2 (0.05 m.min−1.s−1); Ramp12, a ramp protocol with a starting speed of 3 m.min−1 and an acceleration of 12 m.min−2 (0.2 m.min−1.s−1). Maximal respiratory exchange ratio: RERmax There were no inter-test differences in RERmax (Figure 2). For Ramp3 0°, the mean RERmax value was 1.06 ± 0.01, and RERmax was greater than 1.05 for seven of the eight mice. Figure 2 RERmax in each exercise protocol. IP, an incremental protocol with a starting speed of 10 m.min−1 and an increment of 3 m.min−1 every 3 min; Ramp3, a ramp protocol with a starting speed of 3 m.min−1 and an acceleration of 3 m.min−2 (0.05 m.min−1.s−1); Ramp12, a ramp protocol with a starting speed of 3 m.min−1 and an acceleration of 12 m.min−2 (0.2 m.min−1.s−1). Maximum blood lactate concentration [La]max was above 6 mmol.l−1 for all mice and all protocols (Figure 3). In the Ramp3 0° protocol, the mean [La]max was 13.80 ± 0.34 and [La]max was greater than 12 mol.l−1 for all mice. Figure 3 [La]max (measured 2 min after the end of the run) in each exercise protocol. IP, an incremental protocol with a starting speed of 10 m.min−1 and an increment of 3 m.min−1 every 3 min; Ramp3, a ramp protocol with a starting speed of 3 m.min−1 and an acceleration of 3 m.min−2 (0.05 m.min−1.s−1); Ramp12, a ramp protocol with a starting speed of 3 m.min−1 and an acceleration of 12 m.min−2 (0.2 m.min−1.s−1). Maximal speed: Vmax The Vmax of the mice was higher in the ramp protocols (54.88 ± 4.57 m.min−1 for Ramp12 0°; 46.34 ± 2.45 m.min−1 for Ramp3 0°) than in the step protocol (IP 0°: 38.13 ± 1.79 m.min−1) (Figure 4). For all three protocols, Vmax was higher with 0° than with 25°. Figure 4 Vmax in each exercise protocol. IP, an incremental protocol with a starting speed of 10 m.min−1 and an increment of 3 m.min−1 every 3 min; Ramp3, a ramp protocol with a starting speed of 3 m.min−1 and an acceleration of 3 m.min−2 (0.05 m.min−1.s−1); Ramp12, a ramp protocol with a starting speed of 3 m.min−1 and an acceleration of 12 m.min−2 (0.2 m.min−1.s−1); §, a significant difference between 25 and 0° for the same protocol (p < 0.05); *, differs significantly from all other protocols (p < 0.05). Time to exhaustion As shown in Figure 5, the time to exhaustion was significantly longer in IP 0° (29.33 ± 1.58 min) than in the two ramp protocols. For example, the time to exhaustion in Ramp3 0° (15.43 ± 0.8 min) was almost half that observed in IP 0°. Figure 5 Time to exhaustion. IP, an incremental protocol with a starting speed of 10 m.min−1 and an increment of 3 m.min−1 every 3 min; Ramp3, a ramp protocol with a starting speed of 3 m.min−1 and an acceleration of 3 m.min−2 (0.05 m.min−1.s−1); Ramp12, a ramp protocol with a starting speed of 3 m.min−1 and an acceleration of 12 m.min−2 (0.2 m.min−1.s−1); §, a significant difference between 25 and 0° for the same protocol (p < 0.05); *, differs significantly from all other protocols (p < 0.05). Discussion The present study in mice was designed to determine whether (i) a continuous ramp protocol yielded a higher VO2peak than a stepwise, incremental protocol, and (ii) the VO2peak measured in the ramp protocol corresponded to VO2max. This is an important issue, given that mice are frequently studied models in exercise physiology and that a variety of exercise protocols have been applied in this context. Our main findings were that a ramp protocol (with an acceleration of 3 m.min−2 and no treadmill slope) elicited a higher VO2peak than an incremental protocol (regardless of slope), and that the VO2peak does appear to correspond to the VO2max (given that a VO2 plateau was observed and the secondary criteria were met). The Ramp3 0° protocol is therefore relevant for the determination of VO2max inmice. According to the literature data, VO2peak in sedentary male mice ranges from 47 to 94 ml.kg−0.75.min−1 (Dohm et al., 1994; Schefer and Talan, 1996; Desai et al., 1999; Niebauer et al., 1999; Kemi et al., 2002). Furthermore, no major gender differences have been reported. Although gender differences have been observed for voluntary exercise (with young female mice running farther and faster than young males Lightfoot et al., 2004; Bartling et al., 2016), studies of forced exercise on a treadmill have not evidenced gender differences for critical speed, maximum distance (Billat et al., 2005; Lightfoot et al., 2007), or VO2peak in untrained mice (Kemi et al., 2002). Hence, we conclude that aerobic capacity does not depend on gender in untrained mice. Along with heterogeneity in the test protocols, several other factors may influence the observed VO2peak. It has been reported that VO2peak falls from 79 ml.kg−0.75.min−1 in young adult (12-month-old) mice to 56 ml.kg−0.75.min−1 in elderly (24-month-old) mice (Schefer and Talan, 1996). Thus, age differences in various studies may account for some of the discrepancies between reported VO2peak values. Moreover, the mouse's level of performance is known to depend on the strain (Lightfoot et al., 2001; Billat et al., 2005). Given that VO2peak is considered to be an indicator of performance, one can legitimately hypothesize that this variable is also influenced by the strain of mouse studied. The only study to date of VO2peak in FVB mice reported a value corresponding to 60 ml.kg−0.75.min−1 (Chow et al., 2007) which falls within the range of values observed in the present study. Hence, the choice of different strains may also account for some of the discrepancies in VO2peak values. Furthermore, the impact of the exercise protocol used to determine VO2peak values in mice has not previously been assessed. To the best of our knowledge, the only previous study in this field focused on the effect of treadmill slope on VO2peak in an incremental protocol (Wisløff et al., 2001). We hypothesized that the choice of exercise protocol would have a critical impact on the measured VO2peak. For example, Kemi et al.'s (2002) study used an incremental protocol with an increment of 1.8 m.min−1 every 2 min. They reported a mean VO2peak value of 47 ml.kg−0.75.min−1 and a mean time to exhaustion of 30 min. In contrast, Dohm et al. (1994) study used an incremental protocol with an increment of 8.4 m.min−1 every 2 min to obtain a mean VO2peak value of 94 ml.kg−0.75.min−1 and a time to exhaustion of 16 min. The results of the present study showed that the VO2peak value is protocol-dependent (p < 0.05). The highest value was obtained in the Ramp3 0° protocol; hence, ramp protocols are suitable for determining VO2peak in mice. Indeed, the ramp protocol was associated with a shorter time to exhaustion (15 ± 0.82 min in Ramp3 0° and 30 ± 1.51 min in IP 0°). This may explain why VO2peak was higher in the Ramp3 0° protocol than in the IP 0° protocol. In humans, a shorter time to exhaustion is associated with a higher VO2max (Froelicher et al., 1974); this also appears to be true in the mouse. It has been demonstrated that VO2peak is highest when the treadmill slope is between 15 and 35° (Kemi et al., 2002). Accordingly, we chose a value of 25°. This slope was associated with significant differences in the measured VO2peak (relative to the 0° condition, and for both the incremental protocol and the ramp protocols). Interestingly, the IP 25° protocol yielded a higher VO2peak value that the IP 0° protocol. This confirmed the results of Kemi et al.'s study of an incremental protocol (2002). In contrast, VO2peak was lower for Ramp3 25° than for Ramp3 0°. In exercising human (in whom energy expenditure is mainly related to muscle work), concentric work requires 3- to 5-fold more energy than the same amount of eccentric work. The energy cost of running therefore depends on the relative proportions of these two types of work, which in turn depends on the slope; the steeper the slope at a given speed, the greater the proportion of concentric work and thus the greater the energy expenditure. (Minetti et al., 1993, 1994; Pringle et al., 2002). This phenomenon seems to have occurred in the ramp protocols because the mice attained a lower Vmax when the treadmill was inclined. Furthermore, running on a sloping treadmill may recruit a greater muscle mass (Kemi et al., 2002). Consequently, involvement of a greater muscle mass and a greater proportion of concentric work in ramp protocols with slope might be responsible for fatigue and thus a lower VO2peak. However, the data collected in the present study did not enable us to confirm this hypothesis. Furthermore, it is possible that use of a shallower slope would have increased the concentric work without leading to too much fatigue and thus would have yielded a higher VO2peak value. As well as being associated with the highest VO2peak value, the Ramp3 0° protocol produced a VO2max plateau for which two secondary criteria (the blood lactate concentration and the RER) were fulfilled. Thus, a ramp protocol with an acceleration of 3 m.min−2 and no slope enables the determination of the VO2max in mice, according to the definition usually applied in humans. Over the last 15 years, a number of researchers have evaluated the influence of data sampling on changes over time in VO2 and the determination of VO2max in human (Astorino et al., 2005; Midgley et al., 2006, 2007; Astorino, 2009). These studies showed that averaging VO2 over successive 15 s periods provided a more accurate measurement of VO2max and increased the likelihood of observing a VO2 plateau. As breath-by-breath sampling is not possible for mice in a metabolic chamber, we used the device's shortest sampling time (5 s, i.e., below the maximum recommended value of 15 s). Furthermore, very few studies have focused on whether a VO2 plateau (which defines VO2max) can be observed in mice. Many researchers have not distinguished between VO2peak and VO2max, and have defined VO2max in different ways. For example, Gebczynski defined VO2max as the highest mean VO2 value over 1 min (Gebczynski and Konarzewski, 2009), and Ferreira et al. (2007) considered that VO2max was equivalent to VO2peak (Ferreira et al., 2007). In contrast, some researchers have stated that VO2max corresponds to the VO2 plateau; unfortunately, the researchers evaluated the VO2 curve visually and did not define criteria for detecting a plateau (Niebauer et al., 1999; Kemi et al., 2002). In 1955, Taylor et al. stated that the change in VO2 (ΔVO2) should be below 2.1 ml.kg−1.min−1 or 150 ml min−1 for more than 30 s if it is to be considered as a VO2max plateau: (Billat et al., 2013). For a sedentary subject, this ΔVO2 represents around 5% of the difference between the VO2 measured at rest and VO2max. In view of our previous data in mice, (Mille-Hamard et al., 2012; Mouisel et al., 2014) and studies indicating that there is not much difference between VO2 at rest and VO2peak in mice (Ferreira et al., 2007; Mazzucatto et al., 2014), we decided to reduce the value of ΔVO2. Hence, in the present study, the VO2 plateau was determined when the VO2 did not increase by more than 1% of the difference between the VO2 at rest and VO2peak over a 30 s period, despite an increase in running speed. Furthermore, Kemi et al. considered two of the secondary criteria applied in human exercise tests. Given that non-invasive measurement of the heart rate is not practical in mice, Kemi et al. suggested that an RER > 1 and an [La]max > 6 mmol.l−1 can be used to confirm the value of VO2max when a VO2 plateau is not observed (Kemi et al., 2002). Our present data on RERmax and [La]max suggest that the VO2max was attained by all the mice during the Ramp3 0° protocol. The recorded values of RERmax (mean: 1.06 ± 0.01) and [La]max (>12 mmol.l−1) indicated that exercise was strenuous. (Astorino, 2009). In humans, a standardized stepwise protocol is usually preferred because it enables the determination of other performance indices (blood lactate, ventilatory thresholds, heart rate, etc.) as well as VO2max. In mice, these indices cannot be calculated without using non-routine equipment (an implanted heart rate sensor and a mouthpiece, for example), and so the ramp protocol suggested here (which enables the true VO2max to be determined rapidly) should be preferred. However, it remains to be seen whether the ramp protocol is suitable for all strains and age groups and for both sexes. Conclusion The principal findings of this study in the mouse were that (i) the VO2peak observed at the end of exhaustive exercise is protocol-dependent, and (ii) a ramp exercise protocol with an acceleration of 3 m.min−2 (i.e., 0.05 m.min−1.s−1) and no treadmill slope is suitable for determining VO2max as defined in humans. Author contributions MA, RN contributed to the design of the work, the acquisition, analysis, and interpretation of data, drafted the work; LM, IM contributed to the design of the work, the acquisition, analysis, and interpretation of data, drafted the work and revisited it critically for important intellectual content; VB contributed to the design of the work, the interpretation of data, revisited the work critically for important intellectual content. All authors approved the version to be published and agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. Conflict of interest statement The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. ==== Refs References Astorino T. A. (2009 ). Alterations in VOmax and the VO plateau with manipulation of sampling interval . Clin. Physiol. Funct. Imaging 29 , 60 –67 . 10.1111/j.1475-097X.2008.00835.x 19125732 Astorino T. A. Willey J. Kinnahan J. Larsson S. M. Welch H. Dalleck L. C. (2005 ). Elucidating determinants of the plateau in oxygen consumption at VO2max . Br. J. Sports Med. 39 , 655 –660 . discussion: 660. 10.1136/bjsm.2004.016550 16118305 Astrand I. (1960 ). Aerobic work capacity in men and women with special reference to age . Acta Physiol. Scand. Suppl. 49 , 1 –92 . 13794892 Åstrand P. O. (1952 ). Experimental Studies of Physical Working Capacity in Relation to Sex and Age . Copenhagen : E. Munksgaard . Balke B. Ware R. W. (1959 ). An experimental study of physical fitness of Air Force personnel . U. S. Armed Forces Med. J. 10 , 675 –688 . 13659732 Bartling B. Al-Robaiy S. Lehnich H. Binder L. Hiebl B. Simm A. (2016 ). Sex-related differences in the wheel-running activity of mice decline with increasing age . Exp. Gerontol . [Epub ahead of print]. 10.1016/j.exger.2016.04.011 27108181 Billat V. L. Mouisel E. Roblot N. Melki J. (2005 ). Inter- and intrastrain variation in mouse critical running speed . J. Appl. Physiol. (1985) 98 , 1258 –1263 . 10.1152/japplphysiol.00991.2004 15542571 Billat V. Petot H. Karp J. R. Sarre G. Morton R. H. Mille-Hamard L. (2013 ). The sustainability of VO2max: effect of decreasing the workload . Eur. J. Appl. Physiol. 113 , 385 –394 . 10.1007/s00421-012-2424-7 22752344 Bruce R. A. Blackmon J. R. Jones J. W. Strait G. (1963 ). Exercising testing in adult normal subjects and cardiac patients . Pediatrics 32 (Suppl ), 742 –756 . 14070531 Chow L. S. Greenlund L. J. Asmann Y. W. Short K. R. McCrady S. K. Levine J. A. . (2007 ). Impact of endurance training on murine spontaneous activity, muscle mitochondrial DNA abundance, gene transcripts, and function . J. Appl. Physiol. (1985) 102 , 1078 –1089 . 10.1152/japplphysiol.00791.2006 17110513 Cohn J. N. (1987 ). Quantitative exercise testing for the cardiac patient: the value of monitoring gas exchange. Atlanta, Georgia, March 7–8, 1986. Proceedings . Circulation 76 , VI1-58 . 3677363 Desai K. H. Schauble E. Luo W. Kranias E. Bernstein D. (1999 ). Phospholamban deficiency does not compromise exercise capacity . Am. J. Physiol. 276 , H1172 –H1177 . 10199840 Dohm M. R. Richardson C. S. Garland T. Jr. (1994 ). Exercise physiology of wild and random-bred laboratory house mice and their reciprocal hybrids . Am. J. Physiol. 267 , R1098 –R1108 . 7943421 Fernando P. Bonen A. Hoffman-Goetz L. (1993 ). Predicting submaximal oxygen consumption during treadmill running in mice . Can. J. Physiol. Pharmacol. 71 , 854 –857 . 10.1139/y93-128 8143245 Ferreira J. C. Rolim N. P. Bartholomeu J. B. Gobatto C. A. Kokubun E. Brum P. C. (2007 ). Maximal lactate steady state in running mice: effect of exercise training . Clin. Exp. Pharmacol. Physiol. 34 , 760 –765 . 10.1111/j.1440-1681.2007.04635.x 17600553 Froelicher V. F. Jr.Brammell H. Davis G. Noguera I. Stewart A. Lancaster M. C. (1974 ). A comparison of the reproducibility and physiologic response to three maximal treadmill exercise protocols . Chest 65 , 512 –517 . 10.1378/chest.65.5.512 4826023 Gebczynski A. K. Konarzewski M. (2009 ). Metabolic correlates of selection on aerobic capacity in laboratory mice: a test of the model for the evolution of endothermy . J. Exp. Biol. 212 , 2872 –2878 . 10.1242/jeb.030874 19684223 Hawkins M. N. Raven P. B. Snell P. G. Stray-Gundersen J. Levine B. D. (2007 ). Maximal oxygen uptake as a parametric measure of cardiorespiratory capacity . Med. Sci. Sports Exerc. 39 , 103 –107 . 10.1249/01.mss.0000241641.75101.64 17218891 Hill A. V. Lupton H. (1923 ). Muscular exercise, lactic acid, and the supply and utilization of oxygen . QJM 16 , 135 –171 . 10.1093/qjmed/os-16.62.135 Høydal M. A. Wisløff U. Kemi O. J. Ellingsen O. (2007 ). Running speed and maximal oxygen uptake in rats and mice: practical implications for exercise training . Eur. J. Cardiovasc. Prev. Rehabil. 14 , 753 –760 . 10.1097/HJR.0b013e3281eacef1 18043295 Issekutz B. Birkhead N. C. Rodahl K. (1962 ). Use of respiratory quotients in assessment of aerobic work capacity . J. Appl. Physiol. 17 , 47 –50 . Kemi O. J. Loennechen J. P. Wisloff U. Ellingsen Ø. (2002 ). Intensity-controlled treadmill running in mice: cardiac and skeletal muscle hypertrophy . J. Appl. Physiol. (1985) 93 , 1301 –1309 . 10.1152/japplphysiol.00231.2002 12235029 Lerman I. Harrison B. C. Freeman K. Hewett T. E. Allen D. L. Robbins J. . (2002 ). Genetic variability in forced and voluntary endurance exercise performance in seven inbred mouse strains . J. Appl. Physiol. (1985) 92 , 2245 –2255 . 10.1152/japplphysiol.01045.2001 12015333 Lightfoot J. T. Turner M. J. Daves M. Vordermark A. Kleeberger S. R. (2004 ). Genetic influence on daily wheel running activity level . Physiol. Genomics 19 , 270 –276 . 10.1152/physiolgenomics.00125.2004 15383638 Lightfoot J. T. Turner M. J. Debate K. A. Kleeberger S. R. (2001 ). Interstrain variation in murine aerobic capacity . Med. Sci. Sports Exerc. 33 , 2053 –2057 . 10.1097/00005768-200112000-00012 11740298 Lightfoot J. T. Turner M. J. Knab A. K. Jedlicka A. E. Oshimura T. Marzec J. . (2007 ). Quantitative trait loci associated with maximal exercise endurance in mice . J. Appl. Physiol. (1985) 103 , 105 –110 . 10.1152/japplphysiol.01328.2006 17412788 Maritz J. S. Morrison J. F. Peter J. Strydom N. B. Wyndham C. H. (1961 ). A practical method of estimating an individual's maximal oxygen intake . Ergonomics 4 , 97 –122 . 10.1080/00140136108930512 Mazzucatto F. Higa T. S. Fonseca-Alaniz M. H. Evangelista F. S. (2014 ). Reversal of metabolic adaptations induced by physical training after two weeks of physical detraining . Int. J. Clin. Exp. Med. 7 , 2000 –2008 . 25232380 Midgley A. W. McNaughton L. R. Carroll S. (2006 ). Verification phase as a useful tool in the determination of the maximal oxygen uptake of distance runners . Appl. Physiol. Nutr. Metab. 31 , 541 –548 . 10.1139/h06-023 17111008 Midgley A. W. McNaughton L. R. Carroll S. (2007 ). Physiological determinants of time to exhaustion during intermittent treadmill running at vV(.-)O(2max) . Int. J. Sports Med. 28 , 273 –280 . 10.1055/s-2006-924336 17024633 Mille-Hamard L. Billat V. L. Henry E. Bonnamy B. Joly F. Benech P. . (2012 ). Skeletal muscle alterations and exercise performance decrease in erythropoietin-deficient mice: a comparative study . BMC Med. Genomics 5 :29 . 10.1186/1755-8794-5-29 22748015 Minetti A. E. Ardigò L. P. Saibene F. (1993 ). Mechanical determinants of gradient walking energetics in man . J. Physiol. 472 , 725 –735 . 10.1113/jphysiol.1993.sp019969 8145168 Minetti A. E. Ardigò L. P. Saibene F. (1994 ). Mechanical determinants of the minimum energy cost of gradient running in humans . J. Exp. Biol. 195 , 211 –225 . 7964412 Mouisel E. Relizani K. Mille-Hamard L. Denis R. Hourdé C. Agbulut O. . (2014 ). Myostatin is a key mediator between energy metabolism and endurance capacity of skeletal muscle . Am. J. Physiol. Regul. Integr. Comp. Physiol. 307 , R444 –R454 . 10.1152/ajpregu.00377.2013 24965795 Niebauer J. Maxwell A. J. Lin P. S. Tsao P. S. Kosek J. Bernstein D. . (1999 ). Impaired aerobic capacity in hypercholesterolemic mice: partial reversal by exercise training . Am. J. Physiol. 276 , H1346 –H1354 . 10199861 Pringle J. S. Carter H. Doust J. H. Jones A. M. (2002 ). Oxygen uptake kinetics during horizontal and uphill treadmill running in humans . Eur. J. Appl. Physiol. 88 , 163 –169 . 10.1007/s00421-002-0687-0 12436286 Rezende E. L. Chappell M. A. Gomes F. R. Malisch J. L. Garland T. Jr. (2005 ). Maximal metabolic rates during voluntary exercise, forced exercise, and cold exposure in house mice selectively bred for high wheel-running . J. Exp. Biol. 208 , 2447 –2458 . 10.1242/jeb.01631 15939783 Schefer V. Talan M. I. (1996 ). Oxygen consumption in adult and AGED C57BL/6J mice during acute treadmill exercise of different intensity . Exp. Gerontol. 31 , 387 –392 . 10.1016/0531-5565(95)02032-2 9415121 Taylor C. R. Maloiy G. M. Weibel E. R. Langman V. A. Kamau J. M. Seeherman H. J. . (1981 ). Design of the mammalian respiratory system. III Scaling maximum aerobic capacity to body mass: wild and domestic mammals . Respir Physiol. 44 , 25 –37 . 10.1016/0034-5687(81)90075-X 7232885 Whipp B. J. Davis J. A. Torres F. Wasserman K. (1981 ). A test to determine parameters of aerobic function during exercise . J. Appl. Physiol. 50 , 217 –221 . 6782055 Wisløff U. Helgerud J. Kemi O. J. Ellingsen O. (2001 ). Intensity-controlled treadmill running in rats: VO(2 max) and cardiac hypertrophy . Am. J. Physiol. Heart Circ. Physiol. 280 , H1301 –H1310 . Available online at: http://ajpheart.physiology.org/content/280/3/H1301.full.pdf+html 11179077 Yoon B. K. Kravitz L. Robergs R. (2007 ). VO2max, protocol duration, and the VO2 plateau . Med. Sci. Sports Exerc. 39 , 1186 –1192 . 10.1249/mss.0b13e318054e304 17596788

PMC005xxxxxx/PMC5002033.txt

==== Front Med Educ OnlineMed Educ OnlineMEOMedical Education Online1087-2981Co-Action Publishing 3240510.3402/meo.v21.32405Research ArticleBuilding and executing a research agenda toward conducting implementation science in medical education Carney Patricia A. 12*Crites Gerald E. 3Miller Karen H. 4Haight Michelle 5Stefanidis Dimitrios 6Cichoskikelly Eileen 7Price David W. 89Akinola Modupeola O. 10Scott Victoria C. 11Kalishman Summers 121 Department of Family Medicine, Oregon Health & Science University School of Medicine, Portland, OR, USA2 Department of Public Health & Preventive Medicine, Oregon Health & Science University School of Medicine, Portland, OR, USA3 Department of Medicine, Augusta University/University of Georgia Medical Partnership, Athens, GA, USA4 School of Medicine, University of Louisville, Louisville, KY, USA5 Department of Pediatrics, Oregon Health & Science University School of Medicine, Portland, OR, USA6 Department of Surgery, Carolinas Healthcare System, Charlotte, NC, USA7 Department of Family Medicine, College of Medicine, University of Vermont, Burlington, VT, USA8 American Board of Medical Specialties Research and Education Foundation, Chicago, IL, USA9 Department of Family Medicine, University of Colorado School of Medicine, Denver, CO, USA10 Department of Pediatrics, Wake Forest School of Medicine, Winston-Salem, NC, USA11 Department of Psychology, University of North Carolina, Charlotte, NC, USA12 Department of Family and Community Medicine, University of New Mexico, Albuquerque, NM, USA* Correspondence to: Patricia A. Carney, Department of Family Medicine, Oregon Health & Science University School of Medicine, 3181 SW Sam Jackson Park Rd., Portland, OR 97239, USA, Email: carneyp@ohsu.edu25 8 2016 2016 21 10.3402/meo.v21.3240526 5 2016 25 7 2016 25 7 2016 © 2016 Patricia A. Carney et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 International License, allowing third parties to copy and redistribute the material in any medium or format and to remix, transform, and build upon the material for any purpose, even commercially, provided the original work is properly cited and states its license.Background Implementation science (IS) is the study of methods that successfully integrate best evidence into practice. Although typically applied in healthcare settings to improve patient care and subsequent outcomes, IS also has immediate and practical applications to medical education toward improving physician training and educational outcomes. The objective of this article is to illustrate how to build a research agenda that focuses on applying IS principles in medical education. Approach We examined the literature to construct a rationale for using IS to improve medical education. We then used a generalizable scenario to step through a process for applying IS to improve team-based care. Perspectives IS provides a valuable approach to medical educators and researchers for making improvements in medical education and overcoming institution-based challenges. It encourages medical educators to systematically build upon the research outcomes of others to guide decision-making while evaluating the successes of best practices in individual environments and generate additional research questions and findings. Conclusions IS can act as both a driver and a model for educational research to ensure that best educational practices are easier and faster to implement widely. best educational practicescurriculum developmentteam-based careeducational researchundergraduate/graduate medical education ==== Body It takes an estimated 17 years to get healthcare practices proven to be effective into day-to-day clinical practice (1), which delays offering patients the best possible care. Implementation science (IS) was developed to optimize quality of care by narrowing the gap between research and practice (2). It involves the scientific study of methods to promote the uptake of research results and evidence-based practices to improve the quality and effectiveness of health services (3). Specifically, IS seeks to clarify what interventions work where, when, how, and for whom to implement innovations, programs, and processes effectively (3). Recent healthcare reform legislation has heightened awareness of the promise IS has toward reforming our educational systems and improving healthcare outcomes (4). By incorporating IS into the repertoire of learning resources, medical educators are better able to design programs and systems that enhance educational outcomes of our learners. Similar to healthcare delivery systems, educational institutions that train health professionals are highly complex (5). These two types of organizations are typically co-mingled, with many healthcare organizations serving as clinical training centers for medical students, residents, and other allied health professionals. Lessons learned from the application of IS in the clinical environment can inform the application of IS in other health professions education settings. Price et al. (6) recently published an article on the implications that IS has for medical education, including that it can help schools achieve changes in learner performance and competence as well as patient outcomes. These authors also suggest that IS should be incorporated into curricula across disciplines and the health educational continuum to further facilitate achievement of implementing best educational practices. For example, O'Flaherty and Phillips (7) tested their own use of the flipped classroom strategy (independent preparation followed by in-class complex group learning tasks (8)). They did this by conducting a literature review to identify best practices, and then developed research questions related to their own institution and used these to conduct an assessment of strategies, evaluations, and outcomes. These broad views are important and invaluable, but what is missing from the literature on IS in medical education is how best to generate an IS research agenda and identify processes for overcoming typical challenges. Using practical examples on how IS can be applied in diverse educational settings will assist educational leaders, educators, and educational researchers foster IS in their learning settings. In this article, we use a scenario-based educational example to describe how to apply a structured IS approach to overcome implementation challenges in medical education, and we identify additional educational research questions that IS approaches can help address. The scenario is applicable to undergraduate medical education, graduate medical education, and interprofessional medical education, which we purposefully selected, given the importance of team-based care in the health professions. Finally, we suggest next steps to advance the IS research agenda across the medical educational continuum. While our intention is to target educators and educational researchers in this work, it will become evident that collaboration among researchers, educators, and educational leaders as well as many other stakeholders is necessary for IS to be successful and beneficial for all. Research questions that IS can help address Best Evidence Medical Education (BEME) is an international group committed to the development of evidence-informed education in medicine and other health professions (9). Table 1 provides examples of questions we identified based on effective educational practices identified by BEME. It is a useful guide for research questions that, if addressed, could produce effective educational outcomes. Many important questions exist about how to more effectively incorporate best educational practices into learning settings. We organized a set of such examples into three thematic areas: 1) instructional design and teaching process; 2) facilitators and barriers to implementation of best or proven practices; and 3) costs, timing, and policy-related issues associated with successful adoption for medical schools and programs across the educational continuum. Table 1 Research questions that implementation science can help address Implementation science thematic area Example IS research questions based on existing best evidence from BEME Instructional design and teaching processes What facilitators can be employed to rapidly transform instructional design to include techniques that optimize working memory use or cognitive load of medical student learners? Which strategies are most effective for implementing interprofessional education across several institutions (Schools of Medicine, Nursing and Allied Health)? How can effective high fidelity medical simulators be used across health professions schools? How can more effective approaches to giving and getting feedback be incorporated into clinical precepting? Identifying what facilitates or hinders implementation of best practices What barriers and facilitators exist in incorporating best educational practices in undergraduate, residency, and continuing medical education activities? What factors hinder the development of effective self-monitoring and reflective practice? What approaches are best to undertake or avoid in developing e-learning portfolios? What factors affect successful adoption of electronic health records in physician training? Costs, timing, and policy-related issues that are associated with successful adoption for medical schools and programs across the educational continuum What is the cost of adopting interprofessional training? Has the implementation of the ACGME competencies in residency education led to improved resident education and patient outcomes? What are the organizational/institutional cost of IS in medical education? How can multi-institutional studies on the implementation of best educational practices be promoted and funded? How can organizational strategies to adopt IS be improved? What are the costs, policies, regulation, and adaptive institutional responses that enable best IS adoption and implementation? As with other research endeavors, IS for medical education research should be guided by an appropriate framework. While many theories, models, and frameworks exist for IS (10), the specific ones applied to IS research provide the context in which to examine why an implementation succeeds or fails (10). Because the focus of this article is on studying IS in medical education toward understanding and/or explaining influences on implementation outcomes, we have chosen Damschroder et al.'s consolidated framework for implementation research (CFIR). The CFIR (Fig. 1) (11) has become a popular framework for guiding the study of successful implementation within healthcare systems, as it assists in determining how components might be modified to help an innovation move from adoption failure (left side of model) to success (right side of model) (11). In the three-part example below, we apply the concepts of CFIR (Figs. 1 and 2) to different categories of research questions in medical education (Table 1) that IS can help to address. Fig. 1 Consolidated framework for implementation research (CFIR). Fig. 2 Adapted conceptual framework for implementation research in medical education. Example scenario Over utilization of the emergency department (ED) is a long-standing problem for university hospitals, and implementation of the Patient Protection and Affordable Care Act (PPACA) has increased the use of the ED by underserved populations (12). A current challenge in university hospitals as well as community-based settings is how to educate a diverse population on how and when to use community-based clinics versus when to use the ED. Thematic area 1: instructional design and teaching processes Example 1 Education strategy. A first-line strategy would be to train medical, nursing, and allied health students, residents, and nurse practitioner trainees simultaneously with faculty and other health professionals on how to better educate patients about community care compared to ED care and how making the best choice at the right time will benefit both patients and their families. Research question. Which strategies are most effective for implementing interprofessional education across several disciplines (e.g., medicine, nursing, and allied health) within our institution? CFIR asks us to define the entire population of individuals involved. It further emphasizes the need for understanding the culture.To approach this question, use current literature and structured discussions with faculty and learner stakeholders to identify proven practices to teach across multiple professions at the same time. For example, in 2007, BEME published a systematic review on best Search Results interprofessional education (IPE) practices (13), which revealed that faculty development has an important influence on the effectiveness of IPE for learners who have unique values. Customization and authenticity are important mechanisms for positive outcomes of IPE, which has been shown to enhance the development of practice and improvement of services (13). These principles should be implemented in this IPE training program, and IS evaluation strategies should be included to understand how this program was both successfully implemented as well as to identify the barriers that had to be overcome, so the study of further implementation practices can be improved. In the ED scenario, the BEME recommendations would need to be adapted to understand the different learning discipline requirements, focusing on each person's role for educating patients and managing care transitions. It is interesting to note that overuse of the ED was also selected by Lynch et al. (14) as the topic for a 2015 interdisciplinary workshop using IS to address a real-world problem. Example 2 Research question. How can providing feedback to all care team learners be effectively incorporated into clinical work?Because the CFIR model is cyclical, it requires mechanisms in place to assess and evaluate the intervention to continually improve processes. In this case, feedback is a primary mechanism for assessment. As the patient education skills in this scenario are taught and applied, using feedback from peers, attendings, other professional staff, and patients will help with revisions and improvements to the training process, the individual skills of faculty and learners, and will theoretically improve patients’ appropriate use of the ED. For example, Clynes and Raftery (15) found inconsistencies in the provision of feedback to students that were related to inadequate supervisor training, an unfavorable learning environment, and insufficient time spent with students. This study indicates that preceptors need effective training, including an appreciation of the steps of the feedback process, an understanding of the student response to feedback toward developing effective communication skills (15). Given that benefits of feedback include increased confidence, motivation, and self-esteem on the part of the learner, as well as improved clinical practice, implementing effective programs and studying them to ensure they are fully in place is vitally important in many educational settings. In our scenario, using the ‘failure’ data from this research report, the educational leaders involved in patient education and transitions in the ED should meet to plan feedback sources and processes, such as identifying which sources of data (e.g., patient surveys, ED readmission, and literacy improvements) are most useful for feedback data. They should also work with learners to make sense of the data and how to effectively interpret the meaning of various feedback processes. These steps will likely identify additional research questions to pose and address. Tavakol et al. (16) help by differentiating between medical education evaluation and medical education research: ‘Evaluation provides an overview of medical education issues; research is a biopsy of medical education practice’. But both evaluation and research can benefit from a well-founded, structured approach. Thematic area 2: identifying facilitators and barriers for implementation of best educational practices? Example 1 Educational strategy. Create a shared understanding for patients’ frame of reference by determining why patients select ED care rather than neighborhood clinic care. As part of the CFIR model that identified patients as a key stakeholder, processes were needed to include their voices. In doing so, an understanding can be gained regarding both the inner and outer settings of the implementation. Research question. What is the public's perception of the role the ED plays in their own healthcare, and what training models can help address misperceptions?In 2013, Shaw et al. (17) conducted a study on the decision-making process of patients who use the ED for primary care needs. Their findings are displayed as a decision-making flow chart that could be extremely useful in identifying decision points where instructional interventions could be most effective. In another study by Koziol-McLain et al. (18), patients reported that the stress in their lives had influenced their perceptions that they needed emergency care for non-urgent medical problems. In our scenario, the educational leaders in the ED should ask how the new care processes can be taught to learners and how learners can educate the ED system when flaws in the process are found. In addition, including an assessment of patients’ perspectives and motivations for seeking care and having health professionals educate patients about other non-ED options that are better suited to their healthcare problems would likely reduce overuse of the ED. Studying the effectiveness of implementing these new processes will inform this organization on what works. Example 2 Research question. What factors hinder the development of self-monitoring and reflective practice regarding interprofessional communication?Self-monitoring includes all stakeholders; the ED department example includes learners, educators, and patients. However, the development of informed self-monitoring (being aware of one's cognitive and emotional strengths and weaknesses) and reflective practices (ability to reflect on recent learning experiences and during learning experiences) can be challenging (19, 20). For example, a hallmark paper by Boud underscores the importance of context and locating learning in both educational and professional practice for reflective practices and reflecting to be enhanced (19). Yet in the ED example, the context creates challenges for learners from different disciplines to work collaboratively, given the time barriers in this environment (e.g., shifts, attendance to other learning activities). Thus, a focus on which methods of communication would enhance learners’ cross-discipline discourse would be effective given the time barriers. Thematic area 3: costs, timing, and policy-related issues that are associated with successful adoption for medical schools and programs across the educational continuum Example 1 Education strategy. The CFIR model includes attention to financial resources, which can serve as barriers or facilitators for adopting an intervention. An initial step would be to identify a cost analysis model that meets institutional needs. In this case, it would likely include both real costs, such as training materials, and hidden or incidental costs, such as faculty time and curricular time, which can be less precise. Research question. What is the cost of adopting this interprofessional training program?Cost analyses are best conducted when health economists are included to ensure the model chosen matches the activities undertaken and to ensure the outcomes of the enterprise are included in analyses so a cost-to-benefit ratio can be included. Careful planning would need to occur for this type of analysis to produce meaningful results. A caveat regarding educational research in the health professions, IS, or other types of educational research is that funding sources are limited (21), but the following strategies may help:Systematic reviews, literature searches, professional advice, information from professional organizations, and grant searches would provide insights into approaching this research question. Only small-to-moderate funding dollars are available through a variety of sources including medical specialty associations and private foundations. Medical educators must be advocates for support needed to move medical education research from having limited influence on practice to one able to generate sound insights regarding actual practices in education and improved health. In this case, the cost–benefit study may indicate potential long-term savings, as patients choose the most appropriate healthcare delivery site. Such savings could potentially off-set costs, which could drive institutional support. Any implementation project should keep track of incurred costs and savings that will both prove its effectiveness and support its sustainability. Without strong evidence, it is less likely that effective practices that can reduce costs could be successfully implemented. Many additional questions would likely emerge as part of this work. Example 3 Research question. How can the adoption of IS research in medical education be improved?When new programs undergo the truly rigorous research supported by IS, they create a history of success that encourages administrators to support additional IS projects. However, according to Rogers (22), some changes are destined to occur more slowly than others. Educators who use the IS approach (early adopters) (22) can mentor those who wish to adopt these refined methods. The feedback from learners’ discourse about the strengths and flaws of the processes may be generalizable to other situations, and the educational leaders could then ask what scholarly works should come from this, and who would benefit within their educational institution and other educational institutions and how these would be shared. The tiered approach in this example demonstrates application of IS using the CFIR model via cyclical strategies, new questions, processes, and stakeholders entering the project in a structured way. Actions are often determined by outcomes of the previous steps, so educators are not left to ‘reinvent’ processes that have already been shown to be better practices. Rather than the ‘single question’ approach often applied to medical education research, an IS approach steps back to allow a broad view of multiple questions applied sequentially to generate best practices. And, as we evaluate the application of these best practices in our own environment, we can share our outcomes and conceptual frameworks in order to grow this body of literature in a meaningful way. The CFIR provides advantages that are currently lacking with existing educational models. Progressive educational models must take into account that advanced learning activities (such as those that occur during clinical education) are situated in complex environments, comprising problem-solving with authentic cooperative activities with social learning as central components (23). Although a few progressive education models have recently been proposed for the complexities of online learning communities, none focus on the complexities of clinical environments (23, 24). Existing theories, such as cognitivist and humanist traditions, focus on individual learning and development but are less informative with social learning in complex environments (24). Although situated learning and communities of practice add elements of learning through doing, they are often focused on the goals of small learning teams, and not how larger contexts influence learning within and across teams, disciplines, and organizations (24). The CFIR model extends these concepts by providing a framework for organizing research questions and research targets, while helping to identify, capture, and analyze the many variables that influence learning in complex environments. Discussion There are several practical advantages for medical educators to adopt the IS approach. First, by basing decisions on findings that others have already generated through careful investigation, we are less likely to repeat mistakes and better able to focus resources on reforms and/or processes that have a higher probability of success. Second, when we support our proposals for change (and the accompanying research to evaluate the success of change) with research, we are more likely to gain administrative support. And finally, when we consider medical education as a scientific discipline and the learning settings in which we educate as our ‘labs’, we enrich our discipline using a unified, coherent approach to design, application, and evaluation of our processes, expanding a vitally needed body of research. Educational institutions must respond to the needs associated with educational effectiveness, just as healthcare organizations and clinicians must respond to the unique needs of patients and populations they serve, such as existing fiscal realities, regulatory bodies, and external competitors or external pressures (11). Educational institutions are also influenced by communication and informational systems (internal mediating factors), resource limitations, unique cultures of change, and leadership capacity (11). While educators’ choices (individual influences) can, individually and collectively, influence innovation adoption, issues such as perceived competence and self-efficacy become relevant for both learners and educators. Finally, features of the innovation itself (innovation structure) such as complexity, cost, and processes involved, including planning, execution, and evaluation (implementation processes), can affect successful adoption. Bonham and Solomon (25), in examining the relationship of IS to academic medicine, argue that IS is ‘a key component of comparative effectiveness research and essential for evidence-based healthcare reform. This perspective adds an additional application of the IS model’. BEME has now produced 30 guides summarizing evidence for best educational practices. However, more research is needed both to develop new evidence for best educational strategies and to identify best implementation strategies. For IS to advance in medical education, collaboration among all stakeholders is essential. The target audience for the proposed IS strategies described here is not limited to medical educators. This audience includes implementation scientists, academic and healthcare leaders, patients and communities, and all stakeholders, including payors and purchasers, affected by the implemented changes. In addition, IS can act as a driver and a model for medical education research, which would ensure research findings are easier to implement and more widely disseminated to avoid delays in the adoption of best evidence practices. We posit that three additional efforts are needed to promote successful IS in medical education. First, we need faculty development in the study and application of IS methods in educational settings, especially for junior faculty who are tasked with learning many new skills, including clinical teaching. Second, funding is needed at the institutional level and beyond to support this work. Current funding to study the implementation of best educational practices is limited, and using IS methods for curriculum development will likely go beyond traditional resource modeling strategies. Finally, processes used for IS need to be guided by appropriate theoretical frameworks to ensure that the social aspects of change are considered and evaluated. Success will depend on multidisciplinary collaborations and knowledge development among physicians, nurses and other health professions’ educators, health services researchers, and educational policy analysts, in many educational institutions across the spectrum of physician education. Acknowledgements The Association of American Medical Colleges’ Group on Educational Affairs administratively supported this work. The authors gratefully acknowledge the editorial input provided by Brian Mavis, PhD, Associate Professor, Medical Education Research & Development, Division of Evaluation and Dianne Wagner, MD, FACP, Associate Dean for College-wide Assessment, both of Michigan State University. Conflict of interest and funding The authors have not received any funding or benefits from industry or elsewhere to conduct this study. ==== Refs References 1 Blair M Getting evidence into practice – implementation science for paediatricians Arch Dis Child 2014 99 307 9 doi: http://dx.doi.org/10.1136/archdischild-2012-303464 24489364 2 Straus SE Brouwers M Johnson D Lavis JN Légaré F Majumdar SR Core competencies in the science and practice of knowledge translation: description of a Canadian strategic training initiative Implement Sci 2011 6 127 doi: http://dx.doi.org/10.1186/1748-5908-6-127 22152223 3 Eccles MP Armstrong D Baker R Cleary K Davies H Davies S An implementation research agenda Implement Sci 2009 4 18 doi: http://dx.doi.org/10.1186/1748-5908-4-18 19351400 4 Mann S Implementation science: a yardstick for reform Available from: Aamc.org/newsroom/reporter/dec10/165666/Implementation_science-a-yardstick_for_reform.html [cited 23 May 2012] 5 McGaghie WC Implementation science: addressing complexity in medical education Med Teach 2011 33 97 8 21275541 6 Price DW Wagner DP Krane NK Rougas SC Lowitt NR Offodile RS What are the implications of implementation science for medical education? Med Educ Online 2015 20 27003, doi: http://dx.doi.org/10.3402/meo.v20.27003 7 O'Flaherty J Phillips C The use of flipped classrooms in higher education: a scoping review Internet High Educ 2015 25 85 95 8 Knewton Infographics “Flipped Classroom” Available from: https://www.knewton.com/infographics/flipped-classroom/ [cited 14 July 2016] 9 The BEME collaborative: best evidence medical and health professional education Available from: http://www.bemecollaboration.org/Home/ [cited 30 August 2012] 10 Nilsen P Making sense of implementation theories, models and frameworks Implement Sci 2015 10 1 13 25567289 11 Damschroder LJ Aron DC Keith RE Kirsh SR Alexander JA Lowery JC Fostering implementation of health services research findings into practice: a consolidated framework for advancing implementation science Implement Sci 2009 4 1 15 19123945 12 Ginde AA Lowe RA Wiler JL Health insurance status change and emergency department use among US adults Arch Intern Med 2012 23 172 642 7 22450213 13 Hammick M Freeth D Koppel I Reeves S Barr H A best evidence systematic review of interprofessional education: BEME Guide no. 9 Med Teach 2007 29 735 51 18236271 14 Lynch CS Wajnberg A Jervis R Basso-Lipani M Bernstein S Colgan C Implementation science workshop: a novel multidisciplinary primary care program to improve care and outcomes for super-utilizers J Gen Intern Med 2016 31 1 6 15 Clynes MP Raftery SE Feedback: an essential element of student learning in clinical practice Nurse Educ Pract 2008 8 405 11 18372216 16 Tavakol M Gruppen LD Torabi S Using evaluation research to improve medical education Clin Teach 2010 7 192 6 21134182 17 Shaw EK Howard J Clark EC Etz RS Arya R Tallia AF Decision-making processes of patients who use the emergency department for primary care needs J Health Care Poor Underserved 2013 24 1288 305 23974399 18 Koziol-McLain J Price DW Weiss B Quinn AA Honigman B Seeking care for nonurgent medical conditions in the emergency department: through the eyes of the patient J Emerg Nurs 2000 26 554 63 11106453 19 Boud D Avoiding the traps: seeking good practice in the use of self-assessment and reflection in professional courses Int J Soc Work Ed 1999 18 121 32 20 Schumacher DJ Englander R Carraccio C Developing the master learner: applying learning theory to the learner, the teacher, and the learning environment Acad Med 2013 88 1635 45 24072107 21 Carline JD Funding medical education research: opportunities and issues Acad Med 2004 79 918 24 15383346 22 Rogers EM Diffusion of innovations 1995 4th ed New York Simon & Schuster Inc 23 Weichart G The learning environment as a chaotic and complex adaptive system: e-learning support for thrivability Syst Connect Matter Life Cult Technol 2013 1 36 53 24 Mann KV Theoretical perspectives in medical education: past experience and future possibilities Med Educ 2011 45 60 8 21155869 25 Bonham AC Solomon MZ Moving comparative effectiveness research into practice: implementation science and the role of academic medicine Health Aff 2010 29 1901 5

PMC005xxxxxx/PMC5002034.txt

==== Front Glob Health ActionGlob Health ActionGHAGlobal Health Action1654-97161654-9880Co-Action Publishing 3148210.3402/gha.v9.31482Original Article‘You try to play a role in her pregnancy’ - a qualitative study on recent fathers’ perspectives about childbearing and encounter with the maternal health system in Kigali, Rwanda Påfs Jessica 1*Rulisa Stephen 23Musafili Aimable 14Essén Birgitta 1Binder-Finnema Pauline 11 Department of Women's and Children's Health/IMCH, Uppsala University, Uppsala, Sweden2 Department of Obstetrics & Gynecology, School of Medicine, College of Medicine and Health Sciences, University of Rwanda, Kigali, Rwanda3 Department of Clinical Research, University Teaching Hospital of Kigali, Kigali, Rwanda4 Department of Pediatrics and Child Health, School of Medicine, College of Medicine and Health Sciences, University of Rwanda, Butare, Rwanda* Correspondence to: Jessica Påfs, Department of Women's and Children's Health/International Maternal and Child Health (IMCH), Uppsala University, Akademiska Sjukhuset, SE-751 85 Uppsala, Sweden, Email: jessica.pafs@kbh.uu.seResponsible Editor: Virgilio Mariano Salazar Torres, Karolinska Institutet, Sweden. 25 8 2016 2016 9 10.3402/gha.v9.3148202 3 2016 11 5 2016 16 6 2016 © 2016 Jessica Påfs et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 International License, allowing third parties to copy and redistribute the material in any medium or format and to remix, transform, and build upon the material for any purpose, even commercially, provided the original work is properly cited and states its license.Background Rwanda has raised gender equality on the political agenda and is, among other things, striving for involving men in reproductive health matters. With these structural changes taking place, traditional gender norms in this setting are challenged. Deeper understanding is needed of men's perceptions about their gendered roles in the maternal health system. Objective To explore recent fathers’ perspectives about their roles during childbearing and maternal care-seeking within the context of Rwanda's political agenda for gender equality. Design Semi-structured interviews were conducted with 32 men in Kigali, Rwanda, between March 2013 and April 2014. A framework of naturalistic inquiry guided the overall study design and analysis. In order to conceptualize male involvement and understand any gendered social mechanisms, the analysis is inspired by the central principles from relational gender theory. Results The participants in this study appeared to disrupt traditional masculinities and presented ideals of an engaged and caring partner during pregnancy and maternal care-seeking. They wished to carry responsibilities beyond the traditional aspects of being the financial provider. They also demonstrated willingness to negotiate their involvement according to their partners’ wishes, external expectations, and perceived cultural norms. While the men perceived themselves as obliged to accompany their partner at first antenatal care (ANC) visit, they experienced several points of resistance from the maternal health system for becoming further engaged. Conclusions These men perceived both maternal health system policy and care providers as resistant toward their increased engagement in childbearing. Importantly, perceiving themselves as estranged may consequently limit their engagement with the expectant partner. Our findings therefore recommend maternity care to be more responsive to male partners. Given the number of men already taking part in ANC, this is an opportunity to embrace men's presence and promote behavior in favor of women's health during pregnancy and childbirth – and may also function as a cornerstone in promoting gender-equitable attitudes. male involvementintimate relationshipmasculinityrelational theorygendersub-Saharan Africa ==== Body Introduction Men's responsibilities in childbearing first appeared as a global policy agenda in 1994 at the International Conference on Populations and Development (ICPD) (1). Since then, numerous studies have demonstrated that men's involvement promotes positive health outcomes – not only for women, but also for men and their children (2–4). A father's involvement during pregnancy appears to contribute to women's uptake of maternal health services, in addition to increasing his understanding of potential risks and the necessity to seek timely care (5–7). Including men in pregnancy and caretaking activities challenges gendered roles and social structures and promotes gender-equitable attitudes, both of which pose interesting future directions (4, 8). However, there are settings where men report facing barriers to increased involvement. This is partly because of childbearing and its related phenomena being considered the female domain. Studies throughout sub-Saharan Africa, for example, identify how men are often bound to normative social and cultural contexts, fueled by gendered expectations, which are attitudes similarly upheld by health system regulations or health care staff attitudes (9–12). Gender equality was recently added to the political agenda in Rwanda and is defined in the country's National Gender Policy document. This document states that ‘The issue of gender inequality is embedded in patriarchy as a system’ and identifies, among other things, a need to ensure that reproductive health services are accessible to both women and men (13). Another significant undertaking revisits inequitable laws and policies, where it was previously stated in the law that a husband had the dominating power in an intimate relationship; ‘The man is the head of the family and his opinion must prevail’ (14). However, this law is currently updated. Even though such structural changes are currently taking place, and more women are entering into paid work life, the promotion for gender equality has encountered some resistance within households and the community (15). A man is still viewed as the breadwinner, having dominance in decision-making, without expectation for becoming involved in caring activities (15, 16). Yet those Rwandan men having both higher education and income are identified as moving beyond traditionally gendered roles and espousing more gender-equitable attitudes (8, 17). The health system in Rwanda has improved over the past years and decreased its maternal mortality ratio from 1,071 per 100,000 live births in the year 2000 to 210 in 2015 (18, 19). At present, up to 91% of women are giving birth in a health facility, which partly is explained by the rule that imposes a fine if women deliver at home, and the availability of health insurance (20). Currently, 73% of the population is covered by the community-based health insurance, ‘Mutuelle de Santé’ (herein, Mutuelles), covering 90% of the care costs and some medicines (21, 22). Rwanda also reports a distinctly high number of male attendants at first antenatal care (ANC) for HIV testing with up to 87% of women attending with their partner, which is distinctive for this sub-Saharan setting (23, 24). The policy of partner involvement had the intention to prevent mother-to-child transmission, but appears to have turned into a strict requirement for male attendance, a factor that might hinder women from attending alone (25). However, this may also, together with the new political agenda, prompt men to renegotiate definitions of masculinity and challenge traditional gender norms. This study therefore explores recent fathers’ perspectives about their roles during pregnancy, maternal care-seeking, and childbirth within the current context of Rwanda's political agenda for gender equality. Methods Conceptual framework Because of the context of Rwanda with increased focus toward gender equality and involvement of men in ANC, we use Connell's relational theory of gender to conceptualize the participant's perspectives about their experiences (26, 27). This theory posits gender as socially constructed, and that men are gendered beings, having power as a central aspect of their relations, actions, and expectations. Gendered expectations are formed both under the influence and expectations of society, as well as by individual interactions, both between and among women and men. The additional concept of hegemonic masculinity embodies what is ‘currently accepted’ of a man in a certain context. Even though an individual's behavior is complex, and each individual will cope differently with the social setting, Connell argues that gendered structures produce a narrow arena for individuals. The construct of gender is thus multidimensional and arranged simultaneously on intrapersonal, interpersonal, institutional, and society-wide levels (26, 27). The pattern in gender structures on these different levels can be called gender regimes. Institutions, such as a health system, follow the logics of the gender regime within its specific context, and policies are simultaneously constructing and deconstructing the gendered order of a society. Ethical considerations Ethics approval for the project was obtained from the Rwanda National Health Research Committee, Kigali (NHRC/2012/PROT/0045), as well as the Institutional Review Board of Kigali University Teaching Hospital. Permission to recruit participants was sought and obtained separately from each hospital. All participants were approached and included in this study following our precise adherence to the ethics approval, including informed consent for participation, and strict handling and storage of the recorded and transcribed interviews. Research design and sampling This study took place in Kigali, the capital of Rwanda, having 1.2 million inhabitants and vast socioeconomic differences (28). The average inhabitant generates an income from farming or service-related industries, such as driving a motorbike or taxi. The literacy rate is 80% for men and 77% for women (19). Kigali provides public and private health centers, three district hospitals, and three tertiary referral hospitals. Data were collected between March 2013 and April 2014 at three public hospitals in Kigali and this work is part of a project focusing on maternal near-miss (MNM) in the context of Rwanda. MNM refers to women who survive a severe, life-threatening obstetric complication during childbearing (29), and this approach is appropriate for identifying underlying factors to mortality, both including care-seeking delays and inadequate provision of care (30–32). Some participants were recruited purposively with the help of health care workers and others via snowball sampling within the community (33) with assistance from a Rwandan interpreter, fluent in both Kinyarwanda and English. The research team was composed of three female European researchers (a medical doctor, a medical anthropologist, and the doctoral student whose foundation is in social work) and two male medical doctors from Kigali (both holding a PhD in international health). The first author (the doctoral student) and a local interpreter conducted individual interviews with the partners of women who had experienced MNM while they were still at the hospital, yet only after the woman gave her consent for their inclusion. These men were also asked to bring one or two friends to join group interviews, which were conducted at a later session in the community, either at a participant's home or at a conference center. The inclusion criteria for all participants were having a partner who recently experienced MNM or having recently become a father. All except two men consented to take part after learning the aims of the study. No reasons were given for the two men who declined (Table 1). Table 1 Overview of data collection Participants Data collected N (type of data collection) Length Place Partners to MNM 1–3 day(s) post-MNM 9 (IDI) 20–60 min Hospital, private room Partners to MNM 3–6 months post-MNM 4 (IDI, with MNM partner) 60–120 min Participants’ home Partners to MNM and snowballed participants 3–6 months post-MNM 3 (FGD; total 19 participants) 120–180 min Conference center Partners to MNM and snowballed participants 6–12 months post-MNM 3 (GD; total 8 participants) 90–120 min Participants’ home MNM=maternal near-miss; IDI=individual, semi-structured interview; GD=group discussion; FGD=focus group discussion. The framework of naturalistic inquiry guided the study (34). A hermeneutic dialectical approach was used during the interviews for eliciting a conversation-like experience about how men perceived themselves in this shared context (34). All interviews were transcribed into English for analysis and cross-checked by an external translator, which was done to ensure the validity of the transcription (35). Themes of the semi-structured interview questions included men's perceived role during pregnancy, how they experienced their partner's pregnancy outcome – which was sometimes life-threatening, and their attitudes toward fitting in during the care encounter with medical professionals. Analysis Naturalistic inquiry guided the emergent analysis (34), which began during initial data collection, where information from one interview helped to guide new questions in later interviews. Topical saturation of new concepts was met upon repetition of participants’ answers. We performed member checks to validate the findings (34), and the necessity for establishing topical saturation was met upon repetition of participants answers. In-depth analysis occurred once all interviews were transcribed via rereading of all interviews and use of AtlasTi (Scientific Software, 2013) for sorting the data. The first and last authors read the sorted data and coded and interpreted overarching categories that best encapsulated the men's perspectives about their role. These were then sorted into the conceptual framework inspired from Connell (26) and discussed by the full research team. Findings Figure 1 details the outcome of data collection among 32 men, 13 of whom had a partner that experienced MNM. Four of these men's partners had had a miscarriage and three babies had not survived birth. Among the remaining participants, two had prior experiences of a baby not surviving birth. All of the men expressed their desire for the pregnancy, even if it was not necessarily intended. The men were between 23 and 35 years old, and two were in their 50s. One participant was currently studying at the university, whereas all the others had either primary or secondary level of education. The men had varied occupations, and several of them were either unemployed or were daily wageworkers. The men in this study were among those having the community health insurance Mutuelles, which is available for low-income households not covered by other insurance systems, such as those for employees of government, military, or private companies. Fig. 1 Conceptual framework, inspired from Connell (26), mapping recent fathers’ perceived roles during childbearing and maternal care-seeking, illustrated on the different levels. The men openly shared their reflections on their perceived role during their partners’ childbearing. However, the topic of the MNM caused a momentary quietness before the men expressed thoughts of anxiety and frustration over the near loss of their partner and the loss of their coming baby. The event seemed to have brought these men to reflect on their own participation and to be more attentive to flaws in the maternal health system. Men's perspectives were interpreted on an intrapersonal, interpersonal, institutional, and societal level in Fig. 1 and represented separately for clarity. However, these levels should be understood as interlinked. Intrapersonal explanations A pregnancy was perceived as a blessing and embraced with joy, yet also as a responsibility embraced with worries. Men's main purpose was perceived as being the breadwinner, with the role to provide financially for their partner and coming family member. This role was not experienced as short term: ‘As a man you need to provide for the child until s/he is grown up enough to look after her/himself’ (28 years old, first child, partner to MNM). Nevertheless, it appeared important to be strong and engaged and to not allow anxieties about the new responsibility to affect the partner during pregnancy. When asked to define their role toward a pregnant partner, most men presented ideals of an engaged father, using attributes of caring and benevolence. Although some focused mostly on the financial aspects, several of the men further explained the importance of providing care so the woman could rest. Pressuring the partner for sex during pregnancy was seen as selfish, and referred to as something happening earlier, during traditional times, before men's understanding of the need to show respect toward a pregnant partner. The men were thoughtful about such attentiveness, but also toward consequences to the coming child:[During pregnancy] you should listen to whatever your wife tells you, and stop thinking that accepting her ideas or the advice she gives you is some sort of unmanly attitude. You also have to avoid being rude towards her or frightening toward her, because if she gets angry it will affect the baby, too. And then you risk losing both of them because you were a jerk. (FGD 3) Several men regarded their roles as dynamic, emphasized on shared decision-making, concluding that, ‘A man should stop thinking that he is the king of the home’ (FGD 2). Participants felt constrained by their limited insights into maternal health matters, yet perceived themselves as responsible to ensure a woman received maternal care. This was particularly apparent for ANC because the men perceived their presence as mandatory during the first ANC visit. Attendance was understood as required for HIV testing, and ensuring the expectant partner would be received for consultation. The men described attendance at the first ANC as a ‘government rule’ requiring ‘obligatory involvement’. However, complaints were raised about how long wait times actually made men have to take 1 day free from work, which created a liability to their role as breadwinner. This was especially true among those who worked far from home, reflecting that their limited presence had delayed their partner's initial maternal care-seeking. The requirement for a man's attendance at the first visit was perceived as non-flexible at public clinics, which motivated some to ensure their partner went instead to a private clinic. Interpersonal perspectives The men perceived new responsibilities during pregnancy. Their reflections about changed family dynamics were discussed through the lens of changing traditions. Many lived far away from family members after moving to Kigali to find employment, or they had few role models after having lost extended relatives in the 1994 Genocide. The men were particularly aware of the loss of their partner's mother or their own, that is, women who would have otherwise provided support for the pregnant partner. Men described their attempts to become involved, often consulting with relatives and friends to be able to assist: ‘You try to play a role in her pregnancy and you do all you can to make sure she is taken care of’ (33 years old, second child, partner to MNM). However, consulting with others was not always self-evident because pregnancy was esteemed a strictly private matter, shared between a man and woman. One man reflected, ‘Everybody manages their problems. You cannot tell the secrets of the family to other people’ (35 years old, second child, partner to MNM). At most, the couple could share with closest family or people considered trustworthy. Obtaining information on maternal health was generally perceived as challenging and identified as posing potential difficulties on the intimate relationship. Only a few had been present when their partner received maternal health information from a care provider. Not being present at such times seemed to consequently prevent some from absorbing information, for example: ‘When she comes home and tells you what they have told her, you just listen [to her] but do not care, because you were not there’ (GD: 34 years old, third child). Although pregnancy was considered a private matter, a pregnant woman was seen as respected in the community, and men perceived themselves to be under external pressure to fulfill expectations about becoming a father. There was the need to be caring and attentive to a partner's needs, and making sure she attends ANC. A man not attending to these responsibilities risks being publicly referred to as either negligent or cowardly. Most men considered being involved and affectionate in parallel, saying, ‘If you cannot cope with your wife's troubles, then you do not love her’ (FGD 3). Institutional obstacles Nearly all men perceived limited exposure to pregnancy information, as this was not provided at all facilities. They also had different experiences with ANC. Only a few had been included in a group information session conducted at the care facility, covering pregnancy and health risks. None of the men were welcomed during the actual pregnancy consultation, wondering aloud why they could not participate and receive direct, first-hand information from the care provider. For example, ‘[Health care workers] are more interested in talking to the women, but they do not consider informing the men, as well. It would be better if we all could understand more about those symptoms’ (GD: 27 years old, second child). Participants were strongly in favor of facility-based childbirth. They viewed it as their responsibility to bring a woman to a health facility where she would be taken care of by professionals, but also wished to avoid the imposed fine if she delivered at home. The latter was perceived as a bother that required strategizing:You have to keep the umbilical cord uncut and just rush to the hospital, because if you cut it, you will have some issues with the doctors. They are going to fine you. You just need to explain to them that the contractions were sudden and took her by surprise. (GD: 28 years old, first child) Several had accompanied their partner to the health facility at time of delivery, whereas others gave the responsibility to a female relative. Among the men who had accompanied their partners, some emphasized the importance of being there because ‘as a man, you will be listened to’ (FGD2). Many expressed bottlenecks at admission, doubt in the quality of maternal care provided and questioned care provider attitudes. For example, ‘[My wife] told me that when a woman is still in the waiting room and starts shouting, the [health care workers] will not even glance at her. They say that, “if you can still shout, it means you definitely still have strength”’ (33 years old, third child, one died during birth, partner to MNM). Another said his wife had been left alone in the ward and she had given birth while unattended. He explained, ‘I think that the medical staff should be trained in giving better care to their patients’ (GD: 31 years old, third child). One man contemplated that he, as the partner and man, might have been listened to by the care provider, in the case where his partner felt uncomfortable about expressing her needs: ‘There are some women who do not like to scream when they are in pain, which the doctor will interpret as if she is not suffering. But if you, who is closest to her, are there, then she can tell you to advise the doctor’ (26 years old, second child, first died, partner to MNM). The idea of not being heard during the care encounter triggered a lack of trust and a wish to be present during labor. Yet, such presence at a public health care facility was perceived as non-negotiable and only allowed at private facilities. Some had not even questioned being barred from attendance during the birth, perceiving it as ‘not allowing men in is a law at the hospital, and so we have just gotten used to it’ (GD: 31 years old, third child). Other men reasoned their exclusion for logistical causes and the limited privacy found in the labor ward: ‘[Health care workers] do not want to violate the privacy of other women in the shared room’ (FGD 2). Even though the men were unable to surveil the actual procedure, they were held to a sense of responsibility. In one case, where a man was asked to sign a consent form for cesarean section, he raised this paradox:They ask you to give your ID number and sign that you are there to make sure she is safe. But, then, after you sign, they turn around and lock you out of the delivery room. So, how am I supposed to ensure that my wife and kid are safe if am locked out of that room? (29 years old, first child, partner to MNM) Societal/cultural guidelines Despite having limited accessibility on the institutional level, the participants knew of supportive policy guidelines, particularly regarding men's presence for the first ANC visit. Yet, a man's actual right for being present at the consultation and childbirth remained unclear and seemed undefined. This presented strong reasons for why men were hesitant to demand increased involvement as it also clashed with cultural norms, as one man explained:In our culture, they say that if the husband sees his wife giving birth, he is never going feel attracted to her again sexually. We are saying that we should be there, but in the Rwandan culture, it is taboo to see your wife giving birth. Some women would not even accept to have their husbands there. So we are kind of on the fence. (26 years old, second child, first died, partner to MNM) Several participants discussed the cultural aspects of women not wanting the partner to be present during consultation. More than simple tradition of preferring female attendants, in particular, men perceived women as not wanting their partners to witness them in the condition of giving birth. Most men also reflected and were in support of the idea that their presence should always be based on the consent of the expectant mother. A father of three children expressed a clear wish to participate, yet highlighted: ‘I think the decision should come from her. If she is comfortable with it, then [the health care workers] should accept her choice to allow whoever she came with to be there and to watch the entire procedure’ (GD: 34 years old). The men's clear motivation was a wish to surveil the birth procedure, especially among the men whose baby had not survived. One explained, ‘I should have been allowed in there so that I could be sure that whatever happened was nobody's fault’ (33 years old, third child, one died during birth, partner to MNM). A number of men from the public facilities reflected about not having the same access to the childbirth as was possible at private care facilities. They highlighted a wish to be financially able to seek private care. Moreover, men expressed a wish to challenge restrictions at public care facilities, but felt limited in their ability to negotiate and instead blamed the unclear execution of guidelines at these institutions. Discussion These recent fathers in Kigali expressed support for increased engagement during pregnancy and childbirth, as also reported from other African settings (9, 11, 12). The participants voice sheds light on the perspective among men from the lower socioeconomic group in Rwanda and can be interpreted to contradict earlier findings arguing that men of lower education presented more conservative ideas (17). The participants described new family dynamics, which appeared to place novel demands on fatherhood. This also seemed to pave the way for changing role dynamics confronting their perceived rationale for a more traditional gender regime at the institutional and societal levels. However, these men appeared to balance their participation according to their perceived outer demands and cultural norms. The institutions appeared to vary in how willing they were to allow the men to take part, providing an inconsistent, and narrow, arena for increased engagement. This seemed to pose barriers for men as they attempted to access information about pregnancy health and risks. The men's expressed rhetoric in favor of being caring and benevolent, and their reflections about moving beyond unmanly attitudes or being ‘the king of the home’, suggests an increasingly acceptable attitude that challenges traditional masculine ideals (36). Our findings suggest that changed family dynamics, particularly those due to the genocide, which caused disruption in the generational shifts, as well as migration from the rural to the urban setting, have partly changed the requirement for men to take a more active role during pregnancy. Another important aspect appears to be the current societal demand for participation at first ANC visit for HIV testing. As supported by other studies (37), this seems to have challenged earlier ideals of masculinity as well as participation as something socially expected from an expectant man. However, at present, a man's decision-making power is upheld because a woman in this setting will not be admitted if she seeks ANC alone (25). Arguably, the men's increased engagement may be grounded in a latent interest to retain power over women (4, 38). However, it is notable that the participants in our study maintained a concern for the welfare of their partners and seemed to challenge the social expectations placed on them. And where some of the men seemingly dismissed their partners’ needs after not receiving first-hand information, this may be interpreted as distrust with the intimate partner or a need to possess power. This aspect points out the challenges of gendered influences on intimate communication between the couple, which is essential during childbearing (16). The men in this study appeared to have faced institutional barriers, despite their expressed wishes for increased involvement. There is, to our awareness, no written policy prohibiting men from being present during a woman's consultation and childbirth. It is thus presently up to the health facility and its staff to decide, which identifies a potential system-level inconsistency between institutions. The participants’ perspectives about health provider's unwillingness to include and engage them may reflect health providers’ own ‘traditional’ ideals of gendered masculinity. Alternatively, men might be unwelcomed in order for health care staff to protect their professional domain or to uphold maternity as a female domain. Most health care staff in Rwanda's maternity wards are women. If they are unwilling to renegotiate gender structures, this could also be a wish to uphold the power they currently possess over the patient. Moreover, having men present on the ward could be perceived as threatening to that established power and is a subject for further research. However, from the men's perspectives, their feelings of not being welcomed could have prevented them from challenging the established system. This is a subject for further research. Our participants raised their concerns about lack of respect given to their partners during the care encounter, echoing current advice that emphasizes increased focus on offering respectful, woman-centered maternal care (39). The men's wishes for increased presence and surveillance could be triggered by their distrust in care quality, particularly among those having a partner facing severe morbidity or mortality. The institutional demand placed on the man to sign a document admitting he is responsible for the outcome places him in a paradoxical position. Nevertheless, an important aspect of the maternity care wards in Rwanda's public health facilities is that they have limited physical space to allow men in without risking the privacy of other women. This well-known reason for restricting men cannot be neglected (40). However, asking him to sign a document of responsibility might be received as being placed into a powerless position. These aspects highlight the men's desire to afford private care because these venues are perceived as providing better quality care, as well as allowing them to be present during childbirth (41). This socioeconomic disadvantage could consequently underpin the masculine ideal of being the financial provider. Despite the push for gender equality, the lack of explicitly stated public policy about men's involvement at health care facilities in Rwanda contributes to men's disconnection and limits their access to pregnancy information. The men perceived these barriers as being regulated from the maternal health system, which is a situation likely to uphold current gendered domains. Notably, women may not be entirely supportive of challenging the gendered domains, either. Doyle et al. (16) identified Rwandan women as opposing increased male involvement during child care, because they did not want to be reflected upon poorly within their community. It is most likely that women are responding to the gendered expectations placed on them by society and tradition, even though this means they may unwittingly support gender inequitable norms. It is therefore essential to be attentive to the gendered ‘ideal of femininity’ in a setting. The maternal domain, specifically, may be one of the few arenas where a woman feels in power (37). Hence, male involvement should not compromise women's autonomy and decision-making. The men in this study have nevertheless shown themselves to be concerned about pregnancy, as well as the overall health of their partner and child during pregnancy, while also maintaining that women should have autonomous power to choose who is present during consultations and childbirth. However, with the narrow arena currently provided by the health system, it might bring such consequences as decreased motivation for involvement or could trigger distrust in both the available care and partner. Methodological considerations Topical saturation was quickly reached, suggesting strong similarities of perspectives among these men within this peri-urban setting, regardless of whether their partner had a near-miss or not. However, those men partnered with women experiencing complications presented more negative ideas of the health system, which should be considered. These men's experiences about the near-miss event itself are presented only briefly because this is analyzed more in-depth in a separate paper. Face-to-face interviews may provide normative responses of social desirability; however, this in turn tends to mirror the hegemonic norm within a study context. These findings should be considered for this aspect. The design allowed several participants to be revisited, which lent opportunity to respond to earlier reflections and gain trustworthiness of the interpretations. Our study encountered a few limitations because of its reliance on language interpretation. However, the Rwandan co-authors assisted in securing experienced interpreters, and multiple Kinyarwanda-English transcripts were randomly cross-checked for validity using different translators, as recommended for interpreter use (35). The first author and data collection translator are women, which may have impacted the way male participants responded. However, this aspect was not experienced as a barrier for in-depth discussion. Additionally, the foreign nationality of the first author seemed to make men eager to explain their perspectives and experiences to her as an ‘outsider’. Yet, it might be that the first author's cultural background influenced her choice of probing follow-up questions. Thus, to help ensure trustworthiness, the interpretations were member-checked throughout the study, both among the participants and within the community. Importantly, the author also spent a total of 1 year in the setting and took the opportunity to member-check the researchers’ interpretations throughout her time in Rwanda. Another potential limitation is that men's precise caretaking activities are not included here because of being beyond the study's scope. This aspect could have enriched attitudes related to gender equality in this setting. Moreover, a deeper exploration on women's and health care providers’ perspectives are to be explored further in a separate study. Conclusion Our findings imply that these men see an engaged and caring partner as the ideal, yet are facing challenges due to normative gendered expectations in this Rwandan society. From the viewpoint of our participants, they face resistance from the current maternal health policy and care providers in their desire to become increasingly involved. There are hints that the men's limited access and insight may motivate their mistrust of the public health system and its health care staff, but also of their intimate partner. This calls for attention in building men's trust in public maternity care as well as to the quality of care provided to help ensure safety for the pregnant woman. Additionally, because the maternal health system is a core social institution, it plays an important role in challenging gendered structures. The engagement of fathers during childbearing and maternal care-seeking may function as a cornerstone in promoting gender-equitable attitudes. With a number of men already attending ANC in this setting, the opportunity is already in place to inform and motivate men about their role as partner and to promote behaviors that favor women's health during pregnancy and childbirth. Men's interest for increased involvement is evident from our findings, which suggests their willingness to be responsive in a gender sensitive manner to women's needs in maternity care. However, policies aimed at increasing male involvement need to take into account that the inclusion of men could come at the cost of women's empowerment, particularly if a woman is not allowed to seek maternity care alone or with the person of her preference. Acknowledgements The authors thank all the participants and acknowledge the research assistants for their valuable work during data collection. This work was supported by the Swedish International Development Cooperation Agency/SAREC (SWE 2010–060) and the faculty of Medicine at Uppsala University, Sweden. Authors' contributions All authors participated in the design of the study. JP collected the data with assistance from AM and PBF at site. JP and PBF conducted the data analysis in discussion with all authors. JP and PBF drafted the manuscript. All authors gave critical input on the manuscript, and have read and approved the final version. Conflict of interest and funding The authors have not received any funding or benefits from industry or elsewhere to conduct this study. Paper context Male involvement can promote positive maternal health outcomes. With the increased focus on gender equality in Rwanda, this paper present men’s reflections on their roles during childbearing. While an engaged and caring partner was presented as the ideal, gendered expectations from the maternal health system and society were perceived as limiting this ideal. To promote gender equitable ideas, maternity care could be more responsive for men to actualize their wish for increased engagement. ==== Refs References 1 United Nations Report of the International Conference on Population and Development 1995 New York United Nations 2 Mullany BC Becker S Hindin MJ The impact of including husbands in antenatal health education services on maternal health practices in urban Nepal: results from a randomized controlled trial Health Educ Res 2007 22 166 76 16855015 3 Barker G Greene M Nascimento M Segundo M Ricardo C Taylor A Men who Care: a multi-country qualitative study of men in non-traditional caregiving roles 2012 Washington, DC International Center for Research on Women (ICRW) Rio de Janeiro: Instituto Promundo 4 Comrie-Thomson L Tokhi M Ampt F Portela A Chersich M Khanna R Challenging gender inequity through male involvement in maternal and newborn health: critical assessment of an emerging evidence base Cult Health Sex 2015 17 Suppl 2 177 89 doi: http://dx.doi.org/10.1080/13691058.2015.1053412 5 Dudgeon MR Inhorn MC Men's influences on women's reproductive health: medical anthropological perspectives Soc Sci Med 2004 59 1379 95 15246168 6 Tweheyo R Konde-Lule J Tumwesigye NM Sekandi JN Male partner attendance of skilled antenatal care in peri-urban Gulu district, Northern Uganda BMC Pregnancy Childbirth 2010 10 53 20846369 7 Yargawa J Leonardi-Bee J Male involvement and maternal health outcomes: systematic review and meta-analysis J Epidemiol Community Health 2015 69 604 12 doi: http://dx.doi.org/10.1136/jech-2014-204784 25700533 8 Levtov RG Barker G Contreras-Urbina M Heilman B Verma R Pathways to gender-equitable men: findings from the international men and gender equality survey in eight countries Men Masculinities 2014 17 467 501 doi: http://dx.doi.org/10.1177/1097184X14558234 9 Dumbaugh M Tawiah-Agyemang C Manu A Ten AGH Kirkwood B Hill Z Perceptions of, attitudes towards and barriers to male involvement in newborn care in rural Ghana, West Africa: a qualitative analysis BMC Pregnancy Childbirth 2014 14 269 25112497 10 Dunlap J Foderingham N Bussell S Wester CW Audet CM Aliyu MH Male involvement for the prevention of mother-to-child HIV transmission: a brief review of initiatives in East, West, and Central Africa Current HIV AIDS Rep 2014 11 109 18 11 Kaye DK Kakaire O Nakimuli A Osinde MO Mbalinda SN Kakande N Male involvement during pregnancy and childbirth: men's perceptions, practices and experiences during the care for women who developed childbirth complications in Mulago Hospital, Uganda BMC Pregnancy Childbirth 2014 14 54 24479421 12 Mbekenga CK Lugina HI Christensson K Olsson P Postpartum experiences of first-time fathers in a Tanzanian suburb: a qualitative interview study Midwifery 2011 27 174 80 20385433 13 Rwanda Ministry of Gender and Family Promotion National Gender Policy 2010 Kigali, Rwanda 14 Uvuza JN Hidden inequalities: Rwandan female politicians’ experiences of balancing family and political responsibilities 2014 Newcastle University 15 Slegh H Barker G Kimonyo A Ndolimana P Bannerman M ‘I can do women's work’: reflections on engaging men as allies in women's economic empowerment in Rwanda Gender & Development 2013 21 15 30 16 Doyle K Kato-Wallace J Kazimbaya S Barker G Transforming gender roles in domestic and caregiving work: preliminary findings from engaging fathers in maternal, newborn, and child health in Rwanda Gend Dev 2014 22 515 31 17 Wallace C Haerpfer C Abbott P Women in Rwandan politics and society Int J Sociol 2009 38 111 25 18 Rwanda Ministry of Health Rwanda demographic and health survey 2000: key findings 2000 Kigali, Rwanda Rwanda Ministry of Health 19 National Institute of Statistic of Rwanda (NISR) [Rwanda], Ministry of Health (MOH) [Rwanda], ICF International Rwanda Demographic and Health Survey 2014–2015 2015 Rockville, MD NISR 20 Nathan LM Shi Q Plewniak K Zhang C Decentralizing maternity services to increase skilled attendance at birth and antenatal care utilization in rural Rwanda: a prospective cohort study Matern Child Health J 2015 19 1949 55 25652061 21 Pierce H Heaton TB Hoffmann J Increasing maternal healthcare use in Rwanda: implications for child nutrition and survival Soc Sci Med 2014 24 61 7 24607667 22 Logie DE Rowson M Ndagije F Innovations in Rwanda's health system: looking to the future Lancet 2008 372 256 61 18619670 23 Jennings L Na M Cherewick M Hindin M Mullany B Ahmed S Women's empowerment and male involvement in antenatal care: analyses of Demographic and Health Surveys (DHS) in selected African countries BMC Pregnancy Childbirth 2014 14 297 25174359 24 National Institute of Statistic of Rwanda (NISR) [Rwanda], Ministry of Health (MOH), ICF International Rwanda Demographic and Health Survey 2010 2012 Calverton, MD NISR 25 Påfs J Musafili A Binder-Finnema P Klingberg-Allvin M Rulisa S Essén B ‘They would never receive you without a husband’: paradoxical barriers to antenatal care scale-up in Rwanda Midwifery 2015 31 1149 56 26471934 26 Connell R Gender: in world perspective 2009 Cambridge Polity 27 Connell R Gender, health and theory: conceptualizing the issue, in local and world perspective Soc Sci Med 2012 74 1675 83 doi: http://dx.doi.org/10.1016/j.socscimed.2011.06.006 21764489 28 Rwanda Ministry of Health Indicators [Internet] 2012 [cited 2016 Feb 12]. Available from: http://moh.gov.rw/english/ 29 Say L Souza JP Pattinson RC Maternal near miss – towards a standard tool for monitoring quality of maternal health care Best Pract Res Clin Obstet Gynaecol 2009 23 287 96 19303368 30 Pacagnella RC Cecatti JG Osis MJ Souza JP The role of delays in severe maternal morbidity and mortality: expanding the conceptual framework Reprod Health Matters 2012 20 155 63 22789093 31 Haddad SM Cecatti JG Souza JP Sousa MH Parpinelli MA Costa ML Applying the maternal near miss approach for the evaluation of quality of obstetric care: a worked example from a multicenter surveillance study BioMed Res Int 2014 5 1 10 32 Roost M Jonsson C Liljestrand J Essen B Social differentiation and embodied dispositions: a qualitative study of maternal care-seeking behaviour for near-miss morbidity in Bolivia Reprod Health 2009 6 1 10 19126239 33 Bernard HR Research methods in anthropology: Qualitative and Quantitative Approaches Library 2006 Oxford, UK Rowman Altamira 34 Guba EG Lincoln YS Fourth generation evaluation 1989 Beverly Hills, CA Sage 35 Squires A Language barriers and qualitative nursing research: methodological considerations Int Nurs Rev 2008 55 265 73 19522941 36 Dolan A Coe C Men, masculine identities and childbirth Sociol Health Illn 2011 33 1019 34 21561461 37 Siu GE Wight D Seeley JA Masculinity, social context and HIV testing: an ethnographic study of men in Busia district, rural eastern Uganda BMC Public Health 2014 14 33 24417763 38 Mumtaz Z Salway S Understanding gendered influences on women's reproductive health in Pakistan: moving beyond the autonomy paradigm Soc Sci Med 2009 68 1349 56 doi: http://dx.doi.org/10.1016/j.socscimed.2009.01.025 19232810 39 Van LW Matthews Z Achadi E Ancona C Campbell J Channon A Country experience with strengthening of health systems and deployment of midwives in countries with high maternal mortality Lancet 2014 384 1215 25 24965819 40 Ganle JK Dery I ‘What men don't know can hurt women's health’: a qualitative study of the barriers to and opportunities for men's involvement in maternal healthcare in Ghana Reprod Health 2015 12 93 26452546 41 Berry NS Who's judging the quality of care? Indigenous Maya and the problem of ‘not being attended’ Med Anthropol 2008 27 164 89 18464128

PMC005xxxxxx/PMC5002035.txt

==== Front Glob Health ActionGlob Health ActionGHAGlobal Health Action1654-97161654-9880Co-Action Publishing 3226710.3402/gha.v9.32267Original ArticleThe correlation between dengue incidence and diurnal ranges of temperature of Colombo district, Sri Lanka 2005–2014 Ehelepola N. D. B. 1*Ariyaratne Kusalika 21 The Teaching (General) Hospital – Kandy, Kandy, Sri Lanka2 Lanka Hydraulic Institute, Moratuwa, Sri Lanka* Correspondence to: N. D. B. Ehelepola, The Teaching (General) Hospital – Kandy, Kandy, Sri Lanka, Email: drehelepola@gmail.com25 8 2016 2016 9 10.3402/gha.v9.3226714 5 2016 28 7 2016 29 7 2016 © 2016 N. D. B. Ehelepola and Kusalika Ariyaratne2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 International License, allowing third parties to copy and redistribute the material in any medium or format and to remix, transform, and build upon the material for any purpose, even commercially, provided the original work is properly cited and states its license.Background Meteorological factors affect dengue transmission. Mechanisms of the way in which different diurnal temperatures, ranging around different mean temperatures, influence dengue transmission were published after 2011. Objective We endeavored to determine the correlation between dengue incidence and diurnal temperature ranges (DTRs) in Colombo district, Sri Lanka, and to explore the possibilities of using our findings to improve control of dengue. Design We calculated the weekly dengue incidence in Colombo during 2005–2014, after data on all of the reported dengue patients and estimated mid-year populations were collected. We obtained daily maximum and minimum temperatures from two Colombo weather stations, averaged, and converted them into weekly data. Weekly averages of DTR versus dengue incidence graphs were plotted and correlations observed. The count of days per week with a DTR of >7.5°C and <7.5°C were also calculated. Wavelet time series analysis was performed to determine the correlation between dengue incidence and DTR. Results We obtained a negative correlation between dengue incidence and a DTR>7.5°C with an 8-week lag period, and a positive correlation between dengue incidence and a DTR<7.5°C, also with an 8-week lag. Conclusions Large DTRs were negatively correlated with dengue transmission in Colombo district. We propose to take advantage of that in local dengue control efforts. Our results agree with previous studies on the topic and with a mathematical model of relative vectorial capacity of Aedes aegypti. Global warming and declining DTR are likely to favor a rise of dengue, and we suggest a simple method to mitigate this. denguediurnal temperature rangemathematical modelsclimate changeurban heat islands ==== Body Introduction Dengue is a viral infection with life-threatening forms, transmitted by Aedes mosquitoes, and a major global public health problem (1). Meteorological factors, such as rainfall, count of rainy (and wet) days, humidity, temperature, wind, and duration of sunshine, have been demonstrated to influence dengue incidence in hundreds of studies performed around the world (2–4). Even though most of them have studied the relationship between dengue transmission and temperature, only very few have analyzed the correlation between dengue and diurnal ranges of temperature (4). Other factors, such as herd immunity, introduction of new virus phenotypes to the population, efficiency of preventive measures, population movements, urbanization, housing and refuse disposal methods, and knowledge and attitudes of the public, are also known to influence dengue incidence. Diurnal temperature range (DTR) is the difference between daily maximum and minimum temperatures. Several studies concerning DTR dengue correlation were carried out after the first publication appeared in 2011 (4). We recently demonstrated that dengue incidence in Kandy city, in the central hill country of Sri Lanka, is correlated with DTR (5). We decided to perform this study for the following reasons: Dengue is now endemic in various localities in more than 100 countries (1), hyper endemic in Sri Lanka, and one of the world's fastest spreading infections. Although dengue is a global problem, its correlation with DTR has so far only been epidemiologically demonstrated in three localities: Thailand, Bangladesh, and Sri Lanka (4–7). There is a local (Sri Lanka) and global tendency of declining DTR with ongoing climate change (8, 9), which may further facilitate dengue transmission in many areas around the globe (6). Life traits of Aedes aegypti species from different geographical locations were shown to be varied by some laboratory studies (7). Considering the previously mentioned reasons, the need for more epidemiological studies on the topic in more localities is clear. Entomological studies carried out in laboratories mimicking DTR in Mae Sot, Thailand, have shown mechanisms of how DTR affects dengue transmission (4, 7, 10). Aedes vectors’ lifespans are shorter, and their susceptibility to dengue virus infection is reduced, when DTRs fluctuate widely around the same mean temperature, when this temperature is >18°C, as in Colombo (4). Large DTRs prolong the aquatic part of the Aedes life cycle, lower larval survival rates, and reduce adult female reproductive output (7). Wide fluctuations of DTR around a mean of 26°C reduce midgut infection rates of the vector and tend to increase the dengue virus's extrinsic incubation period (10). It is important to ascertain the applicability of those laboratory findings to the real world, through epidemiological studies. Previous locations where dengue DTR correlation was studied (Mae Sot, Dhaka, Kandy) have an approximate mean annual temperature of 25°C–26°C and annual rainfall below 2,000 mm. In contrast, we estimated for our study period that Colombo district had a mean temperature of 28.1°C (average of two weather stations), with annual rainfall in Colombo being approximately 2,350 mm (11). Colombo is closer to the equator and has a lower average DTR (typically DTR becomes smaller toward the equator) compared with the above three locations. Colombo district, located in the western coast low lands of Sri Lanka (Fig. 1), has a warmer and wetter climate than Kandy (located in the central highlands) and has a lower DTR (DTR typically increases with elevation). Fig. 1 A map of Sri Lanka (inset) and our study area. Colombo district in the west coast is shaded in light brown color in contrast to the other 24 districts. Locations of the two weather stations from which we obtained our weather data are depicted. We have estimated two dengue incidences of Sri Lanka. In 2005, it was 31 per 100,000 population, but increased to 228 per 100,000 population in 2014 (12, 13). We used population data from the Sri Lanka Department of Census and Statistics for this estimation. However, those reported dengue cases represent only a fraction of total cases (2). By comparison, one estimate shows the average dengue incidence of the World Health Organization's (WHO) Southeast Asia region for the 2000–2010 period as 13 per 100,000 population (14). This is the world's most severely affected region by dengue and includes Sri Lanka. Therefore, the burden of dengue in Sri Lanka is clearly greater than the average for this region. The dengue case fatality rate in Sri Lanka was 0.5% in 2005, increasing to 1% in 2009 and falling to 0.2% in 2014 (12, 13). There is an interesting mathematical model regarding changes of relative vectorial capacity (rVc) of Aedes aegypti mosquitoes in different DTRs around varying mean temperatures (15). By performing several epidemiological studies, we can ascertain the usefulness of that model in the real world and further improve it. Such models can be used to forecast future dengue incidences with ongoing changes of DTR due to climate change. Reliable predictions will allow better preparation and therefore better management of dengue. Study setting Colombo has a tropical wet climate, category Af according to the Koppen Geiger classification (11). We estimated the mean daily temperature of Colombo district as 28.1°C, and the minimum and maximum average temperatures for our study period as 24.9°C and 31.2°C, respectively. Colombo's population, according to the 2012 census, was 2.31 million. Colombo is the smallest (699 km2), but most densely populated (3,333 people per km2), district in Sri Lanka. Objectives and hypotheses Our objectives were to determine the correlation between Colombo's dengue incidence and DTR, and compare our results with those of similar studies. Then, we endeavored to explore possible ways to improve dengue control using those correlation patterns, as well as other available evidence. Our hypothesis was that a large DTR has a negative correlation with the incidence of dengue in Colombo. Method Ethical statement Only notified dengue patient numbers were collected (without any information about their identity). We obtained clearance (exemption from detailed ethics review) from the ethical review committee of the Peradeniya medical faculty (2015/EC/25). Data The numbers of notified dengue cases from the Colombo district from January 1, 2005 to December 31, 2014 were collected by analyzing weekly health ministry epidemiology reports. Mid-year population data of Colombo for the same period were obtained from the Department of Census and Statistics. Daily minimum and maximum temperatures of both weather stations of the Sri Lanka Department of Meteorology in Colombo district (situated at Colombo seven and Ratmalana=Rathmalana) relevant to our study period were purchased. We decided to take average values of both weather stations, in order to compensate for differences in weather within the district, as there were many people who live in one part and work in another part of Colombo. Most Aedes mosquitoes spend their entire life near the dwelling where they were born (16); so data of a distant weather station may be of little relevance to their life cycle. Hence, we selected weather stations within the study area. Analysis The weekly dengue incidence of Colombo for 2005–2014 was calculated. Temperature data were converted to weekly values and averaged. Weekly averages of diurnal ranges of temperature were also calculated. We plotted time series graphs of median and mean weekly dengue incidences versus weekly averages of DTRs during the course of 52 weeks of the year and observed correlation patterns (Figs. 2 and 3). Fig. 2 Changes in weekly average diurnal temperature range (DTR) in °C (red), and weekly dengue incidence over the course of 52 weeks of the year, 2005–2014. x-axis: weeks; primary y-axis: median weekly dengue incidence (blue) and mean weekly dengue incidence (green); secondary y-axis: DTR in °C. Fig. 3 Changes in weekly average minimum (black), maximum (yellow), and mean (purple) temperatures and median weekly dengue incidence over the course of all 52 weeks of the year, for 2005–2014. x-axis: weeks; primary y-axis: temperature in °C; secondary y-axis: median weekly dengue incidence (blue). We observed that a DTR>7.5°C is followed by a decline of dengue incidence after several weeks (however, a DTR>7.5°C is not intended to be a precise threshold value). We wanted to further verify the observed pattern by analyzing data in another way, using the wavelet analysis technique. During a particular week, there are sometimes several days with extreme DTR values. We performed wavelet analyses to determine the relationship between the count of days per week where DTR was >7.5°C and <7.5°C versus the weekly dengue incidence. Wavelet analysis was carried out following the same procedure as per our Kandy study, published in this journal (5). Here we give a brief description; cross wavelet transform (XWT) and wavelet coherence (WTC) can be used for examining relationships in time frequency space between two time series. A large common power in XWT and a consistent phase relationship in WTC indicate causality between the two time series (17). Continuous wavelet transform (CWT) was calculated for count of days per week with a DTR>7.5°C and <7.5°C. Then, XWT and WTC were calculated for each of them and dengue incidence. Vectors were indicative of the phase difference. A horizontal arrow pointing from left to right signifies ‘in phase’, with an arrow pointing vertically upward indicating the second series lagging the first by 90°. WTC is defined as the square of the cross-spectrum, normalized by the individual power spectra. This gives a quantity between 1 and 0, and measures the cross-correlation between two time series, as a function of frequency. Because the wavelet transform is a band-pass filter with a known wavelet function, by summing over a subset of the scales it is possible to reconstruct the original time series. In this study, the period, which gave the highest coherence among the dengue incidences, and each of the DTR variables were identified. The wavelet filtered time series for this period was reconstructed, and the lagging time was estimated. MATLAB R2013a software was used for wavelet analysis, and Microsoft Office 2007 software was used for other work. Results Figures 2 and 3 clearly illustrate that large weekly DTRs were followed by a reduction of dengue incidence after several weeks. According to Fig. 3, from weeks 22 to 37 the mean weekly temperature was between 28 and 29°C; from Fig. 2, it can be seen that between weeks 23 and 34, DTR was lowest (4°C–5°C). Both mean and median dengue incidence values increased after a few weeks lag. According to Figs. 2 and 3, from weeks 3 to 9 the mean temperature was lower (27°C–28°C) and the DTR was higher (7.2°C–8.7°C). DTR was highest during the fourth week. We observed a fall in both median and mean dengue incidences after a lag of several weeks (Figs. 1 and 3). Wavelet analyses showed that the count of days with a DTR>7.5°C per week and weekly dengue incidence were negatively correlated, with an 8 (7.9) week lag. The count of days with a DTR<7.5°C per week and weekly dengue incidence were positively correlated, with an 8 (7.8) week lag. Figure 4 illustrates the results of the wavelet analysis of weekly dengue incidence versus the count of days per week with a DTR>7.5°C. Fig. 4 Results of wavelet analyses of weekly dengue incidence versus the count of days per week with a DTR>7.5°C: (a) continuous wavelet transform (CWT) variations; (b) wavelet power of CWT; (c) crosswavelet transform (XWT) variations; (d) wavelet power of XWT; (e) wavelet coherence (WTC); (f) wavelet power of WTC; and (g) reconstructed time series for selected periods. The term ‘period’ in the vertical axis indicates duration of cycle (in years). For a, c, and e, there are color-coded columns on the right side of the main figure. They indicate the strength of coherence, in which dark blue and dark red indicate lowest and highest coherence, respectively. Discussion The study's results confirmed our hypothesis of large DTRs having negative correlations with dengue incidence in Colombo. This was shown by our time series graphs and further confirmed by results of our wavelet analyses. We also observed that small DTRs were favorable for dengue transmission. The correlation patterns we found were similar to those from studies carried out in Kandy, Sri Lanka, for 2003–2012 (5), and in Dhaka, Bangladesh, for 2000–2009 (6). However, the lag periods in our study are longer. No information is available regarding lag periods in Mae Sot, Thailand. Past dengue incidence versus meteorological factors correlation studies have shown that lag periods vary from a few weeks to a few months, in various localities (2). Hence, our result is not unusual, although we could not determine the precise reason for longer lag periods. In Sri Lanka, once a vector with dengue virus in its saliva bites a healthy person, it usually takes 2–3 weeks for that person to develop symptoms, go to a hospital, get diagnosed, and then get reported (2). Graphs depicting rVc of the Aedes vector according to different DTRs around different mean temperatures already exist, such as Fig. 1 of reference (15). We use a part of that here as Fig. 5. Fig. 5 The effect of temperature and DTR on the relative vectorial capacity (rVc) of Aedes aegypti mosquitoes. The color-coded column on the right side of the graph describes the rVc value. A higher rVc corresponds to increased dengue epidemic potential. This graph was developed by theoretical mathematical modeling of dengue transmission. As described previously, in Fig. 3 from weeks 22 to 37 the mean weekly temperature was between 28 and 29°C, and according to Fig. 2, between weeks 23 and 34 DTR was lowest (4°C–5°C). According to the rVc graph (Fig. 5), this combination of mean temperature and DTR enhances rVc of the vector mosquito. Therefore, it is highly conducive for dengue transmission. This is confirmed in Fig. 2 by the increase of mean and median dengue incidence values following a lag period. According to Figs. 2 and 3, from weeks 3 to 9 the mean temperature was lower (27°C–28°C) and DTR was higher (7.2°C–8.7°C). DTR was highest during the fourth week. We observe a decrease of both median and mean dengue incidences after several weeks (Figs. 2 and 3). These findings are also in agreement with the rVc graph. Both Ae. aegypti and Aedes albopictus vectors are present in Colombo district, but information regarding their ratios of contribution to Colombo dengue incidence is unavailable at present. There are differences in their biology, and it is interesting to see this theoretical model (15) (which considers only Ae. aegypti as dengue vector) agrees with our findings, which were derived from epidemiological data in a place where two dengue vectors are found. We found a similar occurrence in our Kandy study as well (5). In Kandy, with an average mean temperature of 25.1°C, when DTR is >10°C, dengue incidence declines after a lag period (5). In Colombo, with an average mean temperature of 28.1°C, a lower DTR (>7.5°C) causes similar phenomena. This also agrees with the rVc graph (Fig. 5). Results of the present and Kandy study (5) illustrate that an 8.5°C DTR around mean temperatures 25.1°C and 28.1°C results in opposite effects on dengue transmission in Kandy and Colombo, respectively. We propose a possible explanation of this, in the context of the rVc of the vector, as follows: In the rVc graph, if we consider a zero DTR as the baseline, an 8.5°C DTR reduces the rVc more steeply when mean temperature is 28.1°C, compared with a mean of 25.1°C. Therefore, it appears that dengue transmission in Kandy and Colombo generally follows the pattern of rVc of the vector. This supports the validity, and therefore the usefulness, of this mathematical model in the real world. However, we have also noted the following weakness. In Colombo during week 5, there was a DTR of 7.9°C around a mean temperature of 27.6°C (Fig. 3), which is unfavorable for dengue transmission. Nonetheless, that combination corresponds to a larger rVc than when the mean temperature was 25°C–25.5°C, even with a zero DTR (Fig. 5). However, in Kandy a DTR of 6.5°C–7°C around a 25°C–25.5°C mean was also highly conducive for dengue transmission, as depicted in Fig. 2 of reference (5). Here, our findings disagree with the rVc graph. Our Fig. 2 illustrates a small rise of dengue incidence in the 38–52 weeks, despite a rising DTR during this period. The possible explanations are as follows. One likely reason is that during this period, Colombo receives more rain compared with the first weeks of the year (Colombo receives its lowest rainfall in January and February). Rainfall has also been demonstrated to be correlated with dengue incidence in Colombo (18). We suggest this may be due to another reason. Higher humidity at the same temperature has been demonstrated to increase the longevity of the Aedes vector (19). We made a time series graph of humidity for our study period, using data from the same two weather stations, in order to explore this possibility. Accordingly (Fig. 6), average humidity remained higher (>80%) during the 38–52 weeks, compared with the first 13 weeks (75–80% range). Fig. 6 Changes in weekly average minimum (brown), maximum (black), and mean (purple) humidity% and median weekly dengue incidence, over the course of all 52 weeks of the year, for 2005–2014. x-axis: weeks; primary y-axis: weekly average humidity; secondary y-axis: median weekly dengue incidence (blue). Mean temperatures remained similar during weeks 38–51 and the first 13 weeks (Fig. 3). Increase of longevity exponentially increases rVc of the vector (4). However, Colombo's mean temperatures remained around 28°C during this period, whilst that entomological study was performed at 25°C. During the first 13 weeks, the magnitude of DTR was higher in Colombo (Fig. 2), and average humidity was lower (Fig. 6). We propose that the lowest dengue incidence, seen in week 15, can be attributed to a high DTR of preceding weeks. Low rainfall and average humidity may also have contributed to this to a certain extent. Possible ways to use information gathered for improvement of dengue control In all places where local dengue DTR correlation has been studied in detail so far (Mae Sot, Dhaka, Kandy, and Colombo), mean temperature was relatively low (but in the 18°C–33°C range), and DTR was higher during the first quarter of the year, and was unfavorable for dengue transmission. This period corresponds to the winter in the Northern Hemisphere. On a global scale, winter warming due to global climate changes is more rapid than summer warming (9, 20). Increases of minimum temperatures are greater than the rise of maximum temperatures in Sri Lanka and in the world, although both of them rise (8, 9). However, this does not happen everywhere on the earth. This implies that in many places in the Northern Hemisphere during the first quarter of the year, the mean temperature is likely to rise, and DTR is likely to become smaller in the coming decades. When we consider the effect of these changes on the life cycles of the Aedes vector and the dengue virus (as discussed in the introduction), it is likely to result in more vectors with dengue virus in their saliva, and they will bite more frequently with rising temperatures (5). Considering the great majority of the world's population, and more specifically the population at risk of getting dengue, live in the Northern hemisphere, we may expect more dengue patients even during the first quarter of the year, with ongoing climate changes. To our best knowledge, there have been no previous publications regarding this risk. We also believe that there is a simple, sustainable, and feasible method to mitigate this problem. We propose to popularize local application of mosquito repellents, particularly in the dawn and evenings. This will supplement existing dengue control programs in Colombo and in other areas with high risk of dengue, especially during periods with low DTR and high average temperatures. This proposal has many additional benefits (2, 5). In Colombo and other parts of Sri Lanka, the cornerstone of dengue prevention is elimination of vector breeding sites and immature forms of the vector. These preventive activities become vigorous at the onset of the monsoon rain seasons and when dengue incidence rises. Even though the smallest in area, Colombo is the hub of the nation's administration and economy and makes the largest contribution to the gross national product. At the same time, Colombo district also typically reports the highest number of dengue cases. Hence, control of dengue in Colombo is very important. Despite vigorous preventive campaigns, Colombo's dengue incidence was 81 per 100,000 population in 2005 and rose to 622 per 100,000 population in 2014. Therefore, there is a clear need for additional (or alternative) efforts to control dengue in Colombo. We suggest maintaining vigorous dengue control programs during the first quarter of the year, and capitulate on the weather conditions not conducive for dengue transmission (shown in Figs. 2 and 3), and maximally suppress dengue transmission during this period. This will help minimize the chances of an epidemic during subsequent weeks. In the recent past, there have been many advances in the use of sophisticated mathematical models in dengue epidemiology and in other areas of public health. This is a relatively new, little known concept to most Sri Lankan (and most other developing world) health care workers. We hope our findings of the validity and potential uses of a mathematical model of rVc of the dengue vector (15) will generate interest among our colleagues regarding potential uses of these tools in common public health problems. Limitations of the study Reliability of our results depends upon the fidelity of our data. Notified dengue cases are only the tip of the dengue iceberg. A study performed in Colombo in 2008 showed the existence of approximately 30 primary dengue cases in children (<12 years) for every single case that was notified (21). However, notified dengue cases were the best practical option available to us. Urban heat islands (UHI) and cold islands have been demonstrated in Colombo district (22, 23). In one study (23), a difference in temperature up to 7°C was found between different urban locations, and marked temperature differences between sunlit and shaded areas were recorded during the daytime. Nocturnal heat islands, which were 3°C warmer than the surroundings, were also found. The pattern of change of both temperature and humidity during the course of a day differs in UHIs (22, 23). There are two weather stations for the Department of Meteorology in the district, and we took averages of the recordings of both of them, in order to compensate for intradistrict differences in weather. Most of the heat islands and the majority of the population are on the western side of the district, and both of our weather stations are also from the western side of the district. UHIs in Colombo (and elsewhere) are typically areas with high population density and large floating populations. There, the temperatures are higher in mornings and evenings (23). The Aedes vector bites more frequently in the mornings and evenings, and higher temperatures further increase biting rates. This is more conducive for dengue transmission. However, considering the rVc graph (Fig. 5), the mean temperatures of >31°C are not particularly favorable for dengue transmission. DTRs are also higher in UHIs, which are unfavorable for dengue transmission, but large DTRs around a mean of >33°C again favor dengue transmission. Dedicated future studies may help us to better understand dengue dynamics in UHIs. There are hundreds of published studies on dengue temperature correlation (2) in urban areas. To our best knowledge, none of the other researchers considered the possible influence of UHIs on their results. Conclusions Large DTRs were negatively correlated with dengue transmission in Colombo district. An existing mathematical model of rVc of Ae. aegypti in relation to different average temperatures, and different DTRs around them, agrees with our findings. Such models are likely to be useful for studying and predicting changes in dengue epidemiology due to climate change. Our detection of a negative correlation between large DTR and dengue incidence in Kandy was confirmed in Colombo. Similar studies in areas with different climate and dengue transmission patterns will help to further verify our findings. Acknowledgements The authors thank Dr. Aslak Grinsted for allowing usage of MATLAB codes available online. Authors’ contributions KA performed wavelet analysis and wrote the chapter on wavelet analysis. NDBE contributed to the rest of the work. Both the authors approved the draft. Conflict of interest and funding The authors declare that they have no conflicts of interest. Paper context Entomological studies and a mathematical model show diurnal temperature range (DTR) affects dengue transmission, but there is a paucity of epidemiological evidence. We have demonstrated that large DTRs are unfavorable for dengue transmission in Colombo district, Sri Lanka. We propose ways to use our findings to improve local dengue control and to mitigate a potential rise of dengue due to ongoing global declines in DTR. ==== Refs References 1 World Health Organization Dengue and severe dengue 2016 Available from: http://www.who.int/mediacentre/factsheets/fs117/en [cited 20 February 2016] 2 Ehelepola NDB Ariyaratne K Buddhadasa WMNP Ratnayake S Wickramasinghe M A study of the correlation between dengue and weather in Kandy City, Sri Lanka (2003–2012) and lessons learned Infect Dis Poverty 2015 4 42 doi: http://dx.doi.org/10.1186/s40249-015-0075-8 26403283 3 Centers for Disease Control and Prevention of the U.S.A. Dengue and climate 2016 Available from: http://www.cdc.gov/dengue/entomologyEcology/climate.html [cited 20 February 2016] 4 Lambrechts L Paaijmans KP Fansiri T Carrington LB Kramer LD Thomas MB Impact of daily temperature fluctuations on dengue virus transmission by Aedes aegypti Proc Natl Acad Sci USA 2011 108 7460 5 doi: http://dx.doi.org/10.1073/pnas.1101377108 21502510 5 Ehelepola NDB Ariyaratne K The interrelationship between dengue incidence and diurnal ranges of temperature and humidity in a Sri Lankan city and its potential applications Glob Health Action 2015 8 doi: http://dx.doi.org/10.3402/gha.v8.29359 6 Sharmin S Glass K Viennet E Harley D Interaction of mean temperature and daily fluctuation influences dengue incidence in Dhaka, Bangladesh PLoS Negl Trop Dis 2015 9 e0003901. doi: http://dx.doi.org/10.1371/journal.pntd.0003901 7 Carrington LB Seifert SN Willits NH Lambrechts L Scott TW Large diurnal temperature fluctuations negatively influence Aedes aegypti (Diptera: Culicidae) life-history traits J Med Entomol 2013 50 43 51 doi: http://dx.doi.org/10.1603/ME11242 23427651 8 Climate Change in Sri Lanka 2016 Department of Meteorology of Sri Lanka Available from: http://www.meteo.gov.lk/index.php?option=com_content&view=article&id=14&Itemid=133&lang=en#temperature-change [cited 21 February 2016] 9 Deke A Climate change rule of thumb: cold ‘things’ warming faster than warm things 2016 The National Oceanic and Atmospheric Administration of the U.S.A Available from: https://www.climate.gov/news-features/blogs/beyond-data/climate-change-rule-thumb-cold-things-warming-faster-warm-things [cited 21 February 2016] 10 Carrington LB Seifert SN Armijos MV Lambrechts L Scott TW Reduction of Aedes aegypti vector competence for dengue virus under large temperature fluctuations Am J Trop Med Hyg 2013 88 689 97 doi: http://dx.doi.org/10.4269/ajtmh.12-0488 23438766 11 Climate-data.org Climate-Colombo 2016 Available from: http://en.climate-data.org/location/944 [cited 21 February 2016] 12 National Dengue Control Unit of Sri Lanka Sri Lankan situation 2016 Available from: http://dengue.health.gov.lk/index.php/information-on-dengue/sri-lankan-situation [cited 7 July 2016] 13 Dengue Vaccine Initiative Dengue in Sri Lanka: burden, challenges and prevention strategies – an exclusive interview with Dr. Hasitha Tissera 2015 Available from: http://www.denguevaccines.org/dengue-sri-lanka-burden-challenges-and-prevention-strategies-exclusive-interview-dr-hasitha-tissera [cited 7 July 2016] 14 Hoti SL Assessment of epidemiology of dengue in Sri Lanka in relation to intervention measures 2011 Pondicherry, India Vector Control Research Centre Report of visit to Colombo during 9–15 Oct 2011 (as Consultant, WHO/SEARO) 15 Liu-Helmersson J Stenlund H Wilder-Smith A Rocklov J Vectorial capacity of Aedes aegypti : effects of temperature and implications for global dengue epidemic potential PLoS One 2014 9 e89783. doi: http://dx.doi.org/10.1371/journal.pone.0089783 16 World Health Organization Dengue control – the mosquito 2016 Available from: http://www.who.int/denguecontrol/mosquito/en/ [cited 22 February 2016] 17 Grinsted A Moore JC Jevrejeva S Application of the cross wavelet transform and wavelet coherence to geophysical time series Nonlinear Process Geophys 2004 11 561 6 doi: http://dx.doi.org/10.5194/npg-11-561-2004 18 Pathirana S Kawabata M Goonatilake R Study of potential risk of dengue disease outbreak in Sri Lanka using GIS and statistical modelling J Rural Trop Publ Health 2009 8 8 17 19 Costa EA Santos EM Correia JC Albuquerque CM Impact of small variations in temperature and humidity on the reproductive activity and survival of Aedes aegypti (Diptera, Culicidae) Rev Bras Entomol 2010 54 488 93 20 berkeleyearth.org Global warming and changing the range of seasonal temperatures 2016 Available from: http://berkeleyearth.org/graphics/physical-effects-of-warming [cited 22 February 2016] 21 Tam CC Tissera H de Silva AM De Silva AD Margolis HS Amarasinge A Estimates of dengue force of infection in children in Colombo, Sri Lanka PLoS Negl Trop Dis 2013 7 e2259. doi: http://dx.doi.org/10.1371/journal.pntd.0002259 22 Emmanuel R Johansson E Influence of urban morphology and sea breeze on hot humid microclimate: the case of Colombo, Sri Lanka Clim Res 2006 3 189 200 23 Halwatura RU Nishad NM Effects of greenery on city comfort in different micro climatic conditions 2013 Available from: http://www.civil.mrt.ac.lk/conference/ICSBE2012/SBE-12-139.pdf [cited 24 February 2016]

PMC005xxxxxx/PMC5002036.txt

==== Front Glob Health ActionGlob Health ActionGHAGlobal Health Action1654-97161654-9880Co-Action Publishing 3211610.3402/gha.v9.32116Original ArticleSeasonality of hospital admissions for mental disorders in Hanoi, Vietnam Trang Phan Minh 1*Rocklöv Joacim 1Giang Kim Bao 2Nilsson Maria 11 Department of Public Health and Clinical Medicine, Epidemiology and Global Health, Umeå University, Umeå, Sweden2 Institute for Preventive Medicine and Public Health, Hanoi Medical University, Hanoi, Vietnam* Correspondence to: Phan Minh Trang, University Hospital, Building 9B, Entrance X5, SE-90187, Umeå, Sweden, Email: tranghealth@gmail.comResponsible Editor: Stig Wall, Umeå University, Sweden. 25 8 2016 2016 9 10.3402/gha.v9.3211602 5 2016 30 6 2016 15 7 2016 © 2016 Phan Minh Trang et al.2016This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 International License, allowing third parties to copy and redistribute the material in any medium or format and to remix, transform, and build upon the material for any purpose, even commercially, provided the original work is properly cited and states its license.Background Some studies have shown a relationship between seasonality in weather patterns and depressive and behavioural disorders, especially in temperate climate regions. However, there is a lack of studies describing the seasonal patterns of hospital admissions for a variety of mental disorders in tropical and subtropical nations. The aim of this study has been to examine the relationship between seasons and daily hospital admissions for mental disorders in Hanoi, Vietnam. Designs A 5-year database (2008–2012) compiled by Hanoi Mental Hospital covering mental disorder admissions diagnosed by the International Classification of Diseases 10 was analysed. A negative binominal regression model was applied to estimate the associations between seasonality and daily hospital admissions for mental disorders, for all causes and for specific diagnoses. Results The summer season indicated the highest relative risk (RR=1.24, confidence interval (CI)=1.1–1.39) of hospital admission for mental disorders, with a peak in these cases in June (RR=1.46, CI=1.19–1.7). Compared to other demographic groups, males and the elderly (aged over 60 years) were more sensitive to seasonal risk changes. In the summer season, the RR of hospital visits among men increased by 26% (RR=1.26, CI=1.12–1.41) and among the elderly by 23% (RR=1.23, CI=1.03–1.48). Furthermore, when temperatures including minimum, mean, and maximum increased 1°C, the number of cases for mental disorders increased by 1.7%, 2%, and 2.1%, respectively. Conclusion The study results showed a correlation between hospital admission for mental disorders and season. seasonalitysummertemperaturemental disordersdaily hospital admissions ==== Body Introduction The World Health Organization (WHO) has recognised mental and neurological disorders, including Alzheimer's disease, as important diseases contributing essentially to the global non-communicable burden of disease and disability (1). In 2004, mental disorders accounted for 13% of the global disease burden. The proportion of patients living with disability from mental disorders accounted for 25.3 and 33.5% in low- and middle-income populations, respectively (1). More than 450 million people suffer from mental disorders, and among those depression, dementia, epilepsy, and schizophrenia are amounting to 350, 35.6, 50, and 24 million people, respectively. According to WHO, major depressive disorder may become the principal cause of disability worldwide by 2030 (2). The global burden of disease study from 2010 found that mental, neurological, and substance consumption disorders contributed to 10.4% of global disability-adjusted life years (DALYs) as well as 28.5% of global years lived with disability (YLDs) (3). Mental disorders accounted for the highest proportion of DALYs (56.7%) compared with neurological disorders (28.6%) and substance use disorders (14.7%) (3). Moreover, mental and substance consumption disorders became the leading cause of YLDs all over the world. There were 40.5, 14.6, 10.9, 9.6, and 7.4% of DALYs for depressive disorders, anxiety disorders, drug use disorders, alcohol abuse, and schizophrenia, respectively (3, 4). A study carried out in 17 nations on lifetime prevalence of mental disorders showed that there were risks among Asian populations, ranging from 13.2% in China and 14.4% in Iran to 18.0% in Japan (5). No official studies at the national level on mental disorders have been conducted in Vietnam. However, a representative epidemiological survey in Vietnam covering 10 common mental disorders in the population between 2001 and 2003 indicated a prevalence of 14.9%, of which alcohol abuse, depression, and anxiety accounted for 5.3%, 2.8%, and 2.6%, respectively (6). Many studies have described a relationship between seasonality and depressive disorders, particularly in countries with temperate climate characteristics (7–9). Most of the previous research on depressive disorders and seasonality has shown evidence of a spring/summer peak incidence for manic episodes (10, 11). For example, hospital entries for manic episodes in England revealed a summer peak. Research in Australia and New Zealand also indicated a similar peak, with the highest rate for manic episodes in the warm spring/summer season (from December to February) (10, 12–14). Several studies have identified more cases of patients with schizophrenia in the summer. Mechanistically, high ambient temperature has been suggested to contribute to more frequent onset of psychotic exacerbation (15). Research in the Northern Hemisphere including Ireland, England, and Scotland has shown a seasonal influence on first admissions for schizophrenia, and the peak time of the first hospital entries for mental disorders occurs in the middle of the year (15, 16). Research that was conducted in Israel showed that there was a significant relationship between higher ward temperature and more severe symptoms among inpatients with schizophrenia (16). Ambient temperature has long been suspected of being a cause of psychotic exacerbation in mental and behavioural disorders (17). Several studies from both developed and developing nations have indicated an increase in hospital admissions and emergency room visits during periods of a rise in temperature for individuals with psychological problems (9, 16, 18, 19). As we can see, there is evidence of a relationship between specific mental disorders, seasonality, and temperature variation with an increased trend of hospital admissions for mental disorders. However, most research was conducted in developed populations where weather characteristics, socio-economic conditions, and culture differ from those in Vietnam. It is therefore important to examine the association between variations in mental disease admission and season in Hanoi, Vietnam. The study aim The aim of this study has been to examine the relationship between seasons, mean temperature, and daily hospital admissions for mental disorders in Hanoi, Vietnam. Methods Hospitalisation data One of the two big mental hospitals taking care of mental patients, Hanoi Mental Hospital has captured both emergency visits and outpatients from Hanoi City and neighbouring provinces. In this study, we analysed a database from the hospital that covered mainly emergency cases for mental disorders in Hanoi City. All hospital admissions for mental diseases in Hanoi, in northern Vietnam, were collected from Hanoi Mental Hospital over a 5-year period from 2008 to 2012. The hospital admissions were diagnosed by mental health specialists from the hospital using the International Classification of Diseases 10 (ICD 10). Each patient was assigned a principal diagnostic ICD-10 code. ICD 10 classifies mental disorders as codes F00–F99. In this study, mental disorders were classified into eight groups of psychiatric disorders comprising Organic, including symptomatic, mental disorders (F0–9); Mental and behavioural disorders due to psychoactive substance use (F10–19); Schizophrenia, schizotypal, and delusional disorders (F20–29); Mood (affective) disorders (F30–39); Neurotic, stress-related, and somatoform disorders (F40–48); Behavioural syndromes associated with physiological disturbances and physical factors (F50–59); Mental retardation (F70–79); and Unspecified mental disorder (F99). Meteorological data Located at 21°2′0′′N 105°51′00′′E, Hanoi has a typical subtropical weather with four defined seasons. Hanoi's area of 3,329 km2 includes 13 districts (referred to as the urban area) and 17 suburbs (referred to as the rural area). Hanoi, the capital of Vietnam, has a population of around 7 million people living in a typical subtropical climate with four seasons. Seasons are defined as follows: spring (March–May), summer (June–August), autumn (September–November), and winter (December–February). In recent decades, the mean daily temperatures have steadily increased, affected by climate change and urbanisation (20). Meteorological data including daily maximum, minimum, and mean temperatures were collected from several monitoring stations in the Hanoi region, including the same urban and rural areas for the period 2008–2012. Data from the different monitoring stations were summed to estimate an average of a city mean value. The daily mean values of ambient temperature were used to estimate a daily increase in mental disorder admissions associated during the 5 years. A map of Hanoi, Vietnam Statistical analysis A negative binomial regression model for time series was applied to estimate the relationship between daily hospital admissions as a function of season and month, for which the winter season and January were used as annual reference times. We adjusted for year in the analysis to remove confounding of time trends. Moreover, we also established negative binomial time-series regression models of the aggregated counts of daily admissions as an outcome variable and indicator variables for daily temperature variations, the day of the week, season, and the time trend in the study period as explanatory variables. An electronic data set with 5 years (2008–2012) from Hanoi Mental Hospital including identified code, the date and time of admission, the age for both children and adults, gender, diagnosis of mental disorders, and the treatment for mental disorders were anonymised and identified prior to analysis. In addition, the daily hospital admissions for psychiatric disorders were analysed by stratifying by age, sex, and geographic location. Estimates were generated as relative risks (RRs) with 95% confidence intervals. Ethics statement Electronic health data sets including the code, age, sex, date, and treatment of admissions for mental disorders, both children and adults, were anonymised and de-identified prior to analysis. All procedures were approved by Hanoi Medical University and the Mental Hospital Ethics Board in Vietnam. Results Weather variation During the 5-year period (2008–2012), the seasonal temperature reached a high of around 30°C in summer (June–August) and a low of 15–17°C in winter (December–February). The average daily mean temperature was 24.3°C with a standard deviation of±5.6°C. In addition, the average of the daily maximum temperatures was 28°C, with a daily maximum of 40.4°C. The daily maximum temperatures in the winter and summer seasons were around 18–21°C and 33°C, respectively. In Hanoi, during the same study period, the month of June had the hottest weather with an average of daily maximum and mean temperatures of around 30°C±2°C and 34.4°C±2.4°C (standard deviations), respectively (Fig. 1). Fig. 1 The average of monthly mean temperature and maximum temperature during the 5-year period 2008–2012. Max. temperature (°C): Mean =28, SD=± 6.3, Min=8 and Max=40.4; Mean temperature (°C): Mean=24.3, SD=± 5.6, Min =7.4 and Max=34.5. Hospital admissions for mental disorders Throughout the study period (2008–2012), 23,525 subjects were admitted to the hospital with psychiatric problems, including both inpatients and outpatients (this refers to the first time they were registered at Hanoi Mental Hospital; see Fig. 2). During this period, the numbers of male and female patients with mental disorder admissions were 17,546 (74.6%) and 5,979 (25.4%), respectively; the numbers of admissions of patients from urban and rural areas were 13,486 (57.3%) and 10,017 (42.6%), respectively, whereas the age distribution in groups 0–17, 18–40, 41–60, and over 60 years was 734 (3.1%), 11,645 (49.5%), 9,380 (39.9%), and 1,766 (7.5%), respectively (Fig. 3). Fig. 2 The number of cases for specific mental disorders during 5 years 2008–2012. F0–9: Organic, including symptomatic, mental disorders; F10–19: Mental and behavioural disorders due to psychoactive substance use; F20–29: Schizophrenia, schizotypal, and delusional disorders; F30–39: Mood (affective) disorders; F40–48: Neurotic, stress-related, and somatoform disorders; F50–59: Behavioural syndromes associated with physiological disturbances and physical factors; F70–79: Mental retardation; F99: Unspecified mental disorder. Fig. 3 The characteristics of study population. The relationship between seasonality and mental disorders For Mental and behavioural disorders (ICD-10 codes F00–99), during the study period, the number of cases of mental disorders had a seasonal high in the summer season (RR=1.24, CI=1.1–1.39) with a peak in June (RR=1.46, CI=1.19–1.7; Figs. 4 and 5). The annual seasonal low occurred in the winter season (the reference season). The RRs of the mental disorders estimated an increase in the autumn season (RR=1.20, CI=1.08–1.36) and the spring season (RR=1.22, CI=1.08–1.37; Fig. 4). In general, all groups of hospital admissions for mental disorders showed higher risks of admission during the hotter months from May to August (Fig. 5). Fig. 4 The relationship between daily hospital admissions for mental disorders and seasonality. F0–9: Organic, including symptomatic, mental disorders; F10–19: Mental and behavioural disorders due to psychoactive substance use; F20–29: Schizophrenia, schizotypal, and delusional disorders; F30–39: Mood (affective) disorders; F40–48: Neurotic, stress-related, and somatoform disorders; F50–59: Behavioural syndromes associated with physiological disturbances and physical factors; F70–79: Mental retardation; F99: Unspecified mental disorder. Fig. 5 The relation between daily hospital admissions for mental disorders and monthly trends. For Organic, including symptomatic, mental disorders (ICD-10 codes F00–09), the number of patients admitted estimated to 1,530 (6.5%), with higher rates in the summer season (RR=1.41, CI=1.14–1.75) than in the spring season (RR=1.17, CI=0.94–1.45) and autumn season (RR=1.29, CI=1.04–1.6; Figs. 3 and 4). The annual peak of cases occurred in June (RR=2.24, CI=1.53–3.29; Fig. 5). For Mental and behavioural disorders due to psychoactive substance use (ICD-10 codes F10–19), the number of cases coming under F10–19 accounted for 3,963 (16.8%) admissions (Fig. 3). Hospital visits peaked in the spring season (RR=1.42, CI=1.23–1.64) with slightly lower risks in summer (RR=1.31, CI=1.13–1.51) and autumn (RR=1.33, CI=1.15–1.53; Fig. 4). The greatest risk increase was seen in May (RR=1.66, CI=1.3–2.13; Fig. 5). For Schizophrenia, schizotypal, and delusional disorders (ICD-10 codes F20–29), the highest proportion of cases among all diagnoses occurred and amounted to 11,381 (48.4%) patients (Fig. 3). However, this group had a less strong seasonal difference in admissions over the year. Nevertheless, admissions increased in the summer (RR=1.15, CI=1.02–1.3), autumn (RR=1.15, CI=1.02–1.3), and spring (RR=1.15, CI=1.01–1.29) compared to admissions in the winter (Fig. 4). The peak of cases was seen in May and June with RRs of 1.31 (CI=1.06–1.62) and 1.29 (CI=1.04–1.59) respectively (Fig. 5). For Mood (affective) disorders (ICD-10 codes F30–39), the number of 620 hospital visits peaked in the autumn (RR=1.33, CI=1.04–1.7) and was slightly higher in the summer (RR=1.08, CI=0.84–1.4) and spring (RR=1.17, CI=0.91–1.51) than in the winter (Fig. 4). Admissions peaked sharply in May and November with RRs of 2.05 (CI=1.29–3.26) and 2.05 (CI=1.29–3.27), respectively (Fig. 5). Neurotic, stress-related, and somatoform disorders (ICD-10 codes F40–48) include anxiety disorders, panic disorder, agoraphobia, obsessive compulsive disorder, and posttraumatic stress disorder. The total number of hospitalisations was 508 (2.2%) during the study period, with a peak in the summer season (RR=1.67, CI=1.25–2.25) and with slightly lower increases in the autumn season (RR=1.51, CI=1.12–2.04) and spring season (RR=1.37, CI=1.01–1.85; Fig. 4). From May to August, the number of admissions posed the highest risk with RRs from 2.11 to 2.47, especially in June (RR=2.47, CI=1.45–4.22; Fig. 5). For Behavioural syndromes associated with physiological disturbances and physical factors (ICD-10 codes F50–59), cases peaked in the spring season (RR=1.07, CI=0.86–1.33) and in August (RR=1.23, CI=0.85–1.77) (see Fig. 5). This category involves eating and sleeping disorders and amounted to 1,002 (4.3%) admissions (Fig. 3). For Mental retardation (ICD-10 codes F70–79), admissions encompassing mild, moderate, and severe as well as profound intellectual disabilities accounted for 2,439 (10.4%) cases (Fig. 3). There were differences among three of the seasons, where the number of cases in the summer (RR=1.36, CI=1.03–1.8) was higher than that in the autumn (RR=1.09, CI=0.82–1.44) and spring (RR=1.23, CI=0.93–1.63) relative to that in the winter season (Fig. 4). Admissions peaked in June and August with RRs of 1.79 (CI=1.11–2.9) and 1.76 (CI=1.1–2.83), respectively (Fig. 5). For Unspecified mental disorder (ICD-10 codes F99), there were estimated to be 2,082 (8.8%) admissions in this category (Fig. 3). The highest RR for admissions was found in the autumn season (RR=1.53, CI=1.3–1.8), whereas the RRs in the summer and spring seasons were 1.42 (CI=1.2–1.68) and 1.27 (CI=1.07–1.5), respectively (Fig. 4). The peak of admissions for unspecific mental disorders reached an annual high in October (RR=2.92, CI=2.14–3.97; Fig. 5). The relationship between seasonality and mental disorders stratified by sex, location, and age In the summer season, hospital admissions for mental disorders among males (RR=1.26, CI=1.12–1.41) were more common than for those among females (RR=1.18, CI=1.03–1.36) compared to the winter season. A peak in cases in both men and women was observed in early summer with RRs of 1.49 (CI=1.22–1.82) among men and 1.37 (CI=1.08–1.74) among women (Figs. 6 and 7) compared to January. Fig. 6 The association between daily hospital admissions for mental disorders and seasonality stratified by sex, location, and age. Fig. 7 The associations between daily hospital admissions for mental disorders stratified by sex, location and age groups. Admissions among residents in urban and rural areas increased in the summer season (RR=1.28, CI=1.14–1.44) and spring season (RR=1.21, CI=1.06–1.38) compared to the winter season (Fig. 6). Moreover, the peaks in cases in urban and rural areas occurred in June with an RR of 1.47 (CI=1.2–1.8) and in May with an RR of 1.49 (CI=1.19–1.87) respectively compared to January (Fig. 7). In this study, those aged 0–17 years accounted for the highest risk of mental disorders in the summertime (RR=1.29, CI=0.92–1.81) (Fig. 6). A peak in cases was seen in August with an RR of 1.95 (CI=1.09–3.47). Among the elderly, the number of admissions for mental disorders was also high in the summer season and especially in July with RRs of 1.23 (CI=1.03–1.48) and 1.75 (CI=1.26–2.42), respectively. In addition, the group aged 18–40 years had the highest risk of mental disorders in the summer (RR=1.26, CI=1.11–1.41) and in May (RR=1.51, CI=1.23–1.86). The highest rates among patients aged 41–60 years (RR= 1.23, CI= 1.09–1.39) were reported in the spring season (RR= 1.45, CI= 1.16–1.7), especially in April (Fig. 7). The relationship between temperatures and mental disorders In this study, when minimum, mean, and maximum temperatures increased 1°C, the number of cases for mental disorders increased by 1.7%, 2%, and 2.1%, respectively (Table 1). Table 1 The relationship between temperatures and daily hospital admissions for mental disorders during 5 years (2008–2012) Admissions Relative risk Confidence interval (95%) Max. temperature 1.017 1.004–1.03 Mean temperature 1.021 1.005–1.04 Min. temperature 1.020 1.003–1.04 Among patients with mood disorders (F30–39), it increased 5% (RR =1.05, CI=1.01–1.09; Fig. 8). The group of the elderly had the highest risk of mental disorders when mean temperature increased one unit (1°C) with RR of 1.04 (1.01–1.06) (Fig. 8). Fig. 8 The relationship between daily hospital admissions for mental disorders and the variation in ambient mean temperature. Discussion We found a correlation between hospital admissions for mental disorders and season in which the number of hospital visits for mental disorders increased by 24% in the summer season with a peak in these admissions in June. Compared to other demographic groups, males and the elderly (aged over 60 years) were more sensitive to seasonal risk changes. The RR of hospital admissions among men increased by 26% and among the elderly by 23% in the summer season. The study results show that the admission rates for males were more than three times higher than those for females. According to the literature in mental health, admission rates for psychiatric problems among males are always higher than those among females, except in the case of depression disorders (21). In this study, patients with mental disorders living in rural regions were similar to those living in urban regions. There were differences in hospital entries in between age groups whereby the group aged 18–40 years accounted for the highest rate. Moreover, the greatest proportion of mental disorders accounted for the number of cases of schizophrenia (37.17%), in contrast to a WHO report which states that the rate of depressive disorders is usually higher than that of schizophrenia (1, 2). In this study, mood disorders (F30–39) and unspecified mental disorders (F99) had the highest risks in the autumn season. The results were similar to the findings in research on depressive disorders and seasonality, especially in countries with a temperate climate (22, 23). However, when taking into account monthly trends in mental disorders, admissions of mood disorders (F30–39) and unspecified mental disorders (F99) showed quite high risks in May and June. Studies in Brazil, India, and Egypt showed a high number of cases of mood disorders in the hotter months (15, 24). The results of this study indicated, however, that there was a peak in admissions for mental disorders in the summer season with its fairly extreme weather pattern of high temperatures. The groups of mental disorders including psychoactive substance use (F0–9), schizophrenia (F20–29), somatoform disorders (F40–48), and mental retardation (F70–79) had the highest risks during the summer season. Moreover, our data demonstrated positive associations between temperature and admissions for mental disorders both for all and for specific mental disorders in the month of June when the weather is at its hottest. This increase in summertime may explain the increase in the number of cases of mental disorders and vice versa. Research worldwide has studied the relationship between ambient temperature and schizophrenia or depressive disorders, whereas other specific mental diseases such as dementia and drug and substance abuse have not been studied enough (19, 15, 25, 26). A relationship was reported between environmental temperature and schizophrenia morbidity, where the mean maximum monthly environmental temperature was associated with the monthly intake of inpatients suffering from schizophrenia especially during the summer season (16, 25). Similar findings have been mentioned in other research conducted in both developed and developing countries (15, 19, 27, 28). Lane K et al. reported from the United States that there was a statistically significant difference in the relationship between ambient temperature and visits to a psychiatric emergency department for psychiatric problems (27). In addition, Hansen et al. in Australia indicated that admissions during 1993–2006 for behavioural and mental disorders were correlated with a maximum daily temperature above a threshold of 26.7°C (18). Similar outcomes have been found in other countries such as France, Spain, India, and Israel, whereby France had a high rate of visits to a psychiatric emergency department during the period of heatwaves in 2003 (25, 29–31). Several studies conducted in Israel showed that there was a higher risk of admissions for schizophrenia when this was associated with mean maximal monthly temperature (16, 25). Similarly, scientists in India reported that there was a greater use of psychiatric services by patients with mood disorders in the summer when temperatures were at their hottest (9). The impact of daily climatic variables on psychosis admissions to mental hospitals as reported by the Irish Health Research Board revealed that a weak association existed between temperature and psychosis cases (28). Furthermore, there were found to be acute effects of extreme temperature exposure on emergency room admissions for mental and behavioural disorders in Toronto, Canada, where there was a strong relationship between these at 28°C, especially within a period of 0–4 days of exposure to hot weather. An estimate of cases showed that there was an increase of 29% over a cumulative period of 7 days after exposure to mean high ambient temperature in the 99th compared with the 50th (19). In our study, there was a significant difference in the gender distribution, in which the number of hospital admissions for males was greater than for females, especially in hot weather summer and spring seasons. Hospital visits for mental disorders among men were threefold compared to those among women. Previous studies of mental diseases indicated that the proportion of male patients with mental disorders was larger than the proportion of female patients, except for depressive disorders (21). In Vietnam, there have been high proportions of alcohol consumption and brain damage due to traffic accidents among men. Thus, this may contribute to the increased risk of mental disorders for male patients, especially for the groups with alcohol abuse, and delirium (32, 33). However, further studies need to be conducted in the future to ascertain the gender difference in the association between mental disorders and heat/heatwave exposure. Our data also confirmed that there were higher rates of cases in rural populations in May and June. This result corresponded to findings from Australia and Taiwan because an increased impact of hot weather on rural mental health had been observed (34, 35). It is known that high vulnerability in rural communities can be predicted during variations or extremes in weather (36). In addition, the number of cases of mental disorders among residents in urban regions also increased in the summertime, with a peak in June. This may be explained by a rise in temperatures and the problems of urbanisation (20). Regarding the association between admissions and seasonality and month trends between age groups, the group aged 0–17 years had peaks of mental disorders, especially in summertime and August. There were increased trends in daily hospital admissions for psychiatric disorders among other age groups such as those aged 18–40 and 41–60 years, where the number of cases rose in hot weather including the spring and summer seasons and especially in April, May, and June. In this study, patients over 60 years of age peaked in the summer period and July. The elderly had a higher risk of mental disorders than did others on the basis of monthly trends from February to December, and especially in June and July. When the average of temperatures including minimum, mean, and maximum increased one unit (1°C), the number of cases for mental disorders rose between 1.7 and 2.1% in which patients with depressive disorders accounted for the highest risk with an increase by 5%. Findings from Korea, Taiwan, Australia, Brazil, and Egypt indicated that admissions for depressive disorders increased with rising ambient temperature (18, 15, 24, 26, 34). Thus, emerging admissions beyond elevated temperature may explain the peak of daily hospital visits in hottest periods including June and July. Furthermore, when mean temperature increased one unit (1°C), the number of cases for mental disorders in the population aged over 60 years rose by 4%. The results were similar to the outcome of a study conducted in Australia on heat waves and mental disorders, in which the elderly population (65–74 years of age) had the highest risk of suffering mental problems (18). Psychological responses including thermal sensation and human behaviour are greatly affected by thermal environment (37). In spite of the prime importance attached to them, psychological studies on the impact of thermal environment are still at their infancy (18, 38). According to Hensel (1981), thermo-reception resulting in qualities such as warmth is qualitative, originating as it does from sensory experience, and it therefore cannot be based on physics or physiology (37). In addition, another approach taken by research has been to study the association between environmental conditions, physiological responses, and psychological phenomena such as behaviour and sensation (11, 18, 37). Elevated temperature can exacerbate mental and behavioural disorders for a variety of physiological and psychological reasons in which vulnerable neurotransmitters and thermoregulation play in part a key role (36, 37, 39, 40). This has been demonstrated in animal models of impaired dopaminergic transmission in schizophrenia (19, 21). Moreover, many findings on homeostasis maintain that while the body was in the state with mental disorders, its neurotransmitters, including serotonin, dopamine, and norepinephrine, underwent change when environmental change was influenced by ambient temperature (40, 41). Several experiments in mice and rats have proven the impact of ambient temperature on drug-induced dopamine and serotonin neurotoxicity (42). Some medications affecting neurotransmitters may influence thermoregulation, and some can trigger malignant hyperthermia, neuroleptic syndrome, or serotonin syndrome (21, 43). Physiological mechanism-related factors associated with high temperature, brain temperature, and mental-psychological disorders may shed light on the hypothesis in the study of whether the summer season with its high temperatures may aggregate hospital admissions for mental disorders. However, further research on biological, psychological, and physiological areas is required to find out the exact mechanisms associated with heat-related mental illnesses. Climate change has become a big challenge worldwide (44). The effects of rising temperatures on human health, both physical and mental, come through direct and indirect pathways (36, 45). There has been a lack of studies on mental health consequences, but recent studies have shown a high risk of heat-related mortality and morbidity with an increasingly heat-associated mental illness (18, 19, 45–47). Vietnam situated in the subtropics is suffering annual droughts and heat waves. It is one of the nations with a high potential of being most negatively impacted by climate change with temperatures rising by 0.5–0.7°C per 50 years, from the South to the North (48). Moreover, the mean temperatures in all regions in Vietnam are projected to increase by 1.3–1.6°C, 1.6–1.9°C, and 1.9–2.2°C in 2050, 2070, and 2100, respectively, compared with the period 1980–1999 (48). The trend of increased temperatures in Vietnam may most likely increase the risk for heat-related mortality and morbidity in the future, both physical and mental. This study illustrated that the peak of admissions for mental disorders was associated with the warmest season in Vietnam. Thus, further studies on the impact of extreme heat and heat waves on mental disorders and vulnerable populations in Vietnam are very important and considered necessary to conduct. Conclusion There was an association between hospital admissions for mental disorders and seasonality with peaks of cases occurring in the hot weather of the summer season, especially in June. There were gender and age group differences where a significantly higher number of men and elderly patients (>60 years of age) were admitted for mental disorders in times of hot weather. In the planning of mental health care, the results can be used as an indication of expected peaks of mental ill health in relation to seasons and temperatures, and as a guide for resource allocation. It also gives clinical doctors of mental health in Vietnam more information regarding mental diseases and their associations with seasons and hot weather. By having this understanding they can be better prepared. The study results are formative and may pave the way for future research in the region. Further studies will be needed in the future to explore the impacts of prolonged heat waves on mental-psychological disorders, especially in populations at risk. It may help health managers to be better prepared, by setting up good strategies to prevent heat-related mental illnesses and protect vulnerable populations, at risk of mental health problems. Limitations of the study The 5-year database from one psychiatric hospital in northern Vietnam was used to estimate the relationship between seasonality and admissions of mental cases. This may not capture fully all mental disorder patients in Hanoi, especially outpatients. Moreover, there are different regions in Vietnam which have distinctly specific weather patterns; therefore, it is impossible to generalise the results to the national level. This is a formative study. More studies are needed in the future using more information to build a model by taking into account the environment, personal traits, and socio-economics together with mental health problems in order to identify vulnerable populations. Acknowledgements We thank the Umeå Centre for Global Health Research within the unit of Epidemiology and Global Health at Umeå University and the Swedish International Development Cooperation Agency (SIDA) for supporting the project Public Health Preparedness and Response to Critical Global Health Issues in Vietnam and Sweden, of which this study is a part. We also thank the Hanoi Mental Hospital for having supplied the data on hospital admissions. Authors’ contributions Conceived and designed the experiments: PMT, JR, KBG, MN. Performed the experiments: PMT, JR, MN. Analysed the data: PMT, JR. Contributed reagents/materials/ analysis tools: KBG. Wrote the manuscript: PMT, JR, MN. Conflict of interest and funding The authors declare no conflicts of interest and have not received any funding or benefits from industry to conduct this study. Paper context Several studies on seasonality and mood disorders, and schizophrenia have been conducted worldwide, especially in the developed countries. Vietnam, a low-and middle-income nation, has increasingly suffered consequences of extreme heat/heatwaves. Thus, research studies on relationships between mental disorders and seasonality, heat exposure, as well as on limitations of its consequences, are very important. It is difficult to avoid natural disasters, but we can prevent heat-related mental illnesses, particularly for poor people who need real help and protection. ==== Refs References 1 WHO Library Cataloguing-in-Publication Data Plan action 2013–2030 8 Available from: http://www.who.int/mental_health/en/ [cited 11 March 2015]. 2 Marcus M Yasamy MT van Ommeren M Chisholm D Saxena S Depression – a global public health concern WHO Department of Mental Health and Substance Abuse 2012 Geneva, Switzerland World Federation for Mental Health 14 18 Available from: www.who.int/mental_health/en/ [cited 11 March 2015]. 3 Whiteford HA Ferrari AJ Degenhardt L Feigin V Vos T The global burden of mental, neurological and substance use disorders: an analysis from the Global Burden of Disease Study 2010 PLoS One 2015 10 e0116820 25658103 4 Whiteford HA Degenhardt L Rehm J Baxter AJ Ferrari AJ Erskine HE Global burden of disease attributable to mental and substance use disorders: findings from the Global Burden of Disease Study 2010 Lancet 2013 382 1575 86 23993280 5 Kessler RC Angermeyer M Anthony JC De Graaf RON Demyttenaere K Gasquet I Lifetime prevalence and age-of-onset distributions of mental disorders in the World Health Organization's World Mental Health Survey Initiative World Psychiatry 2007 6 168 76 18188442 6 Vuong DA Van Ginneken E Morris J Ha ST Busse R Mental health in Vietnam: burden of disease and availability of services Asian J Psychiatr 2011 4 65 70 23050918 7 Mersch PP Middendorp HM Bouhuys AL Beersma DG van den Hoofdakker RH Seasonal affective disorder and latitude: a review of the literature J Affect Disord 1999 53 35 48 10363665 8 Ohayon MM Hong S-C Prevalence of major depressive disorder in the general population of South Korea J Psychiatr Res 2006 40 30 6 15878179 9 Avasthi A Sharma A Gupta N Kulhara P Varma VK Malhotra S Seasonality and affective disorders: a report from North India J Affect Disord 2001 64 145 54 11313081 10 Morken G Lilleeng S Linaker OM Seasonal variation in suicides and in admissions to hospital for mania and depression J Affect Disord 2002 69 39 45 12103450 11 Shapira A Shiloh R Potchter O Hermesh H Popper M Weizman A Admission rates of bipolar depressed patients increase during spring/summer and correlate with maximal environmental temperature Bipolar Disord 2004 6 90 3 14996147 12 Carney PA Fitzgerald CT Monaghan CE Influence of climate on the prevalence of mania Br J Psychiatry 1988 152 820 3 2901891 13 Mulder RT Cosgriff JP Smith AM Joyce PR Seasonality of mania in New Zealand The Aust N Z J Psychiatry 1990 24 187 90 2396962 14 Jones I Hornsby H Hay D Seasonality of mania: a Tasmanian study Aust N Z J Psychiatry 1995 29 449 53 8573048 15 Amr M Volpe FM Seasonal influences on admissions for mood disorders and schizophrenia in a teaching psychiatric hospital in Egypt J Affect Disord 2012 137 56 60 22244374 16 Shiloh R Weizman A Stryjer R Kahan N Waitman DA Altered thermoregulation in ambulatory schizophrenia patients: a naturalistic study World J Biol Psychiatry 2009 10 163 70 19514098 17 McMichael AJ Woodruff RE Hales S Climate change and human health: present and future risks Lancet 2006 367 859 69 16530580 18 Hansen A Bi P Nitschke M Ryan P Pisaniello D Tucker G The effect of heat waves on mental health in a temperate Australian city Environ Health Perspect 2008 116 1369 75 18941580 19 Wang X Lavigne E Ouellette-kuntz H Chen BE Acute impacts of extreme temperature exposure on emergency room admissions related to mental and behavior disorders in Toronto, Canada J Affect Disord 2014 155 154 61 24332428 20 Pham T Tong T Pham V Murgante B Misra S Carlini M Torre C Nguyen H-Q Taniar D Becoming urban: how urbanization influences the loss of Arable Land in Peri-urban Hanoi Computational science and its applications – ICCSA 2013. Lecture Notes in Computer Science. 7974 2013 Berlin Springer 238 52 21 Sadock BJ Sadock VA Kaplan & Sadocks synopsis of psychiatry – Chapter 3 2007 10th ed The United State Lippincott Williams & Wilkins 301 326 Neurophysiology and Neurochemistry 22 Silverstone T Romans S Hunt N McPherson H Is there a seasonal pattern of relapse in bipolar affective disorders? A dual northern and southern hemisphere cohort study Br J Psychiatry 1995 167 58 60 7551610 23 Sato T Bottlender R Sievers M Moller HJ Distinct seasonality of depressive episodes differentiates unipolar depressive patients with and without depressive mixed states J Affect Disord 2006 90 1 5 16325920 24 Volpe FM Del Porto JA Seasonality of admissions for mania in a psychiatric hospital of Belo Horizonte, Brazil J Affect Disord 2006 94 243 8 16777234 25 Shiloh R Shapira A Potchter O Hermesh H Popper M Weizman A Effects of climate on admission rates of schizophrenia patients to psychiatric hospitals Eur Psychiatry 2005 20 61 4 15642446 26 Lee HJ Kim L Joe SH Suh KY Effects of season and climate on the first manic episode of bipolar affective disorder in Korea Psychiatry Res 2002 113 151 9 12467954 27 Lane K Wheeler K Charles-Guzman K Ahmed M Blum M Gregory K Extreme heat awareness and protective behaviors in New York City J Urban Health 2014 91 403 14 24297476 28 McWilliams S Kinsella A O'Callaghan E The effects of daily weather variables on psychosis admissions to psychiatric hospitals Int J Biometeorol 2013 57 497 508 22855350 29 Ebi KL Lewis ND Corvalan C Climate variability and change and their potential health effects in small Island States: information for adaptation planning in the health sector Environ Health Perspect 2006 114 1957 63 17185291 30 Oberlin M Tubery M Cances-Lauwers V Ecoiffier M Lauque D Heat-related illnesses during the 2003 heat wave in an emergency service Emerg Med J 2010 27 297 9 20385684 31 Bulbena A Sperry L Garcia Rivera C Merino A Mateu G Torrens M Impact of the summer 2003 heat wave on the activity of two psychiatric emergency departments Actas Esp Psiquiatr 2009 37 158 65 19533429 32 World Health Organization 2014's report. Vietnam - traffic accidents 1 Available from: http://www.who.int/substance_abuse/publications/global_alcohol_report/profiles/vnm 33 World Health Organization 2014's report. Vietnam - traffic accidents 1 Available from: http://www.who.int/violence_injuries_prevention/road_safety_status/2013/country_profiles/vietnam 34 Lee HC Tsai SY Lin HC Seasonal variations in bipolar disorder admissions and the association with climate: a population-based study J Affect Disord 2007 97 61 9 16890994 35 Horton G Hanna L Kelly B Drought, drying and climate change: emerging health issues for ageing Australians in rural areas Australas J Ageing 2010 29 2 7 20398079 36 Reser JP Part II: Specific population. Climate change and human well-being 119 162 37 Parsons KC Thermal environments the effect of hot moderate and cold environments on human health comfort and performance 2003 2nd ed London The Taylor & Francis 47 68 Chapter 3. Psychological responses 38 Gupta S Murray RM The relationship of environmental temperature to the incidence and outcome of schizophrenia Br J Psychiatry 1992 160 788 92 1617362 39 Lipton JM Clark WG Neurotransmitters in temperature control Ann Rev Physiol 1986 48 613 23 2871810 40 Meeusen R Roelands B Central fatigue and neurotransmitters, can thermoregulation be manipulated? Scand J Med Sci Sports 2010 20 Suppl 3 19 28 21029187 41 Roelands B Meeusen R Alterations in central fatigue by pharmacological manipulations of neurotransmitters in normal and high ambient temperature Sports Med 2010 40 229 46 20199121 42 Micheal E Thase. Chapter 9: Neurobiological Aspect of Depression. Handbook of depression 2009 New York The Guilford Press 187 217 43 Swim JK Stern PC Doherty TJ Clayton S Reser JP Weber EU Psychology's contributions to understanding and addressing global climate change Am Psychol 2011 66 241 50 21553950 44 McMichael AJ Ebi KL Mearns LO Nyenzi B Climate change and human health: risks and responses 1 43 WHO Library Cataloguing-in-Publication Data. Chapter 1 and 2 45 Smith KR Woodward A Campbell-Lendrum D Chadee DD Honda Y Liu Q Field CB Barros VR Dokken DJ Mach KJ Mastrandrea MD Bilir TE Human health: impacts, adaptation, and co-benefits Climate change 2014: impacts, adaptation, and vulnerability. Part A: Global and sectoral aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change Cambridge 2014 United Kingdom Cambridge University Press 709 54 46 Nitschke M Tucker GR Bi P Morbidity and mortality during heatwaves in metropolitan Adelaide Med J Aust 2007 187 662 5 18072911 47 Vaneckova P Bambrick H Cause-specific hospital admissions on hot days in Sydney, Australia PLoS One 2013 8 e55459 23408986 48 Van Tai Nguyen Ngoc Kim Ti Tuy Hung Phan Tuan Quyen Le Ti Le Anh Nguyen Ti Ngoc Vietnam assessment report on climate change (VARCC) (2009). Institute of Strategy and Policy on natural resources and environment, Viet Nam (ISPONRE), Chapter 3. Climate change impacts and strategy, options to cope with climate change in Viet Nam 2009 Hanoi 73 84

PMC005xxxxxx/PMC5002037.txt

==== Front Soc Indic ResSoc Indic ResSocial Indicators Research0303-8300Springer Netherlands Dordrecht 108310.1007/s11205-015-1083-6ArticleInteractions Between Policy Effects, Population Characteristics and the Tax-Benefit System: An Illustration Using Child Poverty and Child Related Policies in Romania and the Czech Republic Avram Silvia +44(0)1206874824savram@essex.ac.uk 1Militaru Eva 21 ISER, University of Essex, Wivenhoe Park, Colchester, CO43SQ UK 2 National Research Institute for Labour and Social Protection, 6-8 Povernei Street, Sector 1, 010643 Bucharest, Romania 5 5 2016 5 5 2016 2016 128 3 1365 1385 21 8 2015 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.We investigate the impact of the Romanian and Czech family policy systems on the poverty risk of families with children. We focus on separating out the effects of policy design itself and size of benefits from the interaction between policies and population characteristics. We find that interactions between population characteristics, the wider tax benefit system and child related policies are pervasive and large. Both population characteristics and the wider tax-benefit environment can dramatically alter the antipoverty effect of a given set of policies. Keywords Child povertyChild benefitsMicrosimulationPolicy interactionsPopulation interactionsissue-copyright-statement© Springer Science+Business Media Dordrecht 2016 ==== Body Introduction The past 20 years have witnessed prominent policy initiatives to tackle child poverty both at the European and national levels (for example, the Lisbon strategy or the Labour government pledge to halve child poverty in the UK by 2020). However, despite these efforts, child poverty rates have remained stubbornly high. Even more worryingly, they have increased in some countries especially in comparison with overall poverty rates (Oxley et al. 2000; Van Mechelen and Bradshaw 2013). For example, between 2005 and 2012, poverty among children in the 27 Member States has broadly remained stable around 28 % whereas poverty among the population as a whole fell from 26 to 25 % (EUROSTAT). A large body of scholarly work has linked poverty, and low income in general, to deleterious consequences on child developmental trajectories and educational attainment (Black et al. 2000; Engle and Black 2008; Najman et al. 2009; Petterson and Burke Albers 2001), health status (Aber et al. 1997; Case et al. 2002), as well as adulthood outcomes (Duncan et al. 1998, 2010). Given the consequences of material deprivation both on current well-being and future capability and the fact that children generally have little control over what economic resources are available to them, there is overwhelming agreement that child poverty is an area necessitating public intervention. To mitigate child poverty, governments can resort, among other tools, to various forms of income support and child contingent transfers. Previous scholarly work has found considerable evidence that child contingent transfers do have a substantial effect on child poverty outcomes, with typically large cross-national variation in policy effects (Matsaganis et al. 2007; Barrientos and DeJong 2006; Bradshaw 2006; Immervoll et al. 2000; Whiteford and Adema 2007). These studies usually use either pre-transfer post-transfer comparisons or a microsimulation-based approach and attribute any differences in observed poverty or inequality indicators to the policy package they investigate. One aspect left unaddressed in these studies is the extent to which policy effects are shaped by ‘outside’ factors, i.e. population characteristics and/or the wider tax-benefit system in which they operate. Although these studies generally acknowledge the existence of interactions of various sorts and their potential in shaping the impact of family transfers, they fail to explicitly investigate these issues. As a result, there is little evidence on the sensitivity of estimated policy effects to variation in the population profile and the design of other social and fiscal instruments that are present. For example, can these factors alter the ranking of policy instruments with similar objectives? These issues are all the more important as the European Union (EU) has launched various benchmarking exercises that essentially rely on comparisons between countries with potentially very different demographic, labour market and tax-benefit institutions. This paper seeks to bridge this gap and contribute to the understanding of the role of interactions between child contingent policies, population characteristics and the wider tax benefit system in shaping the impact of the former on child poverty. By interactions, we mean that the magnitude of the policy effect is itself contingent on other factors, in particular population characteristics and/or the architecture of the wider tax-benefit system. To this end, we take Romania and the Czech Republic as case studies and examine the reduction in child poverty effected by three family transfers and one tax concession (see Table 1). Romania is a country with high levels of child poverty where the support package available to families with children has been found to be not very effective (TARKI 2010). In contrast, the Czech Republic registers low overall and child poverty rates which have been found to be at least partly the result of generous income support (TARKI 2010). Using microsimulation techniques, we examine to what extent these results are driven by the characteristics of the child-related policy instruments themselves as opposed to being the product of the ‘fit’ between these instruments, other income support measures available to families with children and population features. More specifically, we compute the direct, first-order effect of both the Romanian and the Czech child policy package on relative poverty, while varying the underlying population characteristics and the wider tax-benefit system. Following Salanauskaite and Verbist (2013), we also distinguish between instrument generosity and instrument design in measuring the impact of a given child policy package. The rest of the paper proceeds as follows. Section 2 reviews the existing literature on the links between child related transfers and child poverty. Section 3 describes the Romanian and Czech policies we consider in this exercise. Section 4 describes the data and methods. The various counterfactual scenarios we simulate are explained in Sect. 5. Section 6 discusses our main results. Section 7 concludes.Table 1 Policy instruments included in the child package Policy Eligibility Amounts % Of children in families receivinga Average amount as  % of HH disposable incomeb Romania Allowance for new born children and the outfit for new born children Universal entitlement for all new-borns Lump sum of approx. 354 RON 6 2 Universal child benefit Age <18 or in high school Per month/200 RON for children under 2; 25 RON for children 2 and older 100 7 Means-tested family benefits Means-tested; monthly income <176 RON per person; children are persons <16 or <18 and with family income <50 RON/month Between 36 and 52 RON/month, depending on the number of children for 2 parent families and between 52 and 79 RON per month for single parent families 40 9 Tax allowance for dependent children All employed parents with employment income below 3000 RON/month; the tax allowance is only deductible against employment income; children are considered dependent if aged <16 or having an income below 250 RON/month Maximum 100 RON per child, max 400 RON/month. The tax allowance is reduced on a sliding scale between 1000 and 3000 RON per month; it reduces to zero once gross employment income reaches 3000 RON/month 69 2 Czech Republic Child allowance Means-tested; family income is <4 times the family minimum living standard level; children are individuals younger than 18 or younger than 26 and in education Between 16–36 % of the child’s minimum living standard (which depends on age), depending on family income 74 3 Social allowance Means-tested; income is <2.2 times the family minimum living standard; children are individuals younger than 18 or younger than 26 and in education Child’s minimum living standard from which a share may be deducted based on the size of family income relative the family’s minimum living standard level 27 6 Birth grant Universal entitlement for all new-borns Lump sum between 17,760 and 79,680 CZK, depending on number of children in the family 9 5 Refundable child tax credit Universal entitlement for all parents with dependent children; the tax credit is only refundable if employment income is larger or equal to 6 minimum wages/year 6000 KCZK/month per child, up to a maximum of 5 children 90 3 Children are considered to be individuals aged 17 or less; all policies refer to 2007 Source: Authors’ compilation based on EUROMOD G1.4 aPercentages calculated based on simulated entitlements in EUROMOD, not on actual reported receipt in SILC bFigures calculated based on households receiving only; in the case of tax concessions, figures are based on approximations not exact amounts Child Poverty and Public Transfers: A Review of the Literature There is a long literature trying to evaluate the role of social and fiscal policies on the welfare of families with children (Gornick and Jäntti 2010, 2011; Jäntti and Bradbury 2003; Barrientos and DeJong 2006; Figari et al. 2011; Whiteford and Adema 2007; Oxley et al. 2000; Sutherland and Piachaud 2011; Bradbury and Jäntti 2001). These studies usually compare poverty and inequality indicators based on market incomes alone with the same indicators derived based on disposable incomes and find that taxes and transfers play an important role in reducing poverty among families with children, although there is considerable cross-national variation in the extent of this reduction. For example, examining child poverty rates among high income countries, Gornick and Jäntti (2010, 2011) conclude that cross-national variation is explained not so much by demographic factors as by labour market institutions alongside the existing system of transfers. Similar exercises have been carried out using child related policies (Van Mechelen and Bradshaw 2013; Matsaganis et al. 2007; Förster and Tóth 2001; Immervoll et al. 2000; Bradshaw 2006). Generally, these studies find that transfers targeted at families with children significantly reduce both the prevalence and the depth of child poverty, albeit the size of the reduction varies substantially across countries. Studies directly looking at infant outcomes such as birth weight and neonatal mortality rates also find positive effects of income support availability to disadvantaged women and parents (Hoynes et al. 2011; Almond et al. 2011). Finally, the availability of income support has been found to positively affect not only outcomes measured during childhood but also long run outcomes such as health and economic self-sufficiency in adulthood (Hoynes et al. 2012). Although there is general consensus that directing resources to low income families with children is a good way to invest in the next generation, there is less agreement on what aspects make a policy more effective. Some authors stress the size of the transfer package (Notten and Gassman 2008; TARKI 2010). According to this view, it is mainly the generosity of the transfer system towards families with children that is likely to impact on child poverty rates. However, public child contingent support is rarely equally generous towards all families with children. Explicitly or implicitly, policy instruments are likely to favour families with some characteristics and not others (ex: number and age of children, number of adults/earners in the household, family income, tax-paying status etc.). Clearly, the effect of a given set of policies on child poverty depends to a large extent on the demographic and labour market characteristics of poor families. A different strand in the field has argued that in addition to size, policy design plays an important role in determining policy effectiveness (Salanauskaite and Verbist 2013; Levy et al. 2009; Immervoll et al. 2000). Generally, these studies have relied on cross-national comparisons, and/or microsimulation methods to measure the impact of policies, as well as to estimate the effect of alternative policy designs. Although providing important insights into the importance of policy design, these studies usually stop at concluding that one set of policies would likely have been more effective than another in a particular context. There is little potential to generalize what features of the design are likely to make a policy more effective than another. More importantly, they fail to consider the sensitivity of the results to the demographic and wider institutional context they have been derived from. Finally, a large body of the literature has focused on the role of targeting transfers in general and family benefits in particular in addressing poverty (Atkinson 1995; Jarvis and Micklewright 1995; de Neubourg et al. 2007; Förster and Tóth 2001). While some authors (Nelson 2004; Korpi and Palme 1998) have found evidence of a negative correlation between targeting and the overall budget available for public transfers (the famous redistribution paradox), it is not clear that this relationship holds when child related policies are concerned. On the contrary, countries that combined universal benefits with targeted support for low income families with children appeared to achieve superior poverty reduction (Van Mechelen and Bradshaw 2013). To sum up, extensive research in the area of child poverty consistently finds that public transfers can play an important role in shaping poverty outcomes for families with children. Nonetheless, we still understand relatively little about which aspects of transfer policies, beyond size, matter most and how these interact with demographic characteristics and the wider fiscal institutional context in which they operate. This paper begins to address this gap by examining the extent to which policy impacts are shaped by the characteristics of the population they apply to and the tax-benefit system within which they operate. We address two questions. First, we assess the variation in estimated policy effects when the context, i.e. population characteristics and/or the tax benefit system, changes. Can the same set of policies produce very different estimates when the context is altered? Second, we probe whether the ranking of policy instruments is context specific. For example, is it possible that one set of policies is more effective in the context of the Romanian population but a different set of policies is most effective in the context of the Czech population? Child Poverty and Child Support in Romania and the Czech Republic From a historical perspective, Romania and the Czech Republic share a number of similarities. Both countries have experienced during half a century a foreign-imposed regime based on a command economy combined with suppression of political and civil freedoms. During the nineties, both countries have undergone an extensive political and economic transition that ended with becoming full members of the European Union in 2004 and 2007 respectively. Despite these similarities, the two countries differ in a number of important respects. In particular, the Czech Republic is much richer with a GDP/capita in 2012 of approximately 20,700 PPP compared to Romania’s 13,500 PPP (EUROSTAT). It is also a country with less inequality as measured by the Gini coefficient (25 vs. 33, EUROSTAT). Most importantly, for our purposes, the two countries rank very differently on child poverty indicators (see Fig. 1). While in the Czech Repulic poverty rates for chidlren are relatively low in comparative perspective, Romania is one of the EU Member States with the highest prevalence of child poverty. Finally, the Czech Republic has slightly higher levels of taxation compared to Romania (40 % of GDP is collected in taxes in the Czech Republic versus 35 % in Romania) and spending on cash social transfers (12.5 % of GDP compared to 9.2 % of GDP in Romania) (EUROSTAT).1Fig. 1 Child poverty rates in the EU, 2007. Note Children are defined as aged 17 or less. Source EUROSTAT database, http://epp.eurostat.ec.europa.eu/portal/page/portal/statistics/search_database Since family benefits have been shown to be able to significantly influence poverty and inequality (see Sect. 2), the large discrepancy in child poverty outcomes may partly be explained by differences in child related public transfers. Obviously, children can be made better off through a variety of public measures benefiting their families, ranging from income transfers, to tax advantages and concession, to subsidies or in-kind provision of goods and services.2 In this paper however, we limit our attention to transfers and tax concessions directly linked to the presence of children. Using 2007 as our reference year,3 we isolate four4 policies in each of the two countries (three transfers and a tax concession) which we collectively term the child support package and which will form the focus of our analysis in the remainder of the paper. Table 1 provides an overview of the main policy elements. In 2007, Romania had three child related benefits. The first is a universal, flat-rate child allowance that covers all children younger than eighteen and pays the same amount irrespective of birth rank. There is however substantial age related variation. Children under two benefit from an increased allowance approximately eight times higher than the one available to older children. In fact, the level of the benefit for young children is unusually high, representing approx. 16 % of the average gross wage in 2007. On average, this benefit constitutes approximately 7 % of household disposable income for the families that receive it. Low income families with children may be entitled to a supplementary allowance. Entitlement is subject to passing an income test which is fairly stringent. The benefit amount depends on the number of children present in the household. However, the benefit increases less than proportionally for higher rank children and is capped after the fourth child. Lone parent families are subject to the same income test but are entitled to higher benefit rates. In total, approximately 40 % of children live in families receiving this type of transfer. For households that receive it, the means-tested family allowance represents approximately 9 % of household disposable income. Finally, in 2007, Romania also had a birth grant which was a lump sum payment to all new-borns equal to approximately 28 % of the average gross monthly wage (or 2 % of household disposable income for the families that receive it). In addition to these transfers, families with children also qualify for tax relief in the form of a tax allowance on employment income. The level of the tax allowance is relatively low and its value is further decreased by the low rate of the personal income tax. Receipt of the allowance is income tested and the amount decreases on a sliding scale depending on the value of taxable earnings. Although the tax allowance is available to all employed parents with low and moderately high employment incomes, only 69 % of children live in families that benefit from this allowance. For these families, the gain due to the presence of the tax allowance represents approximately 2 % of household disposable income. In the Czech Republic, the main child benefit is income —tested. Receipt is restricted to families with an income less than four times the family minimum living standard (MLS). Nonetheless, the income conditionality is largely designed to exclude wealthy families rather than restrict transfers to the poor. This aspect is confirmed by the fact that 74 % of children receive this benefit (see Table 1). The benefit amount depends on the child’s age (older children are entitled to increased payments) as well as on family income (families with lower incomes are entitled to more generous rates). Overall, for recipient households, the average benefit amount equals approximately 3 % of household disposable income. In addition to the main child benefit, low income families may be entitled to an additional income-tested transfer, called social allowance. As in the case of the main child benefit, entitlement and benefit amounts are calculated using the family and child MLS levels. However, eligibility is restricted to families with incomes below 2.2 times the family MLS and any family income reduces the value of the benefit. Only 27 % of children live in households receiving this benefit. On average, the benefit represents 6 % of disposable income for the households that receive it. Similarly to Romania, the Czech Republic has a lump-sum grant payable to all new-borns. However, unlike Romania, the benefit increases with higher order births. It is also more generous than in Romania amounting to, on average, 5 % of disposable income for the families that receive it. Lastly, families with children are entitled to a refundable child tax credit. It is refundable only to parents with sufficiently high employment income. The tax credit is the same for all children in the family, irrespective of age and birth rank. Almost all children (90 %) benefit from it with recipient households relying on it for approximately 3 % of their disposable income. To sum up, the two child support packages are relatively similar. Both feature some universalistic elements (the universal child benefit and the child tax allowance in Romania and the child tax credit in the Czech Republic) together with some means-tested components directed at families with few resources (the means-tested family benefit in Romania and the social allowance in the Czech Republic). Means-testing is somewhat more prevalent within the set of Czech policies. Yet, eligibility thresholds are high enough to allow a significant number of families with children to become entitled. Data and Methods To examine the interaction between child policies, the tax-benefit system and population characteristics, we make extensive use of microsimulation techniques to generate a number of counterfactual scenarios (see the next section for a detailed overview). We then compare child poverty measures under these difference scenarios. To carry out our simulations, we use EUROMOD,5 the-EU wide tax-benefit microsimulation model (Sutherland and Figari 2013). EUROMOD combines individual and household data from the EU-Survey of Income and Living Conditions (EU-SILC) with detailed information on social and fiscal national legislation to accurately simulate a wide range of transfer entitlements and tax liabilities at the micro-level. We use the Romanian and Czech components of EUROMOD to simulate all counterfactual scenarios. All our results refer to the policy year 2007 and use the 2008 EU-SILC as the underlying micro data. As SILC 2008 contains income information corresponding to the year 2007, there is no time discrepancy between our policy year and our data year. All the simulations assume full compliance with taxes and full take-up of benefits.6 As a result, simulation results refer to the intended rather than actual policy impacts. We define children as individuals aged 17 or less, irrespective of their educational or labour market status. Although children may be considered dependent (and thus entitled to child related transfers and tax concessions) up to much older ages in both countries (subject to additional conditions being staisfied), we have opted to circumvent potential incongruities in the way children are defined across countries and across policy instruments by restricting the age range. Given our interest lies mainly in the anti-poverty potential of child related transfers and tax concessions among families with children, we need to operationalize poverty. We adopt the current established practice and define poverty in a relative way, based on equivalised household disposable income. Disposable income is calculated as market income plus public transfers minus taxes and social insurance contributions. We use the ʽmodified OECD’7 equivalence scale to account for differences in household size as well as economies of scale in consumption. We assume income pooling across household members and attribute equivalised disposable income to each individual, including children.8 Poverty is operationalized as having an equivalised disposable income lower than 60 % of the median. To check the sensitivity of our results, we use a second, more stringent, threshold set at 40 % of median equivalised disposable income. We use the term severe poverty to denote poverty defined using the lower income threshold. In all cases, we measure the impact of child related policies on a set of three poverty indicators belonging to the Foster-Greer-Thorbecke (FGT) family (Foster et al. 1984). More specifically, we compute the relative reduction in the poverty rate (FGT0), gap (FGT1) and severity (FGT2). Finally, we conclude this section with two caveats. First, to keep the complexity of our analysis manageable, we abstract from any behavioural changes triggered by replacing one set of policies with another. From a policy perspective however, behavioural responses clearly cannot be ignored. Second, we do not consider the issue of policy administration costs. For example, it has long been acknowledged that administering targeted benefits is much more burdensome compared to administering universal ones, albeit the difference will depend on many factors such as the incentives to comply, the professionalization of the service administering delivery etc. Overview of Policy Scenarios To quantify policy effects, most scholarly work compares poverty indicators using income before and after the transfers that are of interest. For example, Table 2 shows poverty rates and the mean poverty gap for children using market and social security replacement incomes9 (before transfers) as well as disposable income that includes all other transfers (after transfers), including child related ones. Based on these figures, one may conclude that non-contributory and means-tested benefits are more effective in reducing child poverty in the Czech Republic compared to Romania, irrespective of which poverty indicator is used. What this type of comparisons cannot tell us is the extent to which the Czech policies would be similalry effective in a different context. More specifically, if the characteristics of the population and/or the wider tax-benefit system in which they operate changed, would the Czech policies still achieve the same impressive level of poverty reduction?Table 2 Child poverty indicators before and after transfers Child poverty rates (%) Child poverty gap (%) 60 % Of median eq. DPI 40 % Of median eq. DPI 60 % Of median eq. DPI 40 % Of median eq. DPI Before transfers After transfers Before transfers After transfers Before transfers After transfers Before transfers After transfers RO 31.30 31.05 21.03 16.8 51.31 34.28 52.49 24.45 CZ 17.6 9.6 9.3 1.5 43.53 20.31 55.1 13.5 DPI disposable income; market incomes include replacement income from the tax-benefit system such as pensions, sickness benefits, unemployment benefits etc Source: Author’s calculations based on EU-SILC 2008 and EUROMOD G1.4 More generally, we are interested in the role and interconnections between three distinct elements, namely population characteristics, the features of the tax-benefit system and the policies contained in each country’s child support package. By population characteristics we mean all individual or household characteristics that can affect tax liabilities or benefit entitlements. They include demographic characteristics (e.g. age, household composition, gender), labour market characteristics (e.g. employment status, hours worked, occupation) and all market incomes. They also include a small number of transfers that are normally considered part of the tax-benefit system but are not simulated by EUROMOD,10 most notably pensions. These elements are taken directly from the underlying EUROMOD input datasets. In using the term ‘tax-benefit system’ we refer to sum of tax and benefit policies simulated in EUROMOD. These include social insurance contributions, income tax, contributory unemployment benefits as well as universal and means-tested transfers [see Münich and Pavel (2012); and Stroe et al. (2012) for a complete description of what is simulated in EUROMOD in each country]. Finally, the child support package is the sum of the four child related policies described in Sect. 3, as simulated by EUROMOD. To gain a better understanding of how each element affects the others, we simulate all possible combinations between population characteristics as captured by the data (Romanian and Czech), tax-benefit system (Romanian and Czech) and child policies (Romanian and Czech—standard and budget neutral). In each country, in addition to the existing systems in 2007, we simulate three types of policy counterfactuals. To proxy for the income distribution that would be observed in the absence of child related transfers (i.e. pre-transfer income), most research simply uses disposable income minus these transfers. This approach assumes that all the other elements of the tax-benefit system would remain unchanged once child related transfers are eliminated. This assumption is however questionable as many elements of the tax-benefit system are income dependent. For example, if child related transfers are included in the means-test of general social assistance, removing child related transfers would make some families eligible for higher social assistance payments. These adjustments would be automatic. Thus, using disposable income minus child related transfers to approximate the pre-transfer income distribution will usually overestimate the impact of the child related policy package. To address this problem, in our first simulated policy counterfactual, we remove the existing child support package and re-calculate disposable incomes, allowing other elements of the tax-benefit system to react to the changed circumstances of previously eligible families. This scenario provides us with a benchmark against which all policy effects are measured. By comparing it with the original systems, we obtain the net additional effect of the existing child support package on child poverty, conditional on the original population characteristics and wider tax-benefit system. In the second set of counterfactual scenarios, we introduce the other country’s child related policies, adjusting the monetary parameters in two ways. In the standard policy swap, we transform all monetary policy parameters (income limits, benefit amounts etc.) based on the value of median equivalised disposable income.11 This allows us to mirror the generosity of transfers and tax concessions relative to the poverty threshold. Subsequently, we perform a budget-neutral swap where monetary parameters are calibrated so as to keep total aggregate costs constant. Note that budget neutrality is imposed at the tax-benefit system level rather than the child policy package level so as to take into account any potential interactions with the other elements in the system. Finally, we run the original policy system and the simulated counterfactuals using the other country’s dataset as input. This allows us to understand the role of population characteristics in determining the final policy impact we are interested in. To perform this last set of simulations, incomes in the input datasets are adjusted based on the exchange rate. We also construct a small number of variables needed for the simulations, replicating as much as possible their construction in the other country’s dataset. Table 3 presents an overview of all the simulated policy scenarios. Thus, we obtain 16 income distributions that allow as to investigate interactions as follows.Table 3 Overview of simulated scenarios Scenario Data T-B system Child policies 1 RO RO None 2 RO RO RO 3 RO RO CZ (standard) 4 RO RO CZ (budget neutral) 5 RO CZ None 6 RO CZ CZ 7 RO CZ RO (standard) 8 RO CZ RO (budget neutral) 9 CZ RO None 10 CZ RO RO 11 CZ RO CZ (standard) 12 CZ RO CZ (budget neutral) 13 CZ CZ None 14 CZ CZ CZ 15 CZ CZ RO (standard) 16 CZ CZ RO (budget neutral) Source: Authors’ compilation First, we start by examining the impact of the child related package on child poverty in the ‘usual’ way. We compare the percent reduction in child poverty indicators achieved by Romanian policies in Romania [Scenario2 (S2) vs. Scenario 1 (S1)] and by Czech policies in the Czech Republic (S14 vs. S13). Second, to examine the extent to which the impact of child related policies depends on population characteristics, we vary the underlying population while keeping the tax-benefit system fixed. We first examine how the effect of Romanian child policies changes when the characteristics of the population change. We thus compare the effect of the Romanian child benefits (and tax concession) in the Romanian tax benefit system using first Romanian data (S2 vs. S1) and then Czech data (S10 vs. S9). We repeat the same exercise for the Czech child related policies (S3 vs S1 and S11 vs. S9). Finally, we examine the effect of Romanian and Czech policies respectively in the Czech tax benefit system while using Romanian (S7 vs. S5 and S6 vs. S5) and Czech data (S15 vs S13 and S14 vs S13). Third, we examine the interaction between child support policies and other elements of the tax-benefit system, given population characteristics. To this end, we keep the population characteristics (i.e. data) fixed while we vary the tax-benefit system. We first look at the extent to which the effect of the Romanian child policies on the Romanian population is different when the policies are applied in conjunction with the Romanian tax-benefit system (S2 vs. S1), and with the Czech tax-benefit system respectively (S7 vs. S5). Similarly, we calculate the effect of the Romanian policies on the Czech data using first the Romanian tax-benefit system (S10 vs S9) and then the Czech system (S15 vs S13). Finally, we repeat the same set of calculations for the Czech policies, first examining their effect on the Romanian population (S3 vs S1 and S6 vs S5) and then on the Czech population (S11 vs. S9 and S14 vs S13). Fourth, we separate child poverty impacts stemming from benefit generosity from those coming from policy design by comparing the effect of introducing the other country’s child support package with monetary parameters adjusted relative to the poverty threshold and relative to the budget size respectively. Policy effects are calculated as the difference between poverty indicators relative to the scenario when no child policies are present (keeping all the other elements constant). More formally, policy effects are calculated as. PEα,Popi,TBi,CPi=FGTαPopi,TBi,CP0-FGTαPopi,TBi,CPiFGTαPopi,TBi,CP0 where α = 0, 1, 2 is the FGT parameter; Pop = population characteristics; TB = tax-benefit system; CP = child related policy package, with CP0 indicating that no child related policies are present; i = CZ, RO. Results Interactions Between Policy Effects and Population Characteristics Table 4 shows the effect of the child policy packages in the Czech Republic and Romania under all data-tax-benefit system combinations. The “standard” approach in the literature covering cross-national comparisons of policy effects is to compare the effect of the policies in the environment from which they originated. In this case, one would compare the effects of the Romanian policies in Romania (column A) with the effects of the Czech policies in the Czech Republic (column H). In this setting, one can conclude that the Czech child related policies are indeed much more effective at poverty reduction among families with children, especially if the higher poverty threshold is chosen. For example, poverty rates (using the higher poverty threshold) are reduced by approximately 38 % by the Czech child related policies, whereas their Romanian counterparts achieve only a 14 % reduction. However, this approach assumes that the policy effect is independent of population characteristics and other policies being present.Table 4 Anti-poverty effects of child related policies across tax-benefit contexts and populations characteristics Indicators Effect of RO policies Effect of CZ policies A S1–S2 B S9–S10 C S5–S7 D S13–S15 E S1–S3 F S9–S11 G S5–S6 H S13–S14 RO TB sys CZ TB sys RO TB sys CZ TB sys RO pop CZ pop RO pop CZ pop RO pop CZ pop RO pop CZ pop Poverty—60 % of median income FGT0 −13.86 −35.77 −11.04 −32.24 −15.72 −43.36 −7.84 −38.18 FGT1 −26.87 −46.49 −16.92 −33.81 −33.41 −56.78 −18.81 −38.20 FGT2 −35.47 −56.17 −22.11 −36.26 −44.20 −61.70 −26.41 −38.05 Severe poverty—40 % of median income FGT0 −27.40 −69.18 −17.49 −38.80 −34.05 −74.64 −22.38 −36.15 FGT1 −41.15 −72.20 −23.14 −41.96 −52.78 −78.19 −29.01 −44.25 FGT2 −49.45 −76.94 −30.93 −41.01 −60.28 −81.75 −37.41 −43.70 All figures represent percentage reduction in the poverty indicators relative to the scenario when no child related policies are present (keeping population and the tax-benefit system constant); all figures refer to households with children. Each column shows which scenarios are being compared to derive policy effects (ex: the effects in column A are derived as the reduction in poverty indicators between scenarios 1 and 2 relative to scenario 1 (S1–S2)/S1) Source: Authors’ calculations based on EUROMOD G1.4 To test the sensitivity of the child policy effects to population characteristics, one can compare for example columns A and B. These show the effects of the Romanian child policy package when applied to the Romanian population and to the Czech population respectively. It is clear from Table 4 that the Romanian policies are much more effective in reducing poverty among families with children when they are applied to the Czech population. For example, the reduction in the child poverty rates achieved in the context of the Czech population is approximately three times as large as that achieved using the Romanian population. Similalry, the reduction in the rate of severe poverty is more than twice as large when Romanian policies are used with Czech data compared to when they are used with Romanian data. Similarily, the effect of the Czech child policies is much lower when these policies are evaluated using Romanian data (columns G and H or columns E and F). Clearly, population characteristics play a very important role in shaping the impact of policies. Both the Romanian and Czech child support policies are much more effective in reducing poverty and severe poverty when applied to the Czech population, irrespective of the wider tax-benefit system. It appears that features of the Romanian population make it harder to achieve poverty reduction for any set of policies aimed at families with children. This aspect would not be captured if one were to compare policy effects in the “standard” way (comparing columns A and H). More specifically, the much larger anti-poverty effect of Czech child related policies is at least in part due to the characteristics of the Czech population. Interactions Between Policies and the tax Benefit System Given Population Characteristics Next, we examine the interactions between the child support packages available in the two countries and the respective tax-benefit systems. We start with the effect of introducing the Romanian child support policies into the Romanian and Czech policy systems respectively, using first Romanian and then Czech data. As shown in Table 4, the anti-poverty effect of the Romanian policies is somewhat stronger when they are introduced in the Romanian tax-benefit system (columns A vs. C and B vs. D). This is true irrespective of using the Romanian or the Czech datasets and concerns almost all poverty indicators. For example, looking at Romanian policies severe child poverty is reduced by 27 % when introducing the Romanian child support package in the Romanian system but only by 17 % when introduced in the Czech system (columns A and C). Similarly, in the case of the Czech children, severe child poverty is reduced by 69 % when policies are combined with the Romanian tax-benefit system but only 39 % in combination with the Czech system (columns B and D). A similar pattern is observed when analysing the poverty gap or the poverty severity. For example, Romanian child related policies reduce the poverty gap by 41–72 % (depending on population characteristics) when introduced in the Romanian system as opposed to 23–42 % when introduced in the Czech system. The Czech child related transfers and tax concessions are also generating stronger poverty reduction when used within the Romanian tax-benefit system. Table 4 illustrates the reduction in child poverty indicators when introducing the Czech child-related policies in the Romanian and Czech tax-benefit systems respectively. For example, using Romanian data, poverty in households with children is reduced by 16 % when the policies are introduced in the Romanian system but only by 8 % when introduced in the Czech system (columns E and G). The difference in the effectiveness of the Czech policy bundle appears even stronger when simulations are performed using Czech data. Again, although the differences vary from indicator to indicator, generally, policies are more effective when introduced within the Romanian tax-benefit system rather than the Czech one. To illustrate, severe poverty among families with children is reduced by 75 % when pairing policies with the Romanian system (column F). In contrast, introducing the policies within the Czech system reduces severe poverty by around 36 % (column H). To sum up, the Romanian tax-benefit system appears to magnify the anti-poverty effects of child income support measures, regardless of population characteristics. Both the Romanian and the Czech child related packages have enhanced effects when applied on top of the Romanian tax-benefit rules. One possible explanation may be that, excluding child related instruments, the Romanian tax-benefit system’s ability to reduce child poverty is lower. As a result of the ineffectiveness of the other instruments in the Romanian tax-benefit system, ‘more poverty’ is left to be dealt with by the child related instruments and hence, the latter appear to be more effective. Can Context Alter the Ranking of Policy Effects? The anti-poverty effect of Romanian and Czech child related policies is highly dependent on the context in which they operate. Both sets of policies are more effective when they operate within the Romanian tax-benefit system and on the Czech population. However, from a policy perspective, it is probably more interesting to find out which set of policies is more effective while keeping the population characteristics and the overall features of the tax-benefit system fixed. In the “standard” comparison where each set of child related policies is assessed using the context where it originated from, Czech child related policies appear to be more effective at reducing poverty among families with children compared to Romanian policies (columns A and H). Whereas the precise level of achieved poverty reduction may vary with population characteristics or the features of the tax-benefit system, are Czech policies always more effective than the Romanian ones irrespective of context? When assessing the two sets of policies in the context of the Romanian tax-benefit system, the Czech child policies outperform the Romanian ones on almost all poverty indicators. This result holds when inputting both Romanian (columns A vs. E) and Czech data (columns B vs. F) into the simulated counterfactuals. Thus, given the characteristics of the Romanian tax-benefit system and of the Romanian population, Czech policies are able to effect greater poverty reduction among families with children. However, when looking in the context of the Czech tax-benefit system, the performance of the two sets of policy packages is very similar. This is the case both when policies are applied to the Romanian (columns C vs. G) and Czech populations (columns D vs. H). In fact, the Romanian set of policies is more effective at reducing poverty rates, gap and severity for some groups such as families with very young children (results not shown). Thus, the context in which a policy operates can affect not only the absolute magnitude of the estimated policy effects but can also, in some instances, reverse rankings. In our case, the interaction between child related policies and the wider tax benefit system can alter which set of policies is deemed more effective. In the context of the Romanian tax-benefit system, Czech policies are more effective. However, in the context of Czech fiscal and social rules, the two sets of policies generate similar anti-poverty effects, with the Romanian package outperforming the Czech one on some indicators. Generosity Versus Policy Design The last issue we investigate is the role of policy design versus the generosity of the child support package. Admittedly, the size of the transfers/tax concessions is a feature of the policy, and thus could be considered as part of policy design. However, since budgetary resources are not unlimited, it is useful to separate out policy effects due to simply increased spending. For this purpose, in addition to our ‘standard’ policy swaps, we simulate corresponding counterfactuals where all monetary parameters have been adjusted so that the total spending equals the cost of the policies we are replacing (for similar studies see Salanauskaite and Verbist 2013; Levy et al. 2009). Two things should be noted. First, since we are simultaneously replacing four policies, there are potentially many possibilities to obtain a budget neutral counterfactual. We solve this problem by adjusting all the parameters by the same ratio. This strategy also has the advantage that it keeps the relative sizes of the four policies we introduce equal to those in the original system. Second, the budget neutrality is enforced at the tax-benefit system level, not at the policy level. In taking this approach, we account for all budgetary effects generated by interactions between the new policies and the rest of the fiscal and social rules. To give an example, the introduction of more generous child benefits will increase the direct costs. However, if these child benefits are taxable/included in the means-test of other benefits, part of the increased costs will be offset by increased revenue/smaller outlays in other policy areas. Comparing standard and budget-neutral scenarios of the Czech policies in Romania, the latter are clearly more effective in all dimensions (see Table 5). The differences are rather large for all indicators, averaging around 10 percentage points. While the Czech system relies on income-testing quite a lot, the Romanian one is more universalist with the result that it is generally more expensive. Thus, swapping the Romanian child package for the Czech one and adjusting the monetary parameters based on the values of the poverty thresholds actually costs less. Therefore, to achieve budget neutrality, the parameters from the ‘standard’ scenario need to be scaled up by 22 %. As a result, the child benefit package is more generous in the budget neutral scenario and thus achieves better poverty reduction. Coming back to results presented in Table 4, the Czech set of policies (in the ‘standard’ version) outperforms the Romanian one despite a lower budget.Table 5 Policy generosity versus policy design: anti-poverty effect of ‘standard’ versus budget neutral policy swaps Indicators RO policies in the CZ TB sys CZ policies in the RO TB sys Standard S13–S15 Budget neutral S13–S16 Standard S1–S3 Budget neutral S1–S4 Poverty—60 % of median income FGT0 −32.24 −26.58 −15.72 −21.42 FGT1 −33.81 −42.82 −33.41 −39.68 FGT2 −37.28 −54.28 −41.19 −51.21 Poverty—40 % of median income FGT0 −38.80 −69.09 −34.05 −41.96 FGT1 −41.96 −78.65 −52.78 −61.07 FGT2 −41.01 −80.54 −58.93 −67.98 Policy effects have been computed relative to the scenario when no child related policies are present (keeping population and the tax-benefit system constant); all figures refer to households with children. Each column shows which scenarios are being compared to derive policy effects (ex: the effects in the first column are derived as the reduction in poverty indicators between scenarios 15 and 13 relative to scenario 13 (S13-S15)/S13 Source: Authors’ calculations based on EUROMOD G1.4 Since the Romanian child package is generally more expensive, its parameters need to be scaled down compared to the standard scenario to achieve budget neutrality. Indeed, the adjustment factor is 0.62 indicating that the needed reduction is quite substantial. Based on this downward adjustment, we would expect the budget neutral swap to perform worse compared to the standard one. Indeed, this is the case when we look at poverty rates defined using the higher income threshold. Nevertheless, differences are small despite the large correction factor. Moreover, both the poverty gap and poverty severity are better mitigated in the budget neutral scenario, despite lowering amounts disbursed via child benefits. In addition, all three poverty indicators show that severe poverty drops much more dramatically in the budget neutral scenario compared with the standard swap. This finding may seem counterintuitive. However, remember that budget neutrality is attained at the system, not at the policy level. It is possible that lower outlays in the form of child related transfers and tax deductions are compensated by increases in other elements of the tax-benefit system. Indeed, disposable income in the first three deciles is virtually unchanged between the two counterfactuals whereas the poverty line is higher (as expected) in the standard scenario (results not shown). This finding highlights (again!) the importance of policy interactions in shaping the overall effect. The capacity of the Czech system in reaching the poor combined the untargeted nature of Romanian policies mean that reducing the latter and increasing parts of the former may lead to better anti-poverty results. Discussion and Conclusions This paper has examined the anti-poverty effect of child contingent policies in Romania and the Czech Republic, paying particular attention to their sensitivity to population characteristics and the wider tax-benefit system they are embedded in. We find that both population characteristics and the other fiscal and social policies exert a substantial influence on policy effects. For example, both the Romanian and the Czech child contingent transfers are more effective when applied in the Romanian tax-benefit system. On the other hand, when applied in the Czech tax-benefit system, both sets of policies have similar effects. This suggests that the Czech tax-benefit system is effective on its own (even in the absence of the child related policies) thus limiting any policy effects coming from the policies themselves. Conversely, the Romanian tax-benefit system has a smaller effect leaving more leeway for policies to have an impact. Thus, policy effects cannot accurately be evaluated independent of the wider tax-benefit system in which they are supposed to operate, as interactions with the other elements of the system are pervasive and play a significant role in determining impact. We find that quite apart from size, policy design matters on its own. As the case of the Czech policies demonstrates, it is possible to achieve enhanced anti-poverty results on a lower budget. Moreover, it is not clear that increasing the size of the transfers will always lead to better poverty related outcomes. On the contrary, as shown in the case of the Czech Republic, there may be substitution and trade-offs at the bottom of the income distribution that are less likely to occur in the middle or at the top, especially if targeting is used extensively. As a result, in the absence of co-ordination with other instruments, increased spending on some transfers may be compensated by lower benefits/higher taxes in another area. Another consistent finding emerging from our analyses is the role of population characteristics. Both the Romanian and the Czech policy packages achieve larger poverty reduction when used together with the Czech population. In our setup, we cannot explicitly disentangle which features of the Czech population are responsible for this result, but we can hypothesize that the much lower inequality of market incomes in the Czech Republic plays a role. If this is the case, it suggests that poverty mitigation is likely to be much harder when the incomes of the poor and the rich are far apart regardless of what transfer instruments are in place. It may thus be more efficient for public policies to focus on limiting inequality of market incomes in the first place (through such measures as activation policies, minimum wage setting, steep taxation of very high incomes to discourage their occurrence etc.) rather than trying to direct more resources to the poor via transfers. Overall, our results point to the importance of interactions between the various policy instruments operating within the same system, as well as to complex linkages between population characteristics and policy design. In principle, the effect of a given set of policies in a particular context cannot be inferred from the effect of the same set of policies in a different context. Unfortunately, these complexities make policy benchmarking and policy learning all the more difficult. What seems to be working very well in one context may not work in another. EU-wide policy reviews recognize these issues explicitly or implicitly when they recommend an ‘appropriate policy mix’ (TARKI 2010). However, what an ‘appropriate policy mix’ should contain still eludes us. Future research should focus on disentangling which population and system characteristics ‘fit’ with which types of policies. Appendix: Parameters of the Child Related Policies See Tables 6 and 7.Table 6 Romanian child related policies RO policies in RO system (S2 and S10) RO policies in CZ system-standard (S7 and S15) RO policies in CZ system- budget neutral (S8 and S16) Universal child benefit-amounts 200 RON/month if age < 2; 25 RON/month if age ≥ and age < 18 5628 CZK/month if age < 2; 704 CZK/month if age ≥ 2 and age < 18 3489 CZK/month if age < 2; 436.5 CZK/month if age ≥ 2 and age < 18 Universal birth grant-amount 150 RON/year if age = 0; in addition, 204 RON/year if age = 0 and total number of children ≤4 4221 CZK/year if age = 0; in addition, 5721 CZK/year if age = 0 and total number of children ≤4 2617 CZK/year if age = 0; in addition, 3547 CZK/year if age = 0 and total number of children ≤4 Means-tested family benefit -threshold 176 RON/per month per person 4953 CZK per month per person 3071 CZK per month per person Means-tested family benefit-amounts Between 36 and 79 RON per month, depending on number of children and type of household Between 1013 and 2223 CZK per month depending on number of children and type of household Between 628 and 1378 CZK per month depending on number of children and type of household Tax allowance for dependent children-eligibility Employment income below 3000 RON/month; tax allowance starts being reduced when employment income surpasses 1000 RON/month Employment income below 84,446 CZK/month; tax allowance starts being reduced when employment income surpasses 28,142 CZK/month Employment income below 52,357 CZK/month; tax allowance starts being reduced when employment income surpasses 17,448 CZK/month Tax allowance for dependent children-eligibility 100 RON/month 2814 CZK/month 1745 CZK/month 1 RON = approx. 5.60 CZK (according to June 2007 exchange rates); simulations to which the parameters apply in parentheses Source: EUROMOD G1.4; the conversion of the parameters has been made by multiplying the values in the Romanian system by median household disposable income in CZ/median household disposable income in RO Table 7 Czech child related policies CZ policies in the CZ system (S6 and S14) CZ policies in the RO system-standard (S3 and S11) CZ policies in the RO system- budget neutral (S4 and S12) Means-tested child allowance-minimum living standard-adults Between 2600 and 3126 CZK/month depending on the type of household Between 92 and 111 RON/month depending on the type of household Between 112 and 135 RON/month depending on the type of household Means-tested child allowance-minimum living standard-children Between 1600 and 2250 CZK/month, depending on age Between 57 and 80 RON/month, depending on age Between 70 and 98 RON/month, depending on age Birth grant amount Between 17,760 and 79,680 CZK/year depending on birth rank order Between 631 and 2831 RON/year, depending on rank order Between 770 and 3454 RON/year, depending on rank order Social allowance-family minimum living standard Between 2600 and 3126 CZK/month/adult, depending on household type Between 92 and 111 RON/month/adult, depending on household type Between 112 and 135 RON/month/adult, depending on household type Social allowance- child minimum living standard Between 1600 and 5400 CZK/month, depending on age and disability status Between 57 and 192 RON/month, depending on age and disability status Between 70 and 234 RON/month, depending on age and disability status Child tax credit amount 6000 CZK/month/child 213 RON/month/child 260 RON/month/child 1 RON = approx. 5.60 CZK (according to June 2007 exchange rates); simulations to which the parameters apply in parentheses Source: EUROMOD G1.4; the conversion of the parameters has been made by multiplying the values in the Czech system by median household disposable income in RO/median household disposable income in CZ 1 Figures refer to 2007 total government revenue and social benefits other than social transfers in kind. 2 A description of the tax-benefit system in Romania and the Czech Republic in 2007 (our policy year) can be found in the respective EUROMOD country reports (Münich and Pavel 2012; Stroe et al. 2012). 3 While newer policy years were available in EUROMOD at the time this study started, we chose 2007 to match it with the year of our dataset. 4 One important policy instrument that is missing from our analysis are maternal/parental leave benefits; we have opted to exclude them in this case because they are only simulated in Romania. 5 We use version G1.4. 6 We opt to assume 100 % compliance and take-up rates as there is very little information on which modelling of tax evasion and/or benefit non-take-up can be based. 7 The modified OECD equaivalence scale assigns a weight of 1 for the first adult, 0.5 for all subsequent adults and 0.3 for children. Children are defined as being aged 13 or less. 8 This is an often made assumption despite it being innacurate. Some studies (Ward-Batts 2005; Lundberg et al. 1997) have shown that public transfers received by the mother are more likely to be spent on children invalidating the complete intrahousehold redistribution assumption. However, we ignore this issue in this study. 9 Replacement incomes include pensions, sickness benefits, unemployment benefits, maternity and parental leave benefits. 10 Pensions and some smaller transfers such as sickness and maternity benefits are not simulated due to data limitations in the EU-SILC. 11 More specifically, we apply the following formula; new monetary parameter = old monetary parameter* (median equivalised disposable income in the “receiver” country/median equivalised disposable income in the “donor” country); for example when introducing Czech policies in Romania, the new policy parameters in Romania will be = CZ policy parameter*(RO median equivalised household disposable income/CZ median equivalised household disposable income). Acknowledgements This paper uses EUROMOD version G1.4. We are indebted to Holly Sutherland and all members of the EUROMOD consortium for making EUROMOD freely accessible for research purposes. EU-SILC micro-data have been provided by EUROSTAT. Avram acknowledges financial support from the Economic and Social Research Council (ESRC) through the Research Centre for Micro-Social Change, grant no RES-518-28-00. All results and their interpretation presented in this paper remain the authors’ responsibility. Compliance with Ethical Standards Conflict of interest The authors declare they have no competing interests. ==== Refs References Aber JL Bennett NG Conley DC Li J The effects of poverty and on child health and development American Review of Public Health 1997 87 463 483 10.1146/annurev.publhealth.18.1.463 Almond D Hoynes H Witmore Schanzenbach D Inside the war on poverty: The impact of food stamps on birth outcomes The Review of Economics and Statistics 2011 93 2 387 403 10.1162/REST_a_00089 Atkinson AB Van de Walle D Nead K On targeting social security: Theory and western experience with family benefits Public Spending and the Poor. Theory and evidence 1995 Baltimore John Hopkins University Press 25 68 Barrientos A DeJong J Reducing child poverty with cash transfers: A sure thing? Development Policy Review 2006 24 5 537 552 10.1111/j.1467-7679.2006.00346.x Black MM Reiner Hess C Berenson-Howard J Toddlers from low-income families have below normal mental, motor, and behavior scores on the revised Bayley scales Journal of Applied Developmental Psychology 2000 21 6 655 666 10.1016/S0193-3973(00)00059-9 Bradbury B Jäntti M Bradbury B Jenkins SP Micklewright J Child poverty across twenty-five countries The dynamic of child poverty in industrialised countries 2001 Cambridge Cambridge University Press 62 91 Bradshaw J A review of the comparative evidence on child poverty 2006 York Joseph Rowntree Foundation Case A Lubotsky D Paxson C Economic status and health in childhood: The origins of the gradient American Economic Review 2002 92 5 1308 1334 10.1257/000282802762024520 de Neubourg, C., Castonguay, J., & Roelen, K. (2007). Social safety nets and targeted social assistance: Lessons from the European experience. SP discussion papers. The World Bank, Washington D.C. Duncan GJ Yeung WJ Brooks-Gunn J Smith JR How much does childhood poverty affect life chances of children? American Sociological Review 1998 63 3 406 423 10.2307/2657556 Duncan GJ Ziol-guest KM Kalil A Early-childhood poverty and adult attainment, behaviour and health Child Development 2010 81 1 306 325 10.1111/j.1467-8624.2009.01396.x 20331669 EUROSTAT. http://circa.europa.eu/Public/irc/dsis/eusilc/library?l=/udb_user_database&vm=detailed&sb=Title, http://circa.europa.eu/Public/irc/dsis/eusilc/library?l=/udb_user_database&vm=detailed&sb=Title. Engle PL Black MM The effect of poverty on child development and educational outcomes Annals of the New York Academy of Sciences 2008 1136 1 243 256 10.1196/annals.1425.023 18579886 Figari F Paulus A Sutherland H Measuring the size and impact of public cash support for children in cross-national perspective Social Science Computer Review 2011 29 1 85 102 10.1177/0894439310370104 Förster MF Tóth IG Child poverty and family transfers in the Czech Republic, Hungary and Poland Journal of European Social Policy 2001 11 4 324 341 10.1177/095892870101100403 Foster J Greer J Thorbecke E A class of decomposable poverty measures Econometrica 1984 53 3 761 766 10.2307/1913475 Gornick, J. C., & Jäntti, M. (2010). Child poverty in upper-income countries: Lessons from the Luxembourg Income Study. In: S. Kamerman, S. Phipps, A. Ben-Arieh (Eds.), From child welfare to child well-being: An international perspective on knowledge in the service of policy making. Children’s well-being: Indicators and research (Vol. 1, pp. 339–368) New York: Springer. Gornick, J. C, & Jäntti, M. (2011). Child poverty in comparative perspective: Assessing the role of family structure and parental education and employment. LIS working paper series (Vol. 570) Luxembourg: LIS. Hoynes H Page M Huff Stevens A Can targeted transfers improve birth outcomes? Evidemce from the introduction of the WIC program Journal of Public Economics 2011 95 7–8 813 827 10.1016/j.jpubeco.2010.12.006 Hoynes, H., Schanzenbach, D. W., & Almond, D. (2012). Long run impacts of childhood access to the safety net. NBER working papers (Vol. 18535). Immervoll H Sutherland H de Vos K Vleminckx K Smeeding T Reducing child poverty in the European Union: The role of child benefits Child well-being, child poverty and child policy in modern nations. What do we know? 2000 Bristol The Policy Press 407 432 Jäntti M Bradbury B Child poverty across industrialized countries Journal of Population and Social Security (Population) 2003 1 385 409 Jarvis S Micklewright J Van de Walle D Nead K The Targeting of family allowances in Hungary Public spending and the poor. Theory and evidence 1995 Baltimore John Hopkins University Press 294 320 Korpi W Palme J The paradox of redistribution and strategies of equality: Welfare state institutions, inequality and poverty in Western countries American Sociological Review 1998 63 5 661 687 10.2307/2657333 Levy H Morawski L Myck M Lelkes O Sutherland H Alterbative tax-benefit strategies to support children in Poland Tax and benefit policies in the enlarged Europe: Assessing the impact with microsimulation models 2009 Farnham Ashgate 125 152 Lundberg SJ Pollak RA Wales TJ Do husbands and wives pool their resources? Evidence from the United Kingdom child benefit Journal of Human Resources 1997 32 3 463 480 10.2307/146179 Matsaganis M O’Donoghue C Levy H Coromaldi M Mercader-Prats M Farinha Rodrigues C Toso S Tsaklouglou P Spadaro A Child poverty and family transfers in Southern Europe Microsimulation as a tool for the evaluation of public policies. Methods and applications 2007 Bilbao Fundación BBVA Münich, D., & Pavel, J. (2012) EUROMOD country report: Czech Republic (CZ), 2007–2010. ISER. Najman JM Hayatbakhsh MR Heron MA Bor W O’Callghan MJ Williams GM The impact of episodic and chronic poverty on child cognitive development The Journal of Pediatrics 2009 154 2 284 289 10.1016/j.jpeds.2008.08.052 19038402 Nelson K Mechanisms of poverty alleviation: Anti-poverty effects of non-means-tested and means-tested benefits in five welfare states Journal of European Social Policy 2004 14 4 371 390 10.1177/0958928704046879 Notten G Gassman F Size matters: Targeting efficiency and poverty reduction effects of means-tested and universal child benefits in Russia Journal of European Social Policy 2008 18 3 260 274 10.1177/0958928708091059 Oxley H Dang T-T Förster MF Pellizzari M Vleminckx K Smeeding T Income inequalities and poverty among children and households with children in selected OECD countries Child well-being, child poverty and child policy in modern nations. What do we know? 2000 Bristol The Policy Press 371 405 Petterson SM Burke Albers A Effects of poverty and maternal depression on early child development Child Development 2001 72 6 1794 1813 10.1111/1467-8624.00379 11768146 Salanauskaite L Verbist G Is the neighbour’s grass greener? Comparing family support in Lithuania and four other new member states Journal of European Social Policy 2013 23 3 315 331 10.1177/0958928713480066 Stroe, C., Militaru, E., Avram, S., & Cojanu, S. (2012). EUROMOD country reports: Romania, 2007–2010. ISER. Sutherland H Figari F EUROMOD: The European Union tax-benefit microsimulation model International Journal of Microsimulation 2013 6 1 4 26 Sutherland H Piachaud D Reducing child poverty in Britain: An assessment of government policy 1997–2011 The Economic Journal 2011 111 469 F85 F101 20027731 TARKI. (2010). Child poverty and child well-being in the European Union. Report prepared for the DG employment, social affairs and equal opportunities (Unite E.2) of the European commission. Budapest: TARKI. Van Mechelen N Bradshaw J Marx I Nelson K Child poverty as a government priority: Child benefit packages for working families, 1992–2009 Minimum income protection in flux 2013 Basingstoke Palgrave Macmillan 81 107 Ward-Batts J Out of the Wallet and into the purse: Using microdata to test income pooling Journal of Human Resources 2005 43 2 325 351 Whiteford, P., & Adema, W. (2007). What works best in reducing child poverty: A benefit or work strategy? OECD social, employment and migration working papers (Vol. 51) Paris: OECD Publishing.

PMC005xxxxxx/PMC5002038.txt

==== Front Abdom Radiol (NY)Abdom Radiol (NY)Abdominal Radiology (New York)2366-004X2366-0058Springer US New York 76910.1007/s00261-016-0769-9ArticleSingle-session alcohol sclerotherapy of symptomatic liver cysts using 10–20 min of ethanol exposure: no recurrence at 2–16 years of follow-up Larssen Trond Bjerke 12Viste Asgaut 23Horn Arild 3Haldorsen Ingfrid Salvesen 12Espeland Ansgar 47-55973898ansgar.espeland@gmail.com 121 Department of Radiology, Haukeland University Hospital, Jonas Liesvei 65, 5021 Bergen, Norway 2 Department of Clinical Medicine, University of Bergen, Pb 7804, 5020 Bergen, Norway 3 Department of Acute and Gastrointestinal Surgery, Haukeland University Hospital, Jonas Liesvei 65, 5021 Bergen, Norway 7 5 2016 7 5 2016 2016 41 9 1776 1781 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Purpose To assess long-term results after single-session alcohol sclerotherapy of symptomatic benign liver cysts performed with maximum 20 min of exposure to alcohol. Methods We included 47 patients aged 32–88 years (42 women, 5 men) with 51 benign non-parasitic liver cysts that were exposed to ethanol for 7–20 min in a single sclerotherapy session and were followed for at least 24 months. Each cyst was emptied before injecting ethanol (10% of cyst volume, but maximum 100 mL) into it. The patient rotated from side to side to facilitate contact between ethanol and the whole cyst wall. Pre-treatment cyst volume was defined as the volume of aspirated cyst fluid after complete emptying of the cyst. Follow-up cyst volume was estimated based on computed tomography images. Results Cyst volumes were 30–4900 (median 520) mL at pre-treatment and 0–230 (median 1) mL at 24–193 (median 56) months follow-up, a reduction of 83–100% (median 99.7%). No cyst required repeated treatment during the follow-up. Median volume reduction was 99.7% at median 49 months of follow-up for 35 cysts exposed to ethanol for 7–10 min vs. 99.6% at median 75 months of follow-up for 16 cysts exposed for 20 min (p = 0.83, Mann–Whitney test). Ethanol intoxication occurred in one patient. There were no other complications except for pain. Conclusion Long-term results of single-session alcohol sclerotherapy performed with maximum 20 min of exposure to ethanol were satisfactory with no sign of recurrence of cyst fluid. Keywords CystsEthanolLiver diseasesSclerotherapyTreatment outcomeissue-copyright-statement© Springer Science+Business Media New York 2016 ==== Body Alcohol sclerotherapy may be an effective treatment for solitary symptomatic liver cysts and for dominant symptomatic cysts in polycystic livers [1–23]. However, there is much variation in the time the cyst is exposed to alcohol during the procedure (10–240 min), the volume of alcohol applied, and in the use of one single [5–9, 13, 16, 21, 22] or multiple [3, 4, 23] sclerotherapy sessions. These factors affect the laboratory time needed, the risk of alcohol intoxication, and the procedural costs. Thus, research has been recommended to standardize and simplify the method of sclerotherapy [24, 25]. Further studies are also needed on long-term results of sclerotherapy of liver cysts [9, 22, 23]. Earlier reports by Larssen et al. suggest good results of exposing liver cysts to alcohol for 10 or 20 min in a single sclerotherapy session [5, 8, 9]. Here we report results of such a treatment in a larger material after a minimum of two-year follow-up. The purpose of the present study was to assess long-term results after single-session alcohol sclerotherapy of symptomatic benign liver cysts performed with maximum 20 min of exposure to alcohol. Materials and methods This study included 47 patients with at least 24 months of follow-up after alcohol sclerotherapy of symptomatic benign liver cysts performed at our hospital from 1993 to 2010 according to the procedure described below. Patient data were recorded prospectively. Among 68 consecutive patients treated with alcohol sclerotherapy for solitary symptomatic liver cysts or for dominant symptomatic cysts in polycystic livers, we excluded 17 patients followed for <24 months and 4 patients lost to follow-up. The follow-up protocol initially included imaging to assess cyst volume within 2 months, and at 6, 12, and 24 months. We later reduced follow-up imaging during the first year, since it typically showed temporary recollection of cyst fluid that did not require treatment (see “Results” section). The material does not include symptomatic polycystic livers containing only innumerable small cysts. Gastrointestinal surgeons at the hospital with extensive experience in evaluating liver disease had evaluated all patients and judged them in need of invasive treatment due to symptoms from a liver cyst, e.g., pain, fullness, early satiety. Since 1993 at our institution, nearly all such patients have undergone alcohol sclerotherapy rather than primary surgery. If the cause of symptoms was uncertain, the cyst was emptied to test if the symptoms vanished. We did not perform alcohol sclerotherapy of hydatid cysts, neoplastic cysts, or cyst-like ecstasies of the intrahepatic biliary ducts. Contraindications to sclerotherapy were coagulopathy, communication between the cyst and the biliary tree or the peritoneal cavity, and failure to aspirate injected contrast medium from the cyst. Sclerotherapy procedure Premedication (50–100 mg pethidine, 0.6 mg atropine) was used from 1993 to 2002. Since 2002, no premedication was used, but an anesthetic nurse was always present during the procedure and provided the sedation and analgesics needed (midazolam, alfentanil). The patient was positioned supine on an angiographic table. The cyst was punctured under ultrasonographic guidance, usually with a 1.2-mm-diameter 20-cm-long needle. If possible, the needle was inserted through 1-3 cm of liver tissue to prevent leakage from the cyst. A stiff guide wire was introduced and a 30-cm 7 French (2.3 mm outer diameter) pigtail catheter (PBN Medicals, Denmark; Argon Medical Devices, Denmark) was pushed as far as possible into the cyst and fixed to the skin by adhesive tapes. After complete emptying of the cyst, contrast material was injected and radiographs taken to confirm the position of the catheter and to exclude communication with the biliary ducts and leakage of contrast into the peritoneal cavity. If this occurred, or aspiration of contrast material from the cyst failed, the procedure was discontinued. Otherwise, ethanol was injected in an amount of 10% of the cyst volume, but never more than 100 mL. To avoid dilution of ethanol, great care was taken to empty the cyst completely before injecting ethanol into it. To achieve satisfactory contact between ethanol and cyst wall, it was routine to change the patient position from prone to supine and right to left lateral decubitus position at least two times during the procedure. We had tested the pigtail catheter for use with alcohol ex vivo. The manufacturer (Argon Medical Devices) has tested the catheter in contact with 100% alcohol and found no sign of degradation or leakage up to 6 weeks (Argon Report REP1-137707, Nov 07, 2013). Alcohol was injected using regular plastic syringes that were usually filled with alcohol within 3 min from the injection time. Cysts were exposed to ethanol for 20 min (from 1993 to 1995) or 10 min (from December 1995 to June 1999). From June 1999, cysts smaller than 1000 mL were exposed to ethanol once for 10 min, whereas cysts with volumes of at least 1000 mL were exposed to ethanol twice in a single session, each time with 100 mL of ethanol for 10 min. We used 96% ethanol during the period 1993–1999 and 99% ethanol since July 1999. Finally, all alcohol was evacuated, the cyst was irrigated with saline (to avoid leakage of ethanol during withdrawal of the catheter), and the catheter was removed. The patient rested in bed for 4 h and if asymptomatic was discharged from the hospital the same afternoon. The patients were observed for any signs of alcohol intoxication during and after the procedure. The blood alcohol level was not measured routinely, since it was ≤0.3 mg/g in initial patients. Cyst volume Pre-treatment cyst volume was estimated as the volume aspirated on the date of sclerotherapy when emptying the cyst. At follow-up, one experienced radiologist measured the three largest perpendicular cyst diameters d1, d2, and d3 on computed tomography (CT) images, and calculated cyst volume using the formula of an ellipsoid: volume = d1 × d2 × d3 × 0.523 [26]. Statistical analysis WINPEPI version 11.60 (http://www.brixtonhealth.com/pepi4windows.html) was applied for statistical analysis. The Mann–Whitney test for comparing medians was used to compare pre-treatment cyst volumes, follow-up cyst volumes, reduction of cyst volumes at follow-up, and follow-up times between cysts exposed to alcohol for 10 min and cysts exposed to alcohol for 20 min. All p values are two-tailed. p < 0.05 indicated statistical significance. Results Forty-seven patients (42 women and five men) aged 32–88 (median 61) years had 51 liver cysts exposed to ethanol for 7–20 (median 10) minutes and were followed for 24–193 (median 56) months. Six cysts in five patients were in polycystic livers and 45 cysts in 42 patients were not. In the total sample, median cyst volumes were 520 mL at pre-treatment and 1 mL—or median 99.7% smaller—at the last follow-up (Table 1). None of the 51 treated cysts required repeated treatment during the reported follow-up period due to insufficient effect of the sclerotherapy.Table 1 Single-session alcohol sclerotherapy of 51 symptomatic benign liver cysts in 47 patients Age (years) Women/men (numbers) Original cyst volume (mL) Follow-up cyst volume (mL) Reduction of cyst volume at follow-up (%) Time of follow-up (months) Total sample 61 (32–88) 42/5 520 (30–4900) 1 (0–230) 99.7 (83–100) 56 (24–193) Alcohol exposure  10 mina (35 cysts, 32 patientsb) 61 (43–88) 27/5 394 (30–4110) 1 (0–230) 99.7 (83–100) 49 (24–193)  20 min (16 cysts, 16 patientsb) 68 (32–86) 15/1 1525 (200–4900) 5 (0–188) 99.6 (91–100) 75 (24–171) Values are median (range) except values for women/men aIncludes two cysts exposed for 7 and 8 min, respectively bOne patient (a man)is included in both alcohol exposure groups because he had one cyst exposed to alcohol for 10 min and another cyst exposed to alcohol for 20 min Median volume reduction was 99.7% for 35 cysts exposed to ethanol for 7–10 min vs. 99.6% for 16 cysts exposed to alcohol for 20 min (p = 0.83, Table 1). Cysts exposed for 7–10 min vs. 20 min did not differ significantly in volume at follow-up (p = 0.28) or time to follow up (p = 0.24) (Table 1), but were originally smaller (median 394 vs. 1525 mL, p = 0.004; Table 1). This was expected, since cysts ≥1000 mL were exposed to alcohol for 20 min from 1999. Seven (of 18) cysts ≥1000 mL were exposed to alcohol for 7–10 min, and were 88–100% (median 99%) smaller at 78–193 months (median 11 years) of follow-up. We followed 11 cysts in 10 patients for more than 10 years. These cysts were 88–100% (median 100%) smaller at 127–193 months (median 12 years) of follow-up. CT performed within three months after sclerotherapy of 17 cysts in 16 patients showed recollection of cyst fluid in all. The fluid later diminished in all patients without therapy (Fig. 1). At the first follow-up at 0.3–3 (median 1) months, the 17 cysts measured 9–100% (median 39%) of their original volume of 30–4900 (median 1250) mL. All 17 cysts then shrank during the first year. Cyst volume was 0–30% (median 5%) of the original volume at 4.5–6 months (data on 13 cysts) and 0–36% (median 2%) at 10–14 months (data on 13 partly different cysts). At the last follow-up, the 17 cysts had shrank to 0–12% (median 0.3%) of their initial volume.Fig. 1 A 1750 mL liver cyst in a 71-year-old woman before sclerotherapy (A). Recollection of 850 mL cyst fluid was observed 24 days after sclerotherapy (B). However, the cyst (long white arrows) was smaller at 6 months without repeated sclerotherapy (C), had shrank to 99% of its pre-treatment size at 17 months (D), and showed no signs of recurrence at 193 months post treatment (E). An adjacent cyst (short white arrow) had grown slowly to 65 mL at 193 months post treatment (E) There was one major procedure related complication: ethanol intoxication in a 77-year-old woman following exposure of one 4900 mL cyst to 100 mL alcohol for 10 min twice in one session (respiratory failure, alcohol smell from breath, blood alcohol concentration 1.95 mg/g; morphine antidote given, uneventful recovery; cyst disappeared at 77 months of follow-up). A minor complication was abdominal pain during the procedure. This was no longer a problem after we abandoned premedication and started to use analgesics according to individual needs and to irrigate the cysts with saline before removing the catheter, preventing remnants of ethanol within it to cause pain during removal. Discussion In this study, liver cysts exposed to alcohol for a maximum of 20 min in one treatment session were 83–100% (median 97.7%) smaller at 2–16 years of follow-up. No cyst required repeated treatment during the follow-up. Even for large cysts, we used maximum 100 mL of alcohol once or twice in a single session. The procedure caused alcohol intoxication in one patient (2%) with an uneventful recovery. The brief exposure of liver cysts to a relatively small amount of alcohol was thus safe and prevented recurrence of cyst fluid in the long term. Alcohol should be in contact with the whole cyst cavity during the treatment in order to fix the epithelial cells and disable their ability to secret fluid [1, 15, 21, 27]. Alcohol can also induce a temporary inflammatory reaction in a cystic wall deprived of vital epithelium [28]. This may explain the transitory re-accumulation of cyst fluid seen in the present and other studies [5, 6]. Importantly, cyst fluid re-accumulated within three months but diminished without treatment. Some parts of our procedure may be essential when exposing liver cysts to limited amounts of alcohol for a short time. First, in order to avoid dilution of alcohol concentration, it is important to empty the cyst completely before injecting alcohol into it. Second, after alcohol injection, the patient should rotate from side to side to facilitate contact between alcohol and every part of the cyst, especially when the cyst is large and may have a multifolded wall when emptied. Third, to make it easier for the patient to co-operate and change position, we advise effective pain management as needed during the procedure rather than general premedication. We found good results of exposing liver cysts to alcohol for only 10 min, based on treatment of 35 cysts. Larssen et al. have previously reported promising results of exposing liver cysts to ethanol for 10 min in a study comprising only 10 cysts [8]. The present findings suggest that an alcohol exposure time of 10 min may be equally effective as an alcohol exposure time of 20 min in many patients. An alcohol exposure time of 10 min may be effective also for cysts >1000 mL, but was used in only seven such cysts in our study. In their recent retrospective study, Akhan et al. reported successful results of exposing 39 liver cysts to alcohol for 10 min, but all cysts in their material were <700 mL [22]. Therefore, the effectiveness of exposing cysts >1000 mL to alcohol for 10 min needs further validation. Provided the sclerotherapy remains effective, briefer exposure to alcohol during the procedure may be advantageous by reducing alcohol intoxication risk, interventional laboratory time, and duration of hospitalization. In studies of liver cysts exposed to alcohol for 60 min [2, 6], all patients had elevated alcohol levels in the blood, and both prolonged exposure and increased volume of alcohol increased the risk of alcohol intoxication. Our study had strengths and limitations. It represented day-to-day radiological work, making the results highly relevant to clinical practice. Patients and data were noted prospectively, avoiding retrospective identification of patients in medical records. The follow-up period was longer and the material slightly larger (24–193 months, median 56 months; 51 cysts) than in recent studies by Akhan et al. [22] (4–173 months, median 38 months; 39 cysts) and Jang et al. [23] (12–106 months, mean 33 months; 43 cysts). We used CT and the same observer (not ultrasonography performed by different observers) to estimate cyst volumes at follow-up by means of a validated method for estimating organ volumes [26, 29]. However, we did not report on course of symptoms, except that no cyst required repeated treatment. Since prior research had shown good effect of reduced cyst volume on symptoms [2, 5–8], we rather focused on whether our treatment procedure ensured small cyst volumes in the long term. As part of the daily clinical practice, we adjusted the procedure slightly during the study period based on accumulated experience. We reduced the alcohol exposure time from 20 to 10 min based on good results in three initial patients in whom the procedure had to be discontinued due to pain. We later exposed cysts >1000 mL for 20 min (10 min twice) to be more certain of effect. However, our study design was not optimal for comparing results between 10 and 20 min of alcohol exposure, which would have required a randomized trial. In conclusion, the employed procedure of alcohol sclerotherapy of liver cysts varies considerably between institutions. This study showed that there is a durable response to sclerotherapy performed with 10–20 min of exposure to ethanol in a single session. The first author Trond Bjerke Larssen has died. Compliance with ethical standards Conflict of interest The authors declare that they have no conflict of interest. Ethical approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. For this type of study, formal consent is not required. The appropriate research ethics committee did not consider its approval necessary. This article does not contain any studies with animals performed by any of the authors. Informed consent Informed consent was obtained from all individual participants included in the study. ==== Refs References 1. Bean WJ Rodan BA Hepatic cysts: treatment with alcohol AJR Am J Roentgenol 1985 144 237 241 10.2214/ajr.144.2.237 3880981 2. Tikkakoski T Makela JT Leinonen S Treatment of symptomatic congenital hepatic cysts with single-session percutaneous drainage and ethanol sclerosis: technique and outcome J Vasc Intervent Radiol 1996 7 235 239 10.1016/S1051-0443(96)70767-4 3. vanSonnenberg E Wroblicka JT D’Agostino HB Symptomatic hepatic cysts: percutaneous drainage and sclerosis Radiology 1994 190 387 392 10.1148/radiology.190.2.8284385 8284385 4. Andersson R Jeppsson B Lunderquist A Bengmark S Alcohol sclerotherapy of non-parasitic cysts of the liver Br J Surg 1989 76 254 255 10.1002/bjs.1800760313 2720320 5. Larssen TB Viste A Jensen DK Single-session alcohol sclerotherapy in benign symptomatic hepatic cysts Acta Radiol 1997 38 993 997 10.3109/02841859709172116 9394655 6. Kairaluoma MI Leinonen A Stahlberg M Percutaneous aspiration and alcohol sclerotherapy for symptomatic hepatic cysts. An alternative to surgical intervention Ann Surg 1989 210 208 215 10.1097/00000658-198908000-00012 2667475 7. Montorsi M Torzilli G Fumagalli U Percutaneous alcohol sclerotherapy of simple hepatic cysts. Results from a multicentre survey in Italy HPB Surg 1994 8 89 94 10.1155/1994/10372 7880778 8. Larssen TB Rosendahl K Horn A Jensen DK Rorvik J Single-session alcohol sclerotherapy in symptomatic benign hepatic cysts performed with a time of exposure to alcohol of 10 min: initial results Eur Radiol 2003 13 2627 2632 10.1007/s00330-003-1923-7 12955449 9. Larssen TB Jensen DK Viste A Horn A Single-session alcohol sclerotherapy in symptomatic benign hepatic cysts. Long-term results Acta Radiol 1999 40 636 638 10.3109/02841859909175601 10598853 10. Moorthy K Mihssin N Houghton PW The management of simple hepatic cysts: sclerotherapy or laparoscopic fenestration Ann R Coll Surg Engl 2001 83 409 414 11777137 11. Yu JH Du Y Li Y Effectiveness of CT-guided sclerotherapy with estimated ethanol concentration for treatment of symptomatic simple hepatic cysts Clin Res Hepatol Gastroenterol 2014 38 190 194 10.1016/j.clinre.2013.09.008 24210773 12. Feng Y-H Qian L-X Gong H-M Sclerotherapy of simple hepatic cysts by repeated aspiration and alcohol instillation Turk J Gastroenterol 2012 23 359 365 10.4318/tjg.2012.0349 22965507 13. Benzimra J Ronot M Fuks D Hepatic cysts treated with percutaneous ethanol sclerotherapy: time to extend the indications to haemorrhagic cysts and polycystic liver disease Eur Radiol 2014 24 1030 1038 10.1007/s00330-014-3117-x 24563160 14. Blonski WC Campbell MS Faust T Metz DC Successful aspiration and ethanol sclerosis of a large, symptomatic, simple liver cyst: case presentation and review of the literature World J Gastroenterol 2006 12 2949 2954 10.3748/wjg.v12.i18.2949 16718826 15. Cheng D Amin P Ha TV Percutaneous sclerotherapy of cystic lesions Semin Intervent Radiol 2012 29 295 300 10.1055/s-0032-1326934 24293802 16. Yang CF Liang HL Pan HB Single-session prolonged alcohol-retention sclerotherapy for large hepatic cysts AJR Am J Roentgenol 2006 187 940 943 10.2214/AJR.05.0621 16985138 17. Erdogan D van Delden OM Rauws EA Results of percutaneous sclerotherapy and surgical treatment in patients with symptomatic simple liver cysts and polycystic liver disease World J Gastroenterol 2007 13 3095 3100 17589926 18. Okano A Hajiro K Takakuwa H Nishio A Alcohol sclerotherapy of hepatic cysts: its effect in relation to ethanol concentration Hepatol Res 2000 17 179 184 10.1016/S1386-6346(99)00067-4 10794971 19. Valette PJ Chataing L Partensky C Paliard P Treatment of hepatic polycystosis by intracystic injection of alcohol Gastroenterol Clin Biol 1987 11 898 900 3449407 20. Zerem E Imamovic G Omerovic S Percutaneous treatment of symptomatic non-parasitic benign liver cysts: single-session alcohol sclerotherapy versus prolonged catheter drainage with negative pressure Eur Radiol 2008 18 400 406 10.1007/s00330-007-0760-5 17899104 21. Simonetti G Profili S Sergiacomi GL Meloni GB Orlacchio A Percutaneous treatment of hepatic cysts by aspiration and sclerotherapy Cardiovasc Intervent Radiol 1993 16 81 84 10.1007/BF02602983 8485748 22. Akhan O Islim F Balci S Percutaneous treatment of simple hepatic cysts: the long term results of PAIR and catheterization techniques as single-session procedures Cardiovasc Intervent Radiol 2015 23. Jang SY Park SY Tak WY Kweon YO Ryeom H Long-term follow-up of large symptomatic hepatic cysts treated by percutaneous ethanol sclerotherapy Acta Radiol 2016 24. Farges O Menu Y Benhamou J-P Blomgart L Fong Y Non-parasitic cystic diseases of the liver and intrahepatic biliary tree Surgery of the liver and biliary tract 2000 London W. B. Saunders Company LTD 1245 1260 25. Nagorney D Blumgart L Fong Y Surgical management of cystic disease of the liver Surgery of the liver and biliary tract 2000 London W. B. Saunders Company LTD 1261 1276 26. Campbell S Goessens L Goswamy R Whitehead M Real-time ultrasonography for determination of ovarian morphology and volume. A possible early screening test for ovarian cancer? Lancet 1982 1 425 426 10.1016/S0140-6736(82)91622-1 6121093 27. Tapani E Taavitsainen M Lindros K Vehmas T Lehtonen E Toxicity of ethanol in low concentrations. Experimental evaluation in cell culture Acta Radiol 1996 37 923 926 10.1177/02841851960373P247 8995467 28. Larssen TB Rorvik J Horn A Biochemical and cytologic analysis of cystic contents in benign non-parasitic symptomatic hepatic cysts before and after ethanol sclerotherapy Acta Radiol 2004 45 504 509 10.1080/02841850410006588 15515510 29. Campbell S Wladimiroff JW Dewhurst CJ The antenatal measurement of fetal urine production J Obstet Gynaecol Br Commonw 1973 80 680 686 10.1111/j.1471-0528.1973.tb16049.x 4725943

PMC005xxxxxx/PMC5002039.txt

==== Front Soc Indic ResSoc Indic ResSocial Indicators Research0303-8300Springer Netherlands Dordrecht 106610.1007/s11205-015-1066-7ArticleDeterminants of Intergenerational Downward Mobility in the Netherlands Thijssen Lex Wolbers Maarten H. J. m.wolbers@maw.ru.nl Department of Sociology, Radboud University, P.O. Box 9104, 6500 HE Nijmegen, The Netherlands 29 8 2015 29 8 2015 2016 128 3 995 1010 4 8 2015 © The Author(s) 2015 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Several studies have shown that Dutch society has become more open in the last few decades as a result of increasing opportunities for upward social mobility. However, recently it has been observed that the likelihood of downward social mobility has increased for the youngest (male) birth cohorts in the Netherlands. Despite this recent finding, social stratification research has paid little attention to test explanations of downward mobility. This article tries to fill in this knowledge gap by testing three theoretical perspectives that aim to explain intergenerational downward occupational mobility of individuals. In addition, we examine historical developments to study whether the role of these explanations have changed over time. To test the predictive validity of these perspectives, we use data from the Family Survey Dutch Population 2009 (N = 1423). The empirical results, first of all, indicate that individuals who were born in younger birth cohorts are more likely to experience downward occupational mobility than individuals, who were born in older cohorts. We thus replicate earlier findings for the Netherlands. Secondly, we find that cognitive skills and, especially, educational attainment provide individuals significant protection against downward mobility. These findings are mainly in line with the meritocratic perspective. Thirdly, the results reveal that the role of the presumed explanations of downward mobility has not changed over time. Keywords Downward mobilitySocial stratificationThe NetherlandsOccupational statusissue-copyright-statement© Springer Science+Business Media Dordrecht 2016 ==== Body Introduction At the beginning of the twentieth century, most Western countries were experiencing a transition from agrarian to (post)industrialized society (Breen 2004). As a result, labour markets modernized, people experienced more prosperity, and meritocratic values became more widespread. Central theme in, for example, social democratic ideology is the idea that class systems in modern society are open and occupational positions are filled in on the basis of merit, like educational attainment, instead of social origin. To attain this ideal, governments improved access to (higher) education (Waslander and Bosman 1997). The increased enrolment in education, in combination with flourishing economies and a great(er) demand for high skilled labour have given more opportunities for people to experience upward social mobility. Echoing these societal trends, social stratification sociologists became interested in research on the openness of societies (Blau and Duncan 1967; Breen 2004; Erikson and Goldthorpe 1992; Lipset and Bendix 1959). In the Netherlands, for example, research revealed that social origin is nowadays to a lesser extent associated with an individual’s educational attainment and occupational achievement than it used to be (de Graaf and Luijkx 1995; Ganzeboom and Luijkx 2004). In addition, although findings are inconsistent regarding whether education has become more important in predicting persons’ occupational position over time (de Graaf and Luijkx 1995; Ganzeboom and Luijkx 2004; Tolsma and Wolbers 2014), education is generally found to be the most important predictor of an individual’s occupational position nowadays (de Graaf and Ultee 1998). Hence, these findings suggest that a process from ascription to achievement has taken place in Dutch society. The question is whether there are limits to this ongoing trend to societal ‘openness’ and upward mobility in the Netherlands? As a result of the increased educational participation of individuals, the educational level of the Dutch (working) population has risen considerably. However, because the upgrading of the occupational structure did not keep pace with the enormous educational expansion that took place, the demand–supply ratio for high skilled work has decreased, resulting into a process of ‘credential inflation’ in the Netherlands (Wolbers et al. 2001). As a consequence, high skilled workers are forced to accept jobs at a lower level than they actually obtained in education to avoid (long-term) unemployment. High educated workers displace middle educated workers, and eventually, middle educated workers push out low educated workers of employment. In other words, a process of ‘crowding out’ has taken place in recent years (Wolbers et al. 2001; Gesthuizen and Wolbers 2010). When this process is lasting, this could lead to a development in which an increasing number of individuals, despite their higher level of education attained, take an occupational position which is lower than that of their parents who attained their position in more advantageous labour market conditions. In other words, (young) people are more likely to become intergenerational downward occupational mobile nowadays. It is important to make a distinction here between absolute and relative social mobility. Absolute mobility refers to the question how many individuals are in higher or lower social positions than their parents. Relative mobility rates describe the association in social class position between parents and their children. To determine this association adequately, changes in the occupational structure (between the two generations) need to be taken into account. Structural changes can have a strong impact on the total amount of mobility in society. Previous research showed that large part of the total mobility rate in the Netherlands can be ascribed to the increased demand for high skilled labour (Tolsma and Wolbers 2010). Both forms of social mobility occur at the same time. In this article, we focus on absolute mobility rates and on absolute downward occupational mobility in particular. Lately, several researchers indeed observed that recent cohorts of individuals are more likely to be downward mobile. In the United Kingdom, the number of upward mobile persons is decreasing, whereas the number of downward mobile persons is increasing over time (Goldthorpe and Jackson 2007; Li and Devine 2011). These trends were also noticed for France (Peugny 2007). For the Netherlands, Tolsma and Wolbers (2010) found that the likelihood of downward mobility has increased for the youngest (male) birth cohorts. Particularly, men are more likely to experience downward educational and occupational mobility, whereas similar findings were not found for women. These findings make clear that downward social mobility is (will be) more present in (future) Western societies than before. Previous social stratification research, however, has paid little attention to test (individual) explanations of downward mobility (for exceptions, see Alm 2011; Peugny 2007). This article tries to fill in this knowledge gap by testing three theoretical perspectives that aim to explain intergenerational downward occupational mobility. So, our focus is rather modest: we do not intend to give reasons why downward mobility is increasing over time (this could be the next step for future research), but first to investigate the determinants of downward mobility. In addition, we study whether the importance of these determinants has changed over time. Therefore, our research questions are as follows: Which determinants underlie intergenerational downward occupational mobility of individuals in the Netherlands? And: To what extent have effects of the underlying determinants of intergenerational downward occupational mobility in the Netherlands changed over time? The study of intergenerational downward mobility is relevant as downward mobility may have negative consequences for an individual’s well-being. First of all, in the past scholars argued that both upward and downward mobility lead to feelings of anomy (Breed 1963; Lipset and Bendix 1959).1 In addition, downward mobility is an important cause of status anxiety (Layte 2012; Wilkinson 1999). Furthermore, earlier research demonstrated that downward mobility negatively affects social trust, subjective well-being and mental health (Tolsma and Wolbers 2010; Dolan and Lordan 2013; Tooth and Mishra 2013). So, in analysing the determinants of downward mobility, we obtain some evidence on how (lacking) social mobility can play an intermediating role between relevant background characteristics of individuals and their well-being. Theoretical Background Determinants of Downward Mobility Although social stratification research has paid little attention to explanations of downward mobility, three theoretical perspectives have been developed, namely: the social casualty perspective, the meritocratic perspective and the parental resources approach. The social casualty perspective (Richardson 1977) assumes that downward mobility is mainly the result of fate. Individuals who face physical or mental problems are to a lesser extent able to attain a similar social position as their parents than healthy persons. This also applies for individuals who are addicted to drugs or alcohol, or have experienced familial disorganization in their youth. Several findings provide support for this perspective. Firstly, children of divorced parents were found to perform worse in education and are more likely to be unemployed in later life (McLanahan and Sandefur 1994). Presumably, they are therefore also more likely to be downward mobile. Secondly, Alm (2008) observed a significant association between registered drug use and downward mobility for men in Sweden. Thirdly, it appears that persons with a mental disease like schizophrenia perform worse in the labour market, and, as a consequence, are more likely to experience downward mobility (Timms 1998). Hence, we hypothesize: (H1) Individuals with (psychical or mental) health problems are more likely to be intergenerational downward mobile. And: (H2) Individuals who experienced familial disorganization in early life are more likely to be intergenerational downward mobile. The second perspective, the meritocratic perspective (Blau and Duncan 1967; Miller 1960; Saunders 1997), is based on the idea that societies have become more open in recent decades. According to this perspective, occupational status attainment is not associated (anymore) with ascribed characteristics, like social origin, but instead, it is based on an individual’s ability and motivation (Blau and Duncan 1967). As a result, people with more ability and motivation will attain higher occupational positions. To explain this process, human capital theory provides relevant insights (Becker 1964). Human capital represents someone’s knowledge and skills and therefore his or her labour market productivity. According to this micro-economic theory, individuals who possess more human capital are considered to be more attractive employees to be hired and will therefore find themselves in better labour market positions. Besides cognitive skills, educational attainment is an important recruitment indicator for employers to signal someone’s ability and motivation (Arrow 1973). When individuals do not have the cognitive skills or educational credentials required to maintain their parents’ occupational position, they have higher risks of being downward mobile. For Great Britain, for example, several scholars discovered that people’s ability and motivation are important predictors of their opportunities for social mobility (Deary et al. 2005; Saunders 1997). Moreover, a meta-analysis of Strenze (2007) revealed that intelligence is a powerful predictor of later educational and occupational success. Dutch studies, in addition, showed that educational attainment is the most important determinant of a person’s occupational position (de Graaf and Luijkx 1995; Ganzeboom and Luijkx 2004). As a result, we expect that an individual’s cognitive skills and educational attainment provide a reduction of the risk of downward mobility. Therefore, we expect: (H3) Individuals with more cognitive skills are less likely to be intergenerational downward mobile. And: (H4) Individuals with higher educational attainment are less likely to be intergenerational downward mobile. The third perspective, the parental resources approach, is based on the cultural reproduction theory of Bourdieu (1973). This theory argues that parents attempt to transmit their position to their children to avoid downward mobility by using various forms of (parental) resources. First of all, parents in high social classes possess more economic resources. These economic resources give children opportunities to attend better schools and provide them better educational means. Consequently, children will have better opportunities to obtain higher qualifications and, hence, better occupational positions (Blanden and Gregg 2004; de Graaf et al. 2000). With respect to downward mobility, we therefore also expect: (H5) Individuals with more parental economic resources are less likely to be intergenerational downward mobile. Furthermore, children raised up in high social classes possess more cultural resources. Cultural resources enhance people’s chances of educational and occupational success and may in turn reduce the risk of downward mobility (Bourdieu 1973; Scherger and Savage 2010). There are several reasons for this claim. First of all, children are brought up in an environment that develops a greater taste to learn abstract and intellectual concepts. In addition, parents show more interest in children’s schooling performances and children receive more encouragement to attain higher educational credentials (de Graaf et al. 2000; de Graaf and Kalmijn 2001). Moreover, some scholars suggest that children with more parental cultural capital are favoured by teachers and employers, because they are familiar with the habits and customs of the elites (Dronkers and de Graaf 1995; Rivera 2012). Consequently, hypothesis 6 reads: (H6) Individuals with more parental cultural resources are less likely to be intergenerational downward mobile. Lastly, the possession of parental social resources might reduce the risk of downward mobility. In their search for a job, children can use the contacts and the social network of their parents (Lin et al. 1981). As a result, children with more parental social resources might find better jobs than children with less parental social resources. In the Netherlands, research demonstrated that parental voluntary association membership is a significant predictor of intergenerational transmission of social status (van Houten et al. 2013). Moreover, close contacts between parents and teacher may stimulate the educational career of children and, in turn, their future career prospects (Dronkers and de Graaf 1995; McLean 1999). Hence, we hypothesize: (H7) Individuals with more parental social resources are less likely to be intergenerational downward mobile. Changes in Effects over Time Several studies found that Dutch society has become more open in the sense that a process from ascription to achievement has taken place (de Graaf and Luijkx 1995; Ganzeboom and Luijkx 2004). Over time, ascribed characteristics have lost importance, whilst achieved characteristics, like education, have become increasingly important in the intergenerational transmission of occupational status (Blau and Duncan 1967). On the one hand, the upgrading of the occupational structure has led to replacement of lower skilled work by more technological advanced occupations for which employees require more skills, and education and training. As a result, the demand for high skilled workers has increased at the expense of the demand for the less educated. On the other hand, over time social norms have changed in such a way that it has become unacceptable to assess people on their social background (Parson 1951). Hence, we expect that parental resources have lost importance in explaining downward mobility, whereas achieved characteristics have become more important. Therefore our final hypotheses read: (H8) Individuals raised in younger birth cohorts profited less from parental resources in preventing intergenerational downward mobility than individuals raised in older birth cohorts. And: (H9) Individuals raised in younger birth cohorts profited more from cognitive skills and educational attainment in preventing intergenerational downward mobility than individuals raised in older birth cohorts. Data and Measurement Data To test the predictive validity of these theoretical perspectives we used data from the Family Survey Dutch Population 2009 (Kraaykamp et al. 2009). The survey contains retrospective information on the life course and life situation of persons in the Netherlands. The target population was the total Dutch speaking population aged 18–70 in 2009. Interviews were conducted with 3269 persons. The survey consisted of two different parts: an oral questionnaire and a web/written questionnaire. In total, 2802 respondents completed both parts. The final response rate was 42.3 %.2 We performed several restrictions for our data analysis. First, to make a genuine cohort comparison—in order to ascertain trends in downward mobility—we account for confounding age or life course effects. In doing so, we selected only respondents who were 35 years or older and who had a job at the moment of the interview. Indeed, prior research showed that most people reach occupational maturity around the age of 35 (Goldthorpe 1980; Wolbers et al. 2011). As a result of this selection, we lose a large number of respondents, particularly women. Finally, we only include respondents with valid scores on all variables. Our final, analytical sample consists of 1423 respondents. Dependent Variable The measurement of intergenerational downward mobility is based on the status of the respondent’s occupation at age 35 and the occupational status of the father when the respondent was 15 years old. The status scale is based on the International Socio-Economic Index of Occupational Status (ISEI) (Ganzeboom et al. 1992). Downward mobility is constructed by dividing the occupational status of the respondent by that of the father. The obtained scores are classified into three groups: intergenerational downward mobile persons (score < 0.9), intergenerational stable persons (0.9 <= score <= 1.1) and intergenerational upward mobile persons (score > 1.1). Table 1 presents the percentage of persons who experienced downward mobility, stability, or upward mobility for successive birth cohorts. It appears that the percentage of downward mobile persons increased. In cohort 1925–1944 28.8 % is downward mobile, and in the youngest cohort (1965–1974) 32.5 %. The percentage of intergenerational stable or upward mobile persons, however, decreased. In the oldest birth cohort 1925–1944 24.4 % is intergenerational stable and 46.8 % is upward mobile, whereas in cohort 1965–1974 these percentages appear to be slightly lower.Table 1 Developments in intergenerational occupational mobility by birth cohort and gender Cohort N Downward mobile (%) Stable (%) Upward mobile (%) Total 1925–1944 156 28.8 24.4 46.8 1945–1954 557 30.5 24.8 44.7 1955–1964 548 32.1 22.6 45.3 1965–1974 510 32.5 23.3 44.1 Total 1771 31.5 23.7 44.1 Men 1925–1944 126 25.4 25.4 49.2 1945–1954 395 26.8 27.6 45.6 1955–1964 325 28.3 22.2 49.5 1965–1974 277 31.0 26.4 42.6 Total 1123 28.1 25.5 46.4 Women 1925–1944 30 43.3 20.0 36.7 1945–1954 162 39.5 17.9 42.6 1955–1964 223 37.7 23.3 39.0 1965–1974 233 34.3 19.7 45.9 Total 648 37.2 20.5 42.3 Source: Family Survey Dutch Population 2009 (own calculations) In addition to general trends, we also analysed these trends for men and women separately (see Table 1). In line with earlier findings (Tolsma and Wolbers 2010), we observe that especially men are more likely to be downward mobile over time. In the oldest cohort for men 25.4 % experienced downward mobility, whereas this percentage is 31.0 % in the youngest cohort. Likewise, the percentage of intergenerational stable persons slightly increased. By contrast, the proportion of upward mobile men decreased. In the oldest cohort 49.2 % experienced upward mobility, in the youngest cohort this is only 42.6 %. Our results among women differ considerably from those among men. Table 1 reveals that the proportion of women experiencing downward mobility has decreased over time. In the oldest cohort 43.3 % is downward mobile, in the youngest cohort this percentage is only 34.3 %. In addition, the percentage of women that experienced no social mobility lies around 20 %. Finally, we notice that women in more recent birth cohorts are more likely to experience upward mobility. This percentage increased from 36.7 % in cohort 1925–1944 to 45.9 % in cohort 1965–1974. Presumably, these deviating findings for women may result from their sharply increased educational attainment and, in turn, increasing levels of labour market participation. Independent Variables We study the social casualty perspective by using four characteristics mainly preceding the moment of labour market entry. Firstly, we use a measure for health, since health problems may hinder later labour market success. On a scale that varied between 1 (‘very bad’) and 5 (‘very good’), people indicated how healthy they were at the age of 20. Several studies showed that subjective measurements of health are good predictors for overall (both mental and psychical) health status (Idler and Benyamini 1997; Jylhä 2009). Secondly, various forms of family disorganization may influence the likelihood of downward mobility. We consider the effects of parental divorce and premature death of at least one of the parents. In doing so, we constructed two dummy variables. One captures whether or not a respondent experienced a parental divorce before the age of 18 (0 = no parental divorce, 1 = parental divorce), the other whether or not a respondent lost at least one of his or her parents before the age of 18 (0 = no premature death of parent, 1 = premature death of parent). Lastly, we investigate the extent to which victimization of violence is associated with a higher likelihood of experiencing downward mobility. We distinguish between whether or not a person was victim of violence before the 18th birthday (0 = no victim, 1 = victim). The meritocratic perspective is operationalized by measurements for cognitive skills and educational attainment. Cognitive skills are measured using two questions. On a scale that varied between 1 (‘very weak’) and 5 (‘very strong’), people had to indicate whether they were good in (1) arithmetic and in (2) language at primary school. We combined these two measures to create a Likert Scale for cognitive skills (Spearmen-Brown = 0.535). Higher scores indicate more cognitive skills. Level of education is measured as the number of years required to successfully complete an education level. Scores on this variable vary between 6 (primary education) and 21 (post-academic education). We have no direct information on the financial resources of the parents. Therefore, we base our measurement of parental economic resources on items about the presence of five luxury goods (car, microwave oven, bathroom, freezer and dishwasher) in the parental home when the respondent was 15 years old.3 We created a Likert Scale whereby higher scores indicate more economic resources (α = 0.662). Cultural resources are measured by questions about parental highbrow cultural participation (e.g. visiting museums and theatres), parental reading preferences (e.g. reading novels) and parental reading socialization practices when the respondent was 15 years old. Based on these items, we created a Likert Scale (α = 0.871) whereby higher scores indicate more cultural resources. Finally, social resources are measured by questions about parents’ participation rate in formal voluntary organizations like political parties, schools, religious organizations and (other) leisure organizations (among others, Rotary club, Lion’s, sport clubs). Based on these four items a sum scale (α = 0.298) is constructed. Higher scores indicate more social resources. Control Variables In the analysis we take several control variables into account. First of all, we include the occupational status of the father to control for floor and ceiling effects. In addition, we control for the age of the father when the child was 15 year old to correct for life course effects of the father. Finally, we add birth cohort (in years, centred around 1957), ethnicity (0 = native, 1 = non-native), and gender (0 = female, 1 = man) as covariates in the multivariate analysis. Descriptive statistics of the independent variables are displayed in Table 2.Table 2 Descriptive statistics of independent variables (N = 1423) Variable Minimum Maximum Mean SD Birth year 1928.00 1974.00 1957.45 9.49 Occupational status father 24.00 86.00 45.10 14.33 Age father (at age 15 respondent) 30.00 77.00 47.18 6.31 Non-native 0.00 1.00 0.09 Male 0.00 1.00 0.63 Health at age 20 3.00 5.00 4.55 0.57 Parental divorce before the age of 18 0.00 1.00 0.04 Death of a parent before the age of 18 0.00 1.00 0.06 Victim of violence before the age of 18 0.00 1.00 0.06 Cognitive skills 1.00 5.00 3.55 0.84 Level of education 6.00 21.00 13.08 3.39 Parental economic resources 0.00 1.00 0.34 0.27 Parental cultural resources 1.00 2.79 1.57 0.36 Parental social resources 0.00 4.00 1.02 0.99 Source: Family Survey Dutch Population 2009 (own calculations) Results To test the determinants of intergenerational downward social mobility, we used multinomial logistic regression analysis, as our dependent variable consists of three categories (downward mobile, stable, and upward mobile persons). In the models (see Table 3 for the results) we estimated the probability of being downward mobile versus the probability of being stable. The probability of being upward mobile (versus being stable) was simultaneously estimated, but is not shown in Table 3. For interpretation purposes, we present for each independent variable the marginal effect at the mean (MEM). This marginal effect reflects the change in the probability of being downward mobile when the independent variable increases by one unit, with the other independent variables fixed on their mean values. In total, five models were estimated in which we all take into account the set of control variables. In Model 1 we include the variables of the social casualty perspective. In Model 2 we study the effects of cognitive skills and educational attainment. The effects of parental economic, cultural, and social resources are investigated in Model 3. In Model 4 we simultaneously estimate the effects of all independent variables. Finally, in Model 5 we add the interaction effects between the independent variables and birth year. To ascertain that multicollinearity does not bias our results, we first included these interaction terms separately. The effects found do not differ from the results presented here.Table 3 Results of multinomial logistic regression analysis of intergenerational downward occupational mobility: downward mobile versus stable (N = 1423) Variable Model 1 Model 2 Model 3 Model 4 Model 5 MEM SE MEM SE MEM SE MEM SE MEM SE Birth year (1957 = 0) −0.114 0.070 −0.014 0.071 −0.008 0.081 0.048 0.083 −0.197 0.288 Occupational status father 1.148** 0.069 1.442** 0.078 1.129** 0.077 1.506** 0.084 1.514** 0.084 Age father (at age 15 respondent) 0.100 0.108 0.090 0.105 0.103 0.105 0.082 0.108 0.072 0.108 Non-native −0.010 0.048 −0.009 0.049 −0.001 0.048 0.000 0.050 −0.002 0.050 Male −0.112** 0.029 −0.107** 0.029 −0.115** 0.029 −0.107** 0.029 −0.112** 0.029 Health at age 20 −0.048 0.048 −0.043 0.049 −0.074 0.164 Parental divorce before the age of 18 0.021 0.068 −0.012 0.071 −0.073 0.325 Death of a parent before the age of 18 0.074 0.058 0.018 0.060 −0.203* 0.196 Victim of violence before the age of 18 0.071 0.056 0.060 0.059 −0.055* 0.247 Cognitive skills −0.278** 0.069 −0.262** 0.070 −0.591 0.231 Level of education −0.638** 0.070 −0.598** 0.072 −0.531 0.216 Parental economic resources −0.087 0.060 −0.057 0.061 −0.126 0.188 Parental cultural resources −0.314** 0.080 −0.130 0.083 0.050 0.274 Parental social resources −0.039 0.061 −0.012 0.062 0.045 0.197 Health at age 20 * birth year 0.052 0.244 Parental divorce before the age of 18 * birth year 0.086 0.412 Death of a parent before the age of 18 * birth year 0.382 0.318 Victim of violence before the age of 18 * birth year 0.162 0.333 Cognitive skills * birth year 0.525 0.344 Level of education level * birth year −0.127 0.343 Parental economic resources * birth year 0.107 0.263 Parental cultural resources * birth year −0.274 0.411 Parental social resources * birth year −0.089 0.302 Change in Chi2 541.422** 693.495** 565.006** 714.356** 724.960** Degrees of freedom 18 14 16 28 46 Source: Family Survey Dutch Population 2009 (own calculations) * p < 0.05; ** p < 0.01 (2-sided) The results in Model 1 indicate no significant effects of the variables of the social casualty perspective. As a result, H1 and H2 do not find empirical support. In Model 2 we examine the influence of the meritocratic perspective. This model, first of all, reveals that the probability of being downward mobile is smaller in case individuals have more cognitive skills. Likewise, it appears that as people are higher educated, they are less likely to experience downward mobility. Hence, these results support H3 and H4. The parental resource approach assumes that individuals with more parental economic, cultural and social resources are less likely to be downward mobile. However, we observe that the likelihood of being downward mobile is only lower for individuals with more cultural resources (see Model 3). These findings are, therefore, in line with H6. In Model 4 we simultaneously analyse the effects of the three theoretical perspectives. Particularly, these results provide strong support for the meritocratic perspective: higher educational credentials and more cognitive skills reduce the risk of downward mobility. Similarly, we notice that the effect of parental cultural resources disappears (that is, becomes indirect) when including the predictors of the meritocratic perspective. Finally, in Model 5 we do not find significant interaction effects between birth year and the predictors of the three theoretical perspectives on the probability to be downward mobile. Hence, these results do not support H8 and H9 that predicted changes over time in the effect of the characteristics of the meritocratic perspective and the parental resources approach. To check the robustness of our results, we specified several alternative models. Firstly, we examined whether we can replicate our results when we do not account for floor and ceiling effects. After excluding occupational status of the father from the model, all significant effects disappear (or become positive for the variables of the parental resources approach). These results imply that in order to get adequate estimates, it is important to account for occupational status of the father. Secondly, we run separate models for individuals whose father had a high, middle, or low status job (classification of the first and the latter group are based on one standard deviation above or below the mean) to study whether the results are different for individuals of various social origins. These alternative findings do not substantially differ from the ones presented in Table 3 and, likewise, indicate that educational attainment is the most important predictor of downward social mobility. Thirdly, we tried to replicate our results by using alternative measures for occupational status: the EGP class scheme (Erikson et al. 1979) and the Dutch Ultee and Sixma occupational prestige scale (Sixma and Ultee 1983). Once again we find that our main results are robust: educational attainment is the most important determinant of intergenerational downward mobility. However, the effect of cognitive skills is weaker. Finally, we performed our analyses for men and women separately. For men the results are highly similar to the findings presented in Table 3. Most important predictor is educational attainment, followed by cognitive skills. Furthermore, these results demonstrate that a better health is negatively associated with downward mobility. This effect disappears though, when including other variables in the model. For women, we find that the loss of a parent is positively associated with the risk of intergenerational downward mobility. Yet, this effect also disappears after inclusion of other predictor variables. Notably, also in these analyses for women our results reveal significant negative effects for educational attainment and cognitive skills on the likelihood of being downward mobile. Lastly, it appears that parental economic resources protect women against downward mobility. Conclusion and Discussion In the 1990s already, a few Dutch sociologists predicted that rates of intergenerational downward mobility would considerably increase in the Netherlands, but also in other Western societies (Dronkers 1994; Waslander and Bosman 1997). Observations that changes in the occupational structure did not keep pace with the educational expansion provided support for this view. As a result of ‘credential inflation’ and ‘crowding-out’ in the labour market, more and more individuals are forced to take lower social positions than their parents. Recently, various studies indeed found that individuals from younger birth cohorts are more likely to experience downward mobility than those from older cohorts (Goldthorpe and Jackson 2007; Li and Devine 2011; Peugny 2007; Tolsma and Wolbers 2010). An important follow-up question then is to investigate whether and why some individuals are more likely of being downward mobile than others. This article attempted to shed more light on this issue by studying (changes in) the determinants of intergenerational downward occupational mobility in the Netherlands. We derived insights from three theoretical perspectives to explain why some individuals are more likely to be downward mobile: the social casualty perspective, the meritocratic perspective, and the parental resources approach. The first perspective argues that downward mobility is the result of ‘fate’. Health problems or familial disorganization in youth are expected to lead to higher probability of experiencing downward mobility. The meritocratic perspective and the parental resources approach posit rivalling hypotheses: whereas the former assumes that downward mobility is the result of selection on achievement, the latter maintains that social origin has still an important impact on someone’s future career prospects and, in turn, on someone’s risk of downward social mobility. The empirical results provide little support for the social casualty perspective. Possibly, the lack of empirical support for this perspective stems from the fact that these events are quite rare or that more extreme cases felt outside our sample population or tend to have higher non-response rates. Accordingly, our results could be an underestimation of the true importance of the social casualty perspective. Similarly, the parental resources approach is not clearly supported by our findings. Although we do find that parental cultural resources provide some protection against downward mobility, this effect largely becomes indirect when we take into account other characteristics of individuals. The meritocratic perspective, by contrast, is more important in explaining downward mobility for individuals. As individuals are higher educated or have more cognitive skills, they are less likely to be downward mobile. These results are much in line with previous results found in France (Peugny 2007) and Sweden (Alm 2011). However, in contrast to our expectations, we did not find that the meritocratic perspective has gained more importance over time. Thus, despite the fact that Dutch society has become more open and achieved characteristics have become more important than ascribed characteristics in attaining high social positions (de Graaf and Luijkx 1995), education and cognitive skills have not become more important in preventing intergenerational downward occupational mobility. Consequently, one of our main conclusions is that downward mobility is mainly the result of the meritocratic principle selection on effort and ability. Accordingly, our results may imply that despite the expected increase in the number of downward mobile individuals, this trend will not affect the efficiency of the labour system. Yet, because studies have shown that children from higher social strata enter secondary education with a knowledge advantage (Kloosterman et al. 2011) they might avert the risk of downward mobility in earlier life phases so that inequalities may still be persistent over time. In addition, the results clearly demonstrate that, despite processes of ‘credential inflation’ and ‘crowding-out’ in the labour market, individual investment in education remains profitable: indeed, these investments reduce the personal risk for intergenerational downward mobility. Thus, from a general point of view, the findings indicate that educational credentials are not so much of importance in absolute, but especially in relative terms. Another interpretation of the results might be that the emergence of a homogenous underclass of lower educated has become ever more likely (Gesthuizen and Kraaykamp 2002). Because talented persons from lower social backgrounds have profited from the educational expansion, those left behind, including downward mobile persons, are the ones with least cognitive skills (and qualifications). And, as a result of this development, especially this vulnerable group, but also upcoming generations of this group, may hold more marginalised positions with all negative consequences (from anomy to mental health problems) associated (Wilson 1987). Although this article provides new insights into the individual determinants of downward mobility, some aspects deserve more scholarly attention in future research. First of all, in this article we focused on individual explanations of intergenerational downward occupational mobility. However, the effects of structural changes, such as cyclical fluctuations of the economy, could be further investigated. In light of the gender differences found, scholars may also investigate to what extent the risen labour market participation of women has had an effect on the risk of downward mobility of men, due to increased direct competition among scarce jobs between men and women. In addition, although studies in several countries showed that rates of downward mobility are rising, comparative research is still lacking. To fully understand the causes of downward mobility, cross-national research is therefore required. Finally, in this article we investigated intergenerational downward occupational mobility as a result of a two-generation process. However, as proposed by the status consolidation theory of Richardson (1977), a three-generation approach might be more appropriate. Specifically, this theory argues that parents, who experienced upward social mobility in the past without a comparable increase in their educational attainment, cannot provide their children enough parental resources to maintain their social position. As a result, their children will end up in the same social positions as their grandparents did, thereby consolidating social status across three generations. To stringently test this hypothesis, social stratification and mobility research with a focus on three generations is necessary (Mare 2011). In the Netherlands, a first step into this direction has been made recently (Wolbers and Ultee 2013). 1 Recent empirical evidence does not strongly support this claim anymore with regard to upward mobility (Clark and D’Angelo 2013). 2 Given that interviews were time-consuming (approximately 45–60 min) and both the respondent and his or her partner were interviewed, the response rate can be considered as relatively high (and very similar to response rates for earlier rounds of the Family Survey Dutch Population). Moreover, the response rate did not lead to any known response bias (Kraaykamp et al. 2009). 3 We acknowledge that this measure of parental economic resources does not fully capture the financial capacities of parents, especially for people in younger birth cohorts. Yet it is important to keep in mind that we investigate respondents who are at least 35 years old. Accordingly, the situation asked refers to the possession of multiple “luxury goods” of parents before 1990. Nevertheless, the validity of this measure over time may still be doubtful, but a better measure of parental economic resources is simply not available in the data. Anyway, excluding parental economic resources from our models, did not lead to different results. ==== Refs References Alm S Social nedatrorlighet mellan generationer—en provning av tre perspective 2008 Stockholm Institutet for Framtidsstudier, arbetsrapportserien Alm S Downward mobility across generations: The role of parental mobility and education Sociological Research Online 2011 16 1 14 10.5153/sro.2416 Arrow K Higher education as a filter Journal of Public Economics 1973 2 193 216 10.1016/0047-2727(73)90013-3 Becker G Human capital: A theoretical and empirical analysis, with a special reference to education 1964 New York National Bureau of Economic Research Blanden J Gregg P Family income and educational attainment: A review of approaches and evidence for Britain Oxford Review of Economic Policy 2004 20 245 263 10.1093/oxrep/grh014 Blau P Duncan D The American occupational structure 1967 New York Wiley Bourdieu P Brown R Cultural reproduction and social reproduction Knowledge, education and social change 1973 London Tavistock 71 112 Breed W Occupational mobility and suicide among white males American Sociological Review 1963 28 179 188 10.2307/2090605 Breen R Social mobility in Europe 2004 Oxford Oxford University Press Clark, A. E., & D’Angelo, E. (2013). Upward social mobility, well-being and political preferences: Evidence from BHPS. Retrieved from http://www.parisschoolofeconomics.com/clark-andrew/dangelo_clark_2013.pdf de Graaf PM Kalmijn M Trends in the intergenerational transmission of cultural and economic status Acta Sociologica 2001 44 51 66 10.1177/000169930104400105 de Graaf PM Luijkx R Dronkers J Ultee WC Paden naar succes: geboorte of diploma’s? Verschuivende ongelijkheid in Nederland 1995 Assen Van Gorcum 31 45 de Graaf PM Ultee WC Shavit Y Müller W Education and early occupation in the Netherlands around 1990: Categorical and continuous scales and the details of a relationship From school to work: A comparative study of educational qualifications and occupational destinations 1998 Oxford Clarendon Press 337 368 de Graaf ND de Graaf PM Kraaykamp G Parental cultural capital and educational attainment in the Netherlands: A refinement of the cultural capital perspective Sociology of Education 2000 73 92 111 10.2307/2673239 Deary IJ Taylor MD Hart CL Wilson V Smith GD Blane D Starr JM Intergenerational social mobility and mid-life status attainment: Influences of childhood intelligence, childhood social factors, and education Intelligence 2005 33 455 472 10.1016/j.intell.2005.06.003 Dolan, P., & Lordan, G. (2013). Moving up and sliding down: An empirical assessment of the effect of social mobility on subjective wellbeing. Retrieved from http://cep.lse.ac.uk/pubs/download/dp1190.pdf Dronkers J Onderwijs als voorportaal van sociale daling De Helling 1994 7 26 30 Dronkers J de Graaf PM Dronkers J Ultee WC Ouders en het onderwijs van hun kinderen Verschuivende ongelijkheid in Nederland 1995 Assen Van Gorcum 31 45 Erikson R Goldthorpe JH The constant flux: A study of class mobility in industrial societies 1992 Oxford Oxford University Press Erikson R Goldthorpe JH Portocarero L Intergenerational class mobility in three Western European societies The British Journal of Sociology 1979 30 415 441 10.2307/589632 Ganzeboom HBG de Graaf PM Treiman DJ A standard international socio-economic index of occupational status Social Science Research 1992 21 1 56 10.1016/0049-089X(92)90017-B Ganzeboom HBG Luijkx R Breen R Recent trends in occupational class reproduction in the Netherlands 1970–99 Social mobility in Europe 2004 Oxford Oxford University Press 345 381 Gesthuizen M Kraaykamp G Ontwikkelingen in verbale capaciteiten van laagopgeleiden in Nederland. De keerzijde van de onderwijsexpansie Mens & Maatschappij 2002 77 361 380 Gesthuizen M Wolbers MHJ Employment transitions in the Netherlands, 1980–2004: Are low educated men subject to structural or cyclical crowding out? Research in Social Stratification and Mobility 2010 28 437 451 10.1016/j.rssm.2010.06.001 Goldthorpe J Social mobility and class structure in modern Britain 1980 Oxford Clarendon Press Goldthorpe JH Jackson M Intergenerational class mobility in contemporary Britain: Political concerns and empirical findings The British Journal of Sociology 2007 58 525 546 10.1111/j.1468-4446.2007.00165.x 18076385 Idler EL Benyamini Y Self-rated health and mortality: A review of twenty-seven community studies Journal of Health and Social Behavior 1997 38 21 37 10.2307/2955359 9097506 Jylhä M What is self-rated health and why does it predict mortality? Towards a unified conceptual model Social Science and Medicine 2009 69 307 316 10.1016/j.socscimed.2009.05.013 19520474 Kloosterman R Notten N Tolsma J Kraaykamp G The effects of parental reading socialization and early school involvement on children’s academic performance: A panel study of primary school pupils in the Netherlands European Sociological Review 2011 27 291 306 10.1093/esr/jcq007 Kraaykamp, G., Ruiter, S., & Wolbers, M. H. J. (2009). Familie-enquête Nederlandse Bevolking 2009 [Family Survey Dutch Population] (dataset). Department of Sociology, Radboud Universiteit, Nijmegen (producer). The Hague: DANS (distributor). Layte R The association between income inequality and mental health: Testing status anxiety, social capital, and neo-materialist explanations European Sociological Review 2012 28 498 511 10.1093/esr/jcr012 Li Y Devine F Is social mobility really declining? Intergenerational class mobility in Britain in the 1990s and the 2000s Sociological Research Online 2011 16 1 15 10.5153/sro.2424 Lin N Ensel WM Vaughn JC Social resources and strength of ties: Structural factors in occupational status attainment American Sociological Review 1981 46 393 405 10.2307/2095260 Lipset SE Bendix R Social mobility in industrial society 1959 Berkeley University of California Press Mare RD A multigenerational view of inequality Demography 2011 48 1 23 10.1007/s13524-011-0014-7 21271318 McLanahan S Sandefur G Growing up with a single parent: What hurts, what helps 1994 Cambridge, MA Harvard University Press McLean RB Parental involvement as social capital: Differential effectiveness on science achievement, truancy, and dropping out Social Forces 1999 78 117 144 10.1093/sf/78.1.117 Miller SM Comparative social mobility Current Sociology 1960 9 1 80 10.1177/001139216000900101 Parson T The social system 1951 Glencoe Free Press Peugny, C. (2007). La mobilité sociale descendante. L´épreuve du déclassement. Retrieved from http://www.sciencespo.fr/osc/sites/sciencespo.fr.osc/files/These_peugny_0.pdf Richardson CJ The problem of downward mobility The British Journal of Sociology 1977 28 303 320 10.2307/589996 Rivera LA Hiring as cultural matching: The case of elite professional service firms American Sociological Review 2012 77 999 1022 10.1177/0003122412463213 Saunders P Social mobility in Britain: An empirical evaluation of two competing explanations Sociology 1997 31 261 288 10.1177/0038038597031002005 Scherger S Savage M Cultural transmission, educational attainment and social mobility The Sociological Review 2010 58 406 428 10.1111/j.1467-954X.2010.01927.x Sixma H Ultee WC Een beroepsprestigeschaal voor Nederland in de jaren tachtig Mens & Maatschappij 1983 58 360 382 Strenze T Intelligence and socioeconomic success: A meta-analytic review of longitudinal research Intelligence 2007 35 401 426 10.1016/j.intell.2006.09.004 Timms D Gender, social mobility and psychiatric diagnoses Social Sciences and Medicine 1998 46 1235 1247 10.1016/S0277-9536(97)10052-1 Tolsma J Wolbers MHJ Onderwijs als nieuwe sociale scheidslijn? De gevolgen van onderwijsexpansie voor sociale mobiliteit, de waarde van diploma’s en het relatieve belang van opleiding in Nederland Tijdschrift voor Sociologie 2010 31 239 259 Tolsma J Wolbers MHJ Social origin and occupational success at labour market entry in the Netherlands, 1931–1980 Acta Sociologica 2014 57 253 269 10.1177/0001699314533807 Tooth L Mishra G Intergenerational educational mobility on general mental health and depressive symptoms in young women Quality of Life Research 2013 22 1589 1602 10.1007/s11136-012-0310-8 23138380 van Houten JMA Gesthuizen M Wolbers MHJ Intergenerational transmission of occupational status: The role of voluntary association membership as an emerging compensatory strategy of reproductions Research in Social Stratification and Mobility 2013 33 13 26 10.1016/j.rssm.2013.04.002 Waslander S Bosman R Sociale daling en onderwijs Beleid & Maatschappij 1997 24 74 84 Wilkinson RG Health, hierarchy, and social anxiety Annals of the New York Academy of Sciences 1999 896 48 63 10.1111/j.1749-6632.1999.tb08104.x 10681887 Wilson WJ The truly disadvantaged: The inner city, the underclass and public policy 1987 Chicago University of Chicago Press Wolbers MHJ de Graaf PM Ultee WC Trends in the occupational returns to educational credentials in the Dutch labor market: Changes in structures and in the association? Acta Sociologica 2001 44 5 19 10.1177/000169930104400102 Wolbers MHJ Luijkx R Ultee WC Educational attainment, occupational achievement, career peaks: The Netherlands in the second part of the 20th century European Societies 2011 13 425 450 10.1080/14616696.2011.568265 Wolbers MHJ Ultee WC Driegeneraties- en tweeoudersvragen over het opleidingsniveau van mannen en vrouwen in Nederland Mens & Maatschappij 2013 88 276 299 10.5117/MEM2013.3.WOLB

PMC005xxxxxx/PMC5002040.txt

==== Front Int J Public HealthInt J Public HealthInternational Journal of Public Health1661-85561661-8564Springer International Publishing Cham 84210.1007/s00038-016-0842-8ReviewEpidemiology and economic burden of measles, mumps, pertussis, and varicella in Germany: a systematic review Damm Oliver +49 521/1064679oliver.damm@uni-bielefeld.de 1Witte Julian 1Wetzka Stefanie 2Prosser Christine 3Braun Sebastian 3Welte Robert 2Greiner Wolfgang 11 Department of Health Economics and Health Care Management, School of Public Health, Bielefeld University, Universitätsstraße 25, 33615 Bielefeld, Germany 2 GlaxoSmithKline Germany, Prinzregentenplatz 9, 81675 Munich, Germany 3 Xcenda GmbH, Lange Laube 31, 30159 Hanover, Germany 4 8 2016 4 8 2016 2016 61 7 847 860 23 12 2015 21 4 2016 2 6 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Objectives Despite the availability of vaccines and the existence of public vaccination recommendations, outbreaks of vaccine-preventable childhood diseases still cause public health debate. The objective of this systematic review was to provide an overview of the current epidemiology and economic burden of measles, mumps, pertussis, and varicella in Germany. Methods We systematically reviewed studies published since 2000. The literature search was conducted using PubMed and EMBASE. Also, we used German notification data to give an up-to-date overview of the epidemiology of the four diseases under consideration. Results Thirty-six studies were included in our review. Results suggest that there is still considerable morbidity due to childhood diseases in Germany. Studies providing cost estimates are scarce. Comparative analyses of different data sources (notification data vs. claims data) revealed a potential underestimation of incidence estimates when using notification data. Furthermore, several studies showed regional differences in incidence of some of the diseases under consideration. Conclusions Our findings underline the need for improved vaccination and communication strategies targeting all susceptible age and risk groups on a national and local level. Electronic supplementary material The online version of this article (doi:10.1007/s00038-016-0842-8) contains supplementary material, which is available to authorized users. Keywords Childhood diseasesEpidemiologyEconomic burdenGermanyGlaxoSmithKline Biologicals SAissue-copyright-statement© Swiss School of Public Health (SSPH+) 2016 ==== Body Introduction Vaccination is regarded as one of the great public health achievements (CDC 1999) and has led to substantial decreases in morbidity and mortality of vaccine-preventable diseases (Roush and Murphy 2007). However, despite the availability of vaccines and the existence of public vaccination recommendations, vaccine-preventable childhood diseases are still a subject of public health debate and research. This is mostly due to outbreaks such as the large measles outbreak in Berlin in 2014/2015 (RKI 2015b). In Germany, current vaccination recommendations of the Standing Vaccination Committee (STIKO) cover, among others, routine childhood vaccination against measles, mumps, pertussis, and varicella (Table 1). Some of these recommendations have existed for decades and undergone several updates. A detailed description of the history of vaccination recommendations in Germany was published by Klein et al. (2012).Table 1 Measles, mumps, pertussis, and varicella vaccination recommendations and mandatory reporting in Germany (Klein et al. 2012; RKI 2014) Disease Introduction of routine childhood vaccination recommendation Current childhood immunisation schedule Current adult immunisation schedule Mandatory reporting since Measles FRG: 1974; GDR: 1966 (voluntary), 1970 (mandatory) 11–14 months and 15–23 months (catch-up until 17 years) One-time vaccination for all adults born after 1970 who are of unclear vaccination status, are unvaccinated, or have received only one vaccination in childhood 2001 Mumps FRG: 1976; GDR: 1977 (voluntary) 11–14 months and 15–23 months (catch-up until 17 years) No routine immunisation 2013 Pertussis FRG: 1969/1991a; GDR: 1964 (mandatory) 2–4 months and 11–14 months (catch-up until 4 years); booster 5-6 and 9-17 years The next due tetanus and diphtheria vaccination as a one-time tetanus, diphtheria and pertussis combination vaccination 2013 Varicella 2004 11–14 months and 15–23 months (catch-up until 17 years) No routine immunisation 2013 FRG Federal Republic of Germany, GDR German Democratic Republic, RKI Robert Koch Institute aReintroduction in 1991 after the recommendation was suspended in 1974 In many countries, the introduction of vaccination recommendations is accompanied by the implementation of infectious disease surveillance systems. The primary aim of such surveillance systems is to collect and analyse infectious disease data on an ongoing basis. These data can be used to detect outbreaks and to evaluate the impact of interventions. Surveillance data can be collected through various methods including passive surveillance, active surveillance, and sentinel surveillance (Declich and Carter 1994; MacDonald 2012; Oleckno 2008; Roush 2011):Passive surveillance mainly refers to national notification systems based on mandatory case reporting. This means that health care providers (e.g. physicians and hospitals) and laboratories are required by law to routinely report the occurrence of certain infectious diseases to public health officials. Passive surveillance is the most common method for collecting data on vaccine-preventable diseases. It captures the entire population and requires relatively few resources, but completeness of reporting is highly dependent on the compliance of health facilities. Moreover, diagnostic accuracy can differ among health care providers, particularly since case definition criteria of notification systems often include not only laboratory-confirmed cases but also clinically diagnosed cases. Active surveillance involves a proactive search for cases by health authorities through contacting health care providers on a regular basis. This data collection procedure can provide a more complete picture of disease frequency, but it is usually more costly than passive surveillance. That is why active surveillance is often limited to outbreaks or other short-term investigations. Sentinel surveillance, which may comprise elements of both active and passive surveillance approaches, relies on a limited number of carefully selected reporting sites. Selection criteria may include representativeness, geographic area, and practical considerations related to feasibility and reporting quality (e.g. willingness to participate, well-qualified staff, and adequate technical resources). Sentinel surveillance requires fewer resources than population-based surveillance and can provide high-quality data. The main shortcoming of this approach is that the generalisability of the findings may be limited. Further data collection procedures include surveillance surveys, epidemic field investigations and the use of secondary data sets (Declich and Carter 1994). In Germany, measles has been a notifiable disease since 2001. Mumps, pertussis, and varicella became notifiable on a national level in March 2013 (Table 1). German sentinel systems were, among others, developed for the surveillance of measles (Siedler and Leitmeyer 2004) and varicella (Siedler and Arndt 2010). In 2010, the 53 member states of the World Health Organization (WHO) European Region renewed their commitment to the elimination of measles by 2015 (WHO 2010). Nevertheless, the outbreak in Berlin in 2014/2015 (RKI 2015b) clearly demonstrates that Germany has not met the elimination target so far. Therefore, the objective of this study is to systematically review the existing literature on the current epidemiology and economic burden of measles and other childhood diseases (namely mumps, pertussis, and varicella) in Germany. Epidemiological measures of interest include incidence, frequency of complications and long-term sequelae, mortality, outbreak descriptions, and disability-adjusted life years (DALYs). The DALY is a summary measure of population health that captures the burden of morbidity and mortality in a single metric. DALYs for a particular disease are calculated as the sum of the time lost due to premature death and the time lived with poor health or disability. The concept includes weighting of disease duration with a weight factor that reflects the severity of the health condition (Murray 1994). Since the development of this measure in the early 1990s, the methods used to calculate DALYs have undergone several changes (Devleesschauwer et al. 2014; Murray et al. 2012; Voigt and King 2014). The DALY approach is primarily used by the WHO to quantify the global burden of disease. Our definition of economic burden covers resource consumption (particularly hospitalisation), illness-related work days lost, and direct and indirect costs. Direct costs primarily capture the cost of medical care (e.g. drug therapy, physician consultations, inpatient treatment), whereas indirect costs mainly refer to productivity losses due to absence from work and premature death. Methods Literature search We searched in the literature databases PubMed and EMBASE for relevant papers published between 1 January 2000 and 8 February 2015. Search terms included controlled vocabulary and free-text terms. Details of the search strategy are provided in the supplementary material (Online Resource 1). Two investigators (OD and JW) independently screened search results and assessed the eligibility of potentially relevant studies according to predefined inclusion and exclusion criteria. Discrepancies were solved through discussion involving a third investigator (SW). Reference lists of identified studies were searched manually for further relevant publications. We followed the PRISMA guidelines for conducting and reporting systematic reviews (Moher et al. 2009). Inclusion and exclusion criteria We included English and German language articles reporting data on the epidemiology and economic burden of measles, mumps, pertussis, and varicella in Germany. We did not apply restrictions concerning the type of study and the method of data collection. We wanted to focus as far as possible on the general population and, therefore, excluded studies restricted to health care workers, military personnel and their families. We also excluded pure review articles, comments, letters, editorials, single case-reports, small case series and outbreak reports with fewer than 10 subjects, articles without full-text (e.g. conference abstracts), and surveillance and outbreak reports lacking a separate methods section. Data extraction Data extraction was performed by one investigator (OD) and verified by a second (JW). The following data were abstracted from the included studies: type of study, data sources and methods, population and setting, time frame, outcome measures, and results. Surveillance data In addition to our literature search, we used notification data of the German Protection against Infection Act (Infektionsschutzgesetz) to give an up-to-date overview on the epidemiology of measles, mumps, pertussis, and varicella in Germany. The numbers of notified cases were extracted using a web-based data query tool (SurvStat@RKI 2.0). Results Literature review After duplicates were removed, the bibliographic search using PubMed and EMBASE databases yielded 1700 records. After screening of titles and abstracts, 1568 records were excluded and 132 full-text articles were subsequently assessed for eligibility. Ninety-six articles were excluded after full-text assessment. Main reasons for exclusion were study objective (e.g. vaccination coverage or vaccine effectiveness), publication type (e.g. review article, case report, or comment), or lack of a sufficient methods section. Thirty-six studies fulfilled the inclusion criteria and were, therefore, included in this review. Figure 1 shows the process of study identification and selection.Fig. 1 Flowchart of the study identification and selection process Of the 36 included studies, 18 reported results for measles, two for mumps, six for pertussis, and ten for varicella. Most of the studies were mainly based on data from (mandatory) notification and/or sentinel surveillance systems. Four of these studies used data from a paediatric hospital-based surveillance system, called ESPED (Erhebungseinheit für seltene pädiatrische Erkrankungen in Deutschland; German paediatric surveillance unit). Three studies used claims data of the Associations of Statutory Health Insurance Physicians (ASHIP). Two studies explicitly applied (decision-analytic) modelling approaches. Other methods and data sources used include survey and interview methods, contact tracing activities, and review of medical files or national statistics. The majority of the included studies (n = 34) provided data on epidemiological outcomes and/or hospitalisation. Only few studies reported on economic aspects in terms of costs and/or work days lost (n = 5). Table 2 summarises the characteristics and results of the included studies. The main results are also described in the following paragraphs.Table 2 Study characteristics and results Publication Methods (study type, data source, population, and time frame) Outcome measures Results Measles  Arenz et al. (2009) Analysis of surveillance data; paediatric hospital-based surveillance data (ESPED); hospitalised children and adolescents <16 years (detailed questionnaire-based information was obtained for 96 children); 2006 Children hospitalised for measles 115 children (42 % were <2 years) Median length of hospital stay 6 days Complications Pneumonia: 54 %; otitis media: 11 %; seizures: 7 % Measles-related deaths 2 children died of measles (with encephalitis)  Carabin et al. (2003) Multi-country cost study; country-specific incidence and cost data; direct costs include physicians’ visits, prescription medication, hospitalisation, and long-term care for sequelae; general population in Germany; 2001 Average annual costs (2001 values) of caring for measles per capita in Germany from a health care provider perspective Approximately EUR 0.02 per capita  Gillesberg Lassen et al. (2014) Outbreak report; notification data and clinical data collected through interviews with case-patients; community members, students of an anthroposophic school, and family members and friends of the students in Berlin; April–July 2011 Cases of measles 73 cases (27 % of all case-patients and 38 % of community case-patients were ≥20 years) Hospitalisation 15 %  Hegasy et al. (2012) Outbreak report; notification data and contact tracing activities; non-Roma inhabitants and Roma community members living in Hamburg; December 2008–June 2009 Cases of measles 216 cases (69 % were confirmed by laboratory analyses); a local Roma community comprised more than 50 % of the notified cases  Mankertz et al. (2011) Study on spread of the D4-Hamburg strain; results of laboratory samples; general population in Germany; 2008–2010 Cases of measles 216 cases in Hamburg; 72 cases in Lower Saxony; 48 cases in Munich; several cases occurred in Roma community members and asylum seekers Hospitalisation 40 % of patients (due to pneumonia or otitis media)  Mette et al. (2011) Claims data analysis and comparison with surveillance data; ASHIP billing data and notification data; 15.4 million people covered by statutory health insurance living in North Rhine-Westphalia; 2006–2007 Confirmed measles diagnoses (ASHIP data) 2534 diagnoses (87 % of billed measles diagnoses occurred in children <10 years) Reported cases of measles (notification data) 2014 cases (40 % of measles cases were reported for children <10 years) Ratio of confirmed measles diagnoses and reported cases of measles 1.26: 1 (underreporting)  Muscat et al. (2009) Multi-country analysis of surveillance data; data of national mandatory notification systems; general population in Germany; 2006–2007 Cases of measles 2307 cases in 2006; 571 cases in 2007 Incidence per 100,000 inhabitants 2.8 per 100,000 in 2006; 0.7 per 100,000 in 2007 Measles-related deaths 2 fatal cases in children (caused by encephalitis)  Muscat et al. (2014) Multi-country analysis of surveillance data; data submitted by national surveillance institutions to the WHO Regional Office for Europe; general population in Germany; 2012–2013 Cases of measles 167 cases in 2012; 1773 cases in 2013 Incidence per million inhabitants per year 2 per million in 2012; 21.4 per million in 2013  Plass et al. (2014) Burden of disease study; DALY estimates are based on a model of the natural history of disease using notification data; general population in Germany; 2005–2007 Average loss of DALYs per year 740 DALYs (93 % was due to acute symptomatic infections and 7 % was due to long-term sequelae) Average loss of DALYs per case 0.26 DALYs  Roggendorf et al. (2012) Outbreak report; surveillance data of a community health centre; 1st outbreak: children attending a free progressive school in Essen and their contacts; 2nd outbreak: children and adults in a low socio-economic setting and with migration background in Essen; March–July 2010 Cases of measles 1st outbreak: 75 cases; 2nd outbreak: 11 cases Hospitalisation 15 %  Siedler et al. (2006) Outbreak report; surveillance data of the mandatory reporting system and data collected through interviews of the local health authorities with physicians and family members; general population in Hesse and Bavaria; January–May 2005 (Hesse) and March–July 2005 (Bavaria) Cases of measles (without sporadic cases) Hesse: 223 cases; Bavaria: 279 cases; 74 % in school aged children Incidence per 100,000 inhabitants Hesse: 14 per 100,000; Bavaria: 12 per 100,000 Hospitalisation in patients ≥20 years Outbreak in Hesse: 34 % Measles-related deaths Outbreak in Hesse: 1 case  Siedler et al. (2013a) Analysis of surveillance data; sentinel data collected by 1488 paediatric and primary care practices; patients of practices participating in the sentinel system; 2001–2010 Cases of measles 3100 cases (2495 cases in children <10 years) Complications 15 % (mostly otitis media and pneumonia)  Takla et al. (2014) Claims data analysis and comparison with surveillance data; ASHIP billing data and mandatory notification data; 68 % (2007) and 79 % (2008–2011) of the total population living in Germany (ASHIP data); 2007–2011 Cases of measles (notification data) 4440 cases Annual incidence per million population (notification data) Total: 10.8 per million (range 6.9–19.6 per year); northern Germany: 8.7; western Germany: 7.2; eastern Germany: 5.5; southern Germany: 20.3 Annual incidence per million residents with statutory health insurance (ASHIP data) 27.5 per million; incidence based on ASHIP data was up to 4.8-fold higher than incidence based on notified cases  Tischer et al. (2001) Analysis of surveillance data; sentinel data and notification data; general population in Germany; October 1999–March 2001 Cases of measles (sentinel data) 1291 cases Complications (sentinel data) 24 % Hospitalisation (sentinel data) 2.2 % Incidence per 100,000 inhabitants 38.9 per 100,000 (sentinel data); <0.5–5.7 per 100,000 (notification data of the first quarter of 2001)  Tischer et al. (2002) Analysis of surveillance data; sentinel data collected by 1271 paediatric and primary care practices; general population in Germany; October 1999–September 2001 Cases of measles 1720 cases Incidence per 100,000 inhabitants 20 per 100,000 (range <1–56 per 100,000, depending on the federal state) Complications 16 % (mostly otitis media and pneumonia) Hospitalisation 2.4 %  Wadl et al. (2011) Outbreak report; surveillance data and data collected through questionnaires; general population in four Bavarian counties (including attendees of an anthroposophic school in Austria); March–July 2008 Cases of measles 217 cases Incidence per 100,000 population 32 per 100,000 Hospitalisation 11 % Complications 18 %  Wichmann et al. (2007) Retrospective cohort study on the initial phase of an outbreak; data collected through questionnaires; 1098 students aged 10–21 years of a public day school in Duisburg; January–May 2006 Cases of measles 53 cases Hospitalisation 4 % Complications Otitis media: 4 cases; pneumonia: 1 case; encephalitis: 1 case  Wichmann et al. (2009) Study on outbreak-related morbidity and costs; surveillance data and data collected through questionnaires/interviews (face-to-face or by telephone); health care provider costs (including physician consultations, laboratory tests, antibiotic treatment, and hospitalisation) are calculated using DRGs, the outpatient fee schedule and medication prices; general population in Duisburg; 2006 Cases of measles 614 cases in Duisburg Hospitalisation 15 % Antibiotic treatment 32 % Complications Otitis media: 19 %; pneumonia: 7 %; encephalitis: 0.6 % Measles-related deaths 2 fatal cases in children (caused by encephalitis) School days missed 2854 days Work days lost 301 days Average costs (2006 values) EUR 373 per measles patient; EUR 1877 per hospitalised patient Mumps  Otto et al. (2010) Outbreak report; laboratory samples and clinical data; adolescents and young adults in Bavaria; July–October 2010 Laboratory-confirmed infections 115 laboratory-confirmed mumps infections (median age: 24.5 years, predominantly male patients) Complications 1 case of meningitis and 21 cases of orchitis  Takla et al. (2013) Claims data analysis and comparison with surveillance data; ASHIP billing data and notification data; statutory health-insured population and general population in the eastern and western federal states of Germany; 2007–2011 Countrywide mean annual incidence per 100,000 people covered by statutory health insurance 10.3 per 100,000 (range 9.3–11.8); incidence was significantly higher in western than in eastern federal states; comparison of notification data with ASHIP data indicated severe underreporting of incidence estimates based on notification data Complications Orchitis: 6.2 % of male cases; meningitis: 0.4 %; pancreatitis: 0.3 %; encephalitis: 0.2 %; proportion of complications in cases ≥15 years was significantly higher than in cases <15 years Pertussis  Hellenbrand et al. (2009) Analysis of surveillance data; surveillance data (notification and sentinel data) and hospital discharge statistics; general population in Germany; 2000–2007 Incidence per 100,000 inhabitants in eastern federal states (notification data) 20.5 per 100,000 in 2000; 39.3 per 100,000 in 2007 Incidence in adults per 100,000 inhabitants 160–169 per 100,000 in 2002–2004 Hospitalisation in eastern federal states (surveillance data) 1.9–4.9 % in 2002–2007 (depending on the federal state) Hospitalisation (cases per 100,000 population, hospital discharge statistics) 1.7 per 100,000 in eastern federal states in 2007; 1.5 per 100,000 in western federal states in 2005; most cases occurred in children <1 year  Juretzko et al. (2001) Analysis of surveillance data; paediatric hospital-based surveillance data (ESPED) and clinical data collected through questionnaires; children <16 years; 1997–1998 Standardised incidence of pertussis requiring hospitalisation per 100,000 person-years 2.70 per 100,000; 2.36 per 100,000 in western federal states; 4.50 per 100,000 in eastern federal states Mean length of hospital stay 14.9 days Complications in hospitalised children All: 44 % (60 % occurred in children <6 months); pneumonia: 28.1 %; apnea: 20.6 %; seizures: 2.5 %; encephalopathy: 2.1 % Pertussis-related deaths in hospitalised children 0.3 %  Liese et al. (2003) Prospective long-term surveillance study; follow-up data of a population-based case–control efficacy study; 11,087 children (3–8 years) of the original study population and all other children of the same age group presenting in the participating paediatric practices; 8.3 % were not vaccinated against pertussis; May 1997–March 1999 Cases of pertussis 180 cases; 64 % were caused by B pertussis infections and 36 % were caused by B parapertussis infections Incidence per 1000 person-years B pertussis infections: 4.8 per 1000; B parapertussis infections: 2.8 per 1000  Riffelmann et al. (2006) Prospective incidence and cost study; laboratory samples, clinical and resource consumption data collected through questionnaires; direct costs include physician consultations, laboratory tests, and medication; indirect costs are based on the number of work days lost and a cost of EUR 114.30 per day; 971 primary care patients having cough for ≥7 days in two German cities (Krefeld and Rostock); economic analysis is based on 45 cases of pertussis; 2001–2004 Proportion of patients with pertussis 10 % Incidence per 100,000 inhabitants 165 per 100,000 Pertussis patients with antibiotic prescription 53 % Average direct costs (2004 values) per case EUR 120 Average indirect costs (2004 values) per case in employed patients EUR 2443  Sin et al. (2009) Outbreak report; active case finding by the local health authorities and a retrospective cohort study performed in 4 affected schools (questionnaires); mostly children and adolescents attending schools in Ludwigslust district, Mecklenburg-Western Pomerania; 2005–2006 Cases of pertussis 104 cases Attack rate 1.9–32.0 % (depending on the time since last vaccine dose); results suggest that vaccine-induced immunity begins to wane approximately 5 years after completion of the primary series  Stojanov et al. (2000) Prospective surveillance study; data of a case–control efficacy study; 11,016 children <2 years presenting with cough ≤7 days at 63 paediatric practices in Germany; March 1993–May 1995 Proportion of patients with pertussis 6.6 % Hospitalisation 4.5 % Complications in hospitalised pertussis patients All: 48 %; bradycardia: 21 %; apnea: 12 %; conjunctivitis: 12 %; pneumonia: 6 %; otitis media: 6 % Mean length of hospital stay 8 days Varicella  Banz et al. (2004) Cost study; cost estimates are based on a decision-analytic model using survey data (Wagenpfeil et al. 2004); direct costs include physician consultations, medication, and hospitalisation; transfer payments are based on parental work days lost; indirect costs are based on the number of work days lost and a cost of EUR 150 per day; general population in Germany; 1999 prices (pre-vaccination era) Annual third-party payer costs (1999 values of direct costs and transfer payments to parents caring for their sick children) EUR 78 million (direct medical costs: 43 %) Annual societal costs (1999 values of direct and indirect costs) EUR 187.5 million (direct medical costs: 18 %)  Grote et al. (2007) Analysis of surveillance data; paediatric hospital-based surveillance data (ESPED); paediatric population up to the age of 17 years; 2003–2004 (pre-vaccination era) Cases of varicella-associated deaths 10 cases (none was vaccinated against varicella) Annual mortality rate (cases per million children) 0.4 per million  Liese et al. (2008) Analysis of surveillance data; nationwide paediatric hospital-based surveillance data (ESPED), practice sentinel network and hospital diagnosis data in one federal state (North Rhine-Westphalia); children ≤16 years; 2003–2004 (pre-vaccination era) Hospitalised varicella cases (ESPED data) 918 cases (with a median age of 3.3 years) Annual incidence of varicella-related hospitalisations (cases per 100,000 children) 3.25 per 100,000 (ESPED data); 14.1 per 100,000 (capture–recapture methodology for two sources); 19.7 per 100,000 (hospital diagnosis data) Median length of hospital stay 5 days Complications in hospitalised cases All: 79.6 %; neurological: 25.4 %; skin: 23.2 %; gastrointestinal tract: 15 %; lower respiratory tract: 11.8 %; severe systemic bacterial infections: 4.4 % Cases of varicella-associated deaths 10 cases  Siedler and Arndt (2010) Analysis of surveillance data; sentinel data collected by paediatricians and general practitioners; general population in Germany; April 2005–March 2009 (vaccination era) Cases of varicella and trend analysis 83,181 cases; sentinel data showed a reduction of 55 % of varicella cases in all ages over time  Siedler et al. (2013b) Analysis of surveillance data; sentinel data, notification data, and hospital diagnosis statistics; general population in Germany; 2005–2012 (vaccination era) Trend analysis of varicella-related morbidity Significant decline of varicella incidence, complications, and hospitalisations over time  Siedler and Dettmann (2014) Analysis of hospitalisation data; national hospital discharge statistics; general population in Germany; 1995–2012 (pre-vaccination era and vaccination era) Trend analysis of varicella hospitalisation incidence No particular trend until 2003, hospitalisation incidence peaked in 2004 (time of vaccine recommendation), and decreased thereafter; hospitalisation incidence per 100,000 was significantly lower in the vaccination period (1.86) than in the pre-vaccination period (3.30)  Spackova et al. (2010) Analysis of surveillance data; sentinel data collected by paediatricians and general practitioners; general population in Germany; April 2005–March 2009 (vaccination era) Cases of varicella and trend analysis 83,075 cases; the total number of varicella cases decreased over time with increasing vaccine uptake Complications All: 0,34 %; bacterial superinfection: 0.13 %; otitis media: 0.06 %; neurological: 0.03 % Hospitalisation in varicella patients with complications 25 %  Streng et al. (2013) Analysis of surveillance and survey data; data of the Bavarian Varicella Surveillance Project (BaVariPro) based on parent surveys (vaccination coverage), paediatric practice surveillance, and paediatric hospital database queries; children <17 years in Munich; October 2006–September 2011 (vaccination era) Cases of varicella and trend analysis 16,054 cases; the mean number of cases decreased by 67 % during the five seasons Incidence per 1000 children (based on the number of reported cases) 26 per 1000 in the first season and 6 per 1000 in the fifth season Complications in practice patients 0.8 % (mostly skin complications and otitis media) Complications in hospitalised patients Central nervous system: 22.5 %; skin: 15.7 %; lower respiratory tract: 9 % Median length of hospital stay 3 days Hospitalisation incidence (cases per 100,000 children) and trend analysis 8.2 per 100,000; hospitalisation incidence decreased by 43 % between 2005 and 2009  Wagenpfeil et al. (2004) Retrospective epidemiological and resource consumption study; clinical data were obtained from medical files through telephone interviews with physicians; 1334 unvaccinated varicella patients of a representative and German-wide sample of paediatric and primary care practices; 1999 (pre-vaccination era) Complications All: 5.7 %; bacterial superinfection: 2.5 % (≤12 years), 1.9 % (>12 years); neurological: 0.2 % (≤12 years); pneumonia or bronchitis: 1.9 % (≤12 years), 1.9 % (>12 years); otitis media: 1.1 % (≤12 years) Hospitalisation period per diagnosed varicella case 0.1 inpatient days Work days lost per diagnosed varicella case 1.3 days  Ziebold et al. (2001) Analysis of surveillance data; paediatric hospital-based surveillance data (ESPED) and clinical data collected through questionnaires; hospitalised children ≤16 years; 1997 (pre-vaccination era) Hospitalised varicella cases 119 Complications in hospitalised varicella cases Neurological: 62.3 %; infectious: 35.2 % ASHIP Associations of Statutory Health Insurance Physicians, BaVariPro Bavarian Varicella Surveillance Project, DALY disability-adjusted life year, DRG diagnosis-related group, ESPED Erhebungseinheit für seltene pädiatrische Erkrankungen in Deutschland (German paediatric surveillance unit), WHO World Health Organization Measles Incidence estimates based on analyses of surveillance data ranged widely across studies and years (from <0.1 to 38.9 per 100,000 inhabitants per year). Outbreak-related incidence estimates ranged between 12 and 32 per 100,000 inhabitants. One study examining the epidemiology of measles from 2007 to 2011 (Takla et al. 2014) showed large geographical differences in incidence of measles with the highest incidence in southern Germany. Another study presenting results for 1999 to 2001 (Tischer et al. 2002) found that most cases of measles occurred in southern Germany and in North Rhine-Westphalia. Two studies (Mette et al. 2011; Takla et al. 2014) compared results of the analyses of surveillance data and ASHIP claims data and found that incidence estimates based on ASHIP data were higher than estimates based on notification data. Some studies only reported absolute numbers of cases instead of presenting data as incidence rates (see Table 2 for details). The proportion of patients developing any complications varied from 15 to 24 %. The most frequent complications of measles were otitis media and pneumonia. In hospitalised children, pneumonia was diagnosed in more than 50 % of the patients (Arenz et al. 2009). The proportion of measles-associated hospitalisations ranged from 2.2 to 40 % in studies examining all ages. One study found a proportion of hospitalisation of 34 % in patients ≤20 years (Siedler et al. 2006). Median length of hospital stay in children was 6 days (Arenz et al. 2009). Four studies provided information on measles-related deaths, which occurred rarely and only in children (see Table 2 for details). In a German substudy of the Burden of Communicable Diseases in Europe (BCoDE) project, the average loss of DALYs per case of measles was estimated to be 0.26 resulting in an average DALY loss per year of 740 (Plass et al. 2014). The average costs per measles patient and per hospitalisation were EUR 373 and 1877, respectively (Wichmann et al. 2009). Mumps Mean annual incidence of mumps based on claims data was estimated to be 10.3 per 100,000 people covered by statutory health insurance. Incidence was significantly higher in western federal states than in eastern federal states. A comparison between claims data and notification data indicated severe underreporting of mumps incidence in the notification surveillance system (see Table 2 for details). The main complication was orchitis affecting 6.2 % of male cases. The proportion of complications in individuals ≥15 years was higher than in younger patients (Takla et al. 2013). Information on the economic burden of mumps was not available. Pertussis In children, incidence of infections with Bordetella pertussis and Bordetella parapertussis was 4.8 and 2.8 per 1000 person-years, respectively. More than 60 % of all pertussis cases in children from 3 to 8 years were caused by Bordetella pertussis (Liese et al. 2003). Incidence of pertussis requiring hospitalisation was 2.7 per 100,000 person-years in children (Juretzko et al. 2001). In adults, incidence of pertussis ranged from 160 to 169 per 100,000 inhabitants (Hellenbrand et al. 2009). The diagnosis of pertussis could be verified in 10 % of the primary care patients having cough for ≥7 days (Riffelmann et al. 2006). In children <2 years presenting with cough for ≥7 days, pertussis was diagnosed in 6.6 % of the cases (Stojanov et al. 2000). Pertussis-associated hospitalisation rate was 1.5 and 1.7 per 100,000 population in the western and eastern federal states, respectively (Hellenbrand et al. 2009). Most of the hospitalisations occurred in children <1 year (Juretzko et al. 2001). More than 40 % of hospitalised children suffered from complications, and the mean length of hospitalisation in children varied between 8 and 14.9 days (Juretzko et al. 2001; Stojanov et al. 2000). Direct costs per case of pertussis were EUR 120 in primary care patients, and indirect costs per case were EUR 2443 in employed patients (Riffelmann et al. 2006). None of the reviewed studies focused on older adults (>60 years). Varicella Studies analysing surveillance data from the vaccination era showed a decline of varicella incidence and hospitalisations over time and with increasing vaccine uptake (Siedler and Arndt 2010; Siedler et al. 2013b; Siedler and Dettmann 2014; Spackova et al. 2010; Streng et al. 2013). The proportion of patients with complications was higher in the pre-vaccination era (5.7 %) (Wagenpfeil et al. 2004) than in the vaccination era (0.34–0.8 %) (Spackova et al. 2010; Streng et al. 2013). Annual incidence of varicella-related hospitalisations varied from 3.25 to 19.7 per 100,000 children depending on the data source used before routine childhood vaccination against varicella was implemented (Liese et al. 2008). About 80 % of these (hospitalised) children suffered from varicella-related complications, and most frequent complications were neurological and infectious complications (Liese et al. 2008; Ziebold et al. 2001). In the pre-vaccination era, annual varicella-associated mortality in children was 0.4 per million (Grote et al. 2007), and societal costs of varicella were estimated at EUR 187.5 million per year (Banz et al. 2004). However, direct medical costs accounted for only 18 % of these costs. Work days lost per diagnosed varicella case were 1.3 days (Wagenpfeil et al. 2004). Analysis of surveillance data The numbers of reported cases of measles, mumps, pertussis, and varicella over time are shown in Fig. 2. Nationwide mandatory notification of measles was introduced in 2001. Since then, the number of reported cases of measles varied from year to year. Mumps, pertussis, and varicella became officially notifiable diseases in Germany in 2013. Hence, the numbers of notified cases of these diseases strongly increased from that point of time. In 2014, about 70 % of all reported mumps and pertussis cases occurred in adults, while most of the notified varicella cases occurred in children. For the period before 2013, notification data on mumps, pertussis, and varicella were available only for federal states in the eastern part of Germany (see dotted lines in Fig. 2).Fig. 2 Numbers of notified cases of measles, mumps, pertussis, and varicella in Germany (2001–2015) based on SurvStat@RKI 2.0. Vertical dashed lines indicate the date of nationwide implementation of mandatory notification. BB Brandenburg, MV Mecklenburg-West Pomerania, SN Saxony, ST Saxony-Anhalt, TH Thuringia Discussion Our review examined available information on the epidemiology and economic burden of measles, mumps, pertussis, and varicella in Germany. In general, results differed widely by reporting year, population, and data source used. More specific key findings are discussed below. Results of the studies investigating the measles epidemiology in Germany and latest reports on the ongoing outbreak in Berlin (RKI 2015b) clearly showed that Germany has failed in achieving the aim of eliminating measles until 2015. Some of the reviewed studies found huge regional difference in measles incidence. Most cases occurred in the context of regionally limited outbreaks, and several outbreaks were linked to transmission from Roma community members, asylum seekers, or anthroposophic communities/schools. These findings are in line with a recently published analysis of the measles epidemiology in 2014/2015 by the RKI (2015b) and indicate the importance of conducting more in-depth analyses at a regional level, increasing public awareness on the benefits of immunisation, and strengthening efforts to identify and close the existing vaccination gaps. The relevance of targeting vulnerable groups such as Rome communities, anthroposophic groups, and immigrants has already been emphasised by an overview of affected groups in Europe (Muscat 2011). Comparative analyses of different data sources (notification data vs. claims data) revealed a potential underestimation of measles incidence estimates when using notification data. Hence, claims data analyses should be used more regularly to complement analyses based on surveillance systems and to provide a more comprehensive picture of the epidemiology of notifiable diseases in Germany. Compared to measles, only very few studies have dealt with the epidemiology of mumps in Germany. However, some of the findings concerning measles also hold true for mumps such as the underreporting of incidence estimates based on surveillance data. Another similarity was found between mumps and pertussis: in 2014, about 70 % of all notified cases occurred in adults. Studies from other countries have also reported an increasing incidence of pertussis in adolescents and adults (Rothstein and Edwards 2005; McGuiness et al. 2013). Moreover, there is evidence that adults are one of the major sources of pertussis in infants (Orenstein 1999; Bisgard et al. 2004) and play a crucial role in transmitting pertussis to household members (Baptista et al. 2009). The reason for the age shift of pertussis is supposed to be the combination of less boosting by natural infections and waning of vaccine-induced immunity (Nitsch-Osuch et al. 2013). One of the reviewed studies showed that pertussis was a common cause of persistent cough in adults, which is in line with published data from other countries (Rothstein and Edwards 2005). In general, it is assumed that standard surveillance systems greatly underestimate the level of pertussis (Crowcroft and Pebody 2006). Several of the included studies on varicella reported on the situation in the pre-vaccination era. Studies that are more current showed that varicella incidence and hospitalisation have decreased after the implementation of routine childhood vaccination. Similar effects have been observed after the introduction of the varicella vaccination programme in the United States (Marin et al. 2008; Baxter et al. 2014). Furthermore, the overall decline in varicella incidence and hospitalisation in the United States was not associated with a shift to older age groups (Baxter et al. 2014). Our review revealed that information on direct and indirect costs of childhood diseases in Germany is scarce. Furthermore, not all studies that collected cost data included all relevant cost components. For example, in the study by Riffelmann et al. (2006), direct costs were calculated without considering hospitalisation cost. In contrast, the study by Banz et al. (2004) considered an additional cost category when adopting a third-party payer perspective, namely transfer payments to parents that stay at home to care for their sick children. Inclusion of this cost category can substantially increase costs from the health care payer perspective, particularly when assuming that all sick children cause parental absence from work. For instance, in the study by Banz et al. (2004), the mean number of parental work days lost per sick child ranged from 0.6 to 4 depending on the course of disease, and as a consequence, the reimbursed costs of parental work days lost accounted for 57 % of the total third-party payer costs. The use of different outcome measures also hampers the comparison of economic results across studies. The study by Carabin et al. (2003), which estimated the costs of measles for 11 countries, used per capita costs (approximately EUR 0.02) as economic outcome measure, whereas the study by Wichmann et al. (2009) presented results in terms of costs per measles patient (EUR 373) and costs per hospitalised patient (EUR 1877). In summary, since only few studies have provided cost estimates so far, future research should concentrate on quantifying the economic burden of disease. A more intensified use of claims data analyses might contribute to this aim. Also, administrative data from health insurance funds might provide a good foundation to supplement surveillance data (Jones et al. 2013) and to extend existing methods of measuring underreporting of notified cases of infectious diseases (Gibbons et al. 2014). However, as current case definitions include not only laboratory-confirmed cases but also clinically diagnosed cases (RKI 2015a), notification data might also be subject to overestimation. Certainly, the same is true for health insurance claims data, since the validity of the recorded diagnoses is largely unknown. Limitations There are several limitations of our systematic review. First, since the scope of our review was limited to studies presenting results for Germany, transferability of results to other countries is limited, too. Second, due to the high heterogeneity of the included studies, an assessment of methodological quality was not performed. Third, many of the included studies were based on surveillance data that might be subject to underreporting. Fourth, case definitions varied among studies, which might partially explain differences in results. Conclusions This review aimed to provide an overview of the epidemiology and economic burden of measles, mumps, pertussis, and varicella in Germany. Most of the reviewed studies presented epidemiological outcomes. Studies providing information on economic aspects except hospitalisation were scarce. Despite the existing immunisation recommendations, results suggest that there is still considerable morbidity due to childhood diseases in Germany. However, not only children are affected. For instance, a high proportion of all pertussis cases occurs in adults. Furthermore, several studies revealed regional differences in incidence of some of the target diseases. These findings underline the need for improved vaccination and communication strategies targeting all susceptible age and risk groups on a national and local level. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary material 1 (PDF 187 kb) The authors would like to thank Roeland Van Kerckhoven (consultant for Keyrus Biopharma) for editorial assistance and public coordination on behalf of GSK Vaccines. Compliance with ethical standards Conflict of interest OD worked on Projects funded by the GSK group of companies, Sanofi Pasteur MSD and AstraZeneca Germany. JW worked on Projects funded by the GSK group of companies and Sanofi Pasteur MSD. CP and SB worked on Projects funded by the GSK group of companies. SW is employed by the GSK group of companies and reports personal fees from GlaxoSmithKline Biologicals SA. RW is employed by the GSK group of companies and reports shares. WG declares that he has no conflict of interest. Ethical approval This article does not contain any studies with human participants or animals performed by any of the authors. Funding This systematic review was fully funded by GlaxoSmithKline Biologicals SA, which also funded costs associated with the development and the publishing of the present manuscript. ==== Refs References Arenz S Fischer R Wildner M Measles outbreak in Germany: clinical presentation and outcome of children hospitalized for measles in 2006 Pediatr Infect Dis J 2009 28 1030 1032 10.1097/INF.0b013e3181aa6a29 19738510 Banz K Wagenpfeil S Neiss A Hammerschmidt T Wutzler P The burden of varicella in Germany: potential risks and economic impact Eur J Health Econ 2004 5 46 53 10.1007/s10198-003-0200-7 15452764 Baptista P Magalhães V Rodrigues L The role of adults in household outbreaks of pertussis Int J Infect Dis 2009 14 e111 e114 10.1016/j.ijid.2009.03.026 19559636 Baxter R Tran TN Ray P Lewis E Fireman B Black S Shinefield HR Coplan PM Saddier P Impact of vaccination on the epidemiology of varicella: 1995–2009 Pediatrics 2014 134 24 30 10.1542/peds.2013-4251 24913796 Bisgard KM Pascual FB Ehresmann KR Miller CA Cianfrini C Jennings CE Rebmann CA Gabel J Schauer SL Lett SM Infant pertussis. Who was the source? Pediatr Infect Dis J 2004 23 985 989 10.1097/01.inf.0000145263.37198.2b 15545851 Carabin H Edmunds WJ Glydmark M Beutels P Lévy-Bruhl D Salo H Griffiths UK The cost of measles in industrialised countries Vaccine 2003 21 4167 4177 10.1016/S0264-410X(03)00534-6 14505896 Centers for Disease Control & Prevention (CDC) Ten great public health achievements United States, 1900–1999 Morb Mortal Wkly Rep 1999 48 241 243 Crowcroft NS Pebody RG Seminar: recent developments in pertussis Lancet 2006 367 1926 1936 10.1016/S0140-6736(06)68848-X 16765762 Declich S Carter AO Public health surveillance: historical origins, methods and evaluation Bull World Health Organ 1994 72 285 304 8205649 Devleesschauwer B Havelaar AH Maertens de Noordhout C Haagsma JA Praet N Dorny N Duchateau L Torgerson PR Van Oyen H Speybroeck N Calculating disability-adjusted life years to quantify burden of disease Int J Public Health 2014 59 565 569 10.1007/s00038-014-0552-z 24752429 Gibbons CL Mangen MJ Plass D Havelaar AH Brooke RJ Kramarz P Peterson KL Stuurman AL Cassini A Fèvre EM Kretzschmar ME Burden of Communicable diseases in Europe consortium Measuring underreporting and under-ascertainment in infectious disease datasets: a comparison of methods BMC Public Health 2014 14 147 10.1186/1471-2458-14-147 24517715 Gillesberg Lassen S Schuster M Stemmler M Steinmüller A Matysiak-Klose D Mankertz A Santibanez S Wichmann O Falkenhorst G Measles outbreak spreading from the community to an anthroposophic school, Berlin, 2011 Epidemiol Infect 2014 142 789 796 10.1017/S0950268813001398 23823129 Grote V von Kries R Springer W Hammersen G Kreth HW Liese J Varicella-related deaths in children and adolescents—Germany 2003–2004 Acta Paediatr 2007 97 187 192 10.1111/j.1651-2227.2007.00595.x 18076716 Hegasy G, Kätzner K, Helle M, Mankertz A, Baumgarte S, Wille A, Fell G (2012) Description of measles D4-Hamburg outbreak in Hamburg, Germany, December 2008 to June 2009, which disproportionally affected a local Roma community. Euro Surveill 17. [pii]:20194 Hellenbrand W Beier D Jensen E Littmann M Meyer C Oppermann H Wirsing von König CH Reiter S The epidemiology of pertussis in Germany: past and present BMC Infect Dis 2009 9 22 10.1186/1471-2334-9-22 19243604 Jones SG Coulter S Conner W Using administrative medical claims data to supplement state disease registry systems for reporting zoonotic infections J Am Med Inform Assoc 2013 20 193 198 10.1136/amiajnl-2012-000948 22811492 Juretzko P Fabian-Marx T Haastert B Giani G von Kries R Wirsing von König CH Pertussis in Germany: regional differences in management and vaccination status of hospitalized cases Epidemiol Infect 2001 127 63 71 10.1017/S0950268801005593 11561976 Klein S Schöneberg I Krause G Vom Zwang zur Pockenschutzimpfung zum Nationalen Impfplan Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz 2012 55 1512 1523 10.1007/s00103-012-1539-7 23114451 Liese JG Renner C Stojanov S Belohradsky BH Munich Vaccine Study Group Clinical and epidemiological picture of B pertussis and B parapertussis infections after introduction of acellular pertussis vaccines Arch Dis Child 2003 88 684 687 10.1136/adc.88.8.684 12876162 Liese JG Grote V Rosenfeld E Fischer R Belohradsky BH Kries RV ESPED Varicella Study Group The burden of varicella complications before the introduction of routine varicella vaccination in Germany Pediatr Infect Dis J 2008 27 119 124 10.1097/INF.0b013e3181586665 18227713 MacDonald PDM Methods of field epidemiology 2012 Burlington Jones & Bartlett Learning Mankertz A Mihneva Z Gold H Baumgarte S Baillot A Helble R Roggendorf H Bosevska G Nedeljkovic J Makowa A Hutse V Holzmann H Aberle SW Cordey S Necula G Mentis A Korukluoğlu G Carr M Brown KE Hübschen JM Muller CP Mulders MN Santibanez S Spread of measles virus D4-Hamburg, Europe, 2008-2011 Emerg Infect Dis 2011 17 1396 1401 10.3201/eid1708.101994 21801615 Marin M Meissner HC Seward JF Varicella prevention in the United States: a review of successes and challenges Pediatrics 2008 122 e744 e751 10.1542/peds.2008-0567 18762511 McGuiness CB Hill J Fonseca E Hess G Hitchcock W Krishnarajah G The disease burden of pertussis in adults 50 years old and older in the United States: a retrospective study BMC Infect Dis 2013 13 32 10.1186/1471-2334-13-32 23343438 Mette A Reuss AM Feig M Kappelmayer L Siedler A Eckmanns T Poggensee G Under-reporting of measles—an evaluation based on data from North Rhine-Westphalia Dtsch Arztebl Int 2011 108 191 196 21505600 Moher D Liberati A Tetzlaff J Altman DG Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. David Moher and colleagues introduce PRISMA, an update of the QUOROM guidelines for reporting systematic reviews and meta-analyses BMJ 2009 339 332 336 10.1136/bmj.b2535 Murray CJL Quantifying the burden of disease: the technical basis for disability-adjusted life years Bull World Health Organ 1994 72 429 445 8062401 Murray CJL Ezzati M Flaxman AD Lim S Lozano R Michaud C Naghavi M Salomon JA Shibuya K Vos T Wikler D Lopez AD GBD 2010: design, definitions, and metrics Lancet 2012 380 2063 2066 10.1016/S0140-6736(12)61899-6 23245602 Muscat M Who gets measles in Europe? J Infect Dis 2011 204 353 365 10.1093/infdis/jir067 Muscat M Bang H Wohlfahrt J Gilsmann S Mølbak K Measles in Europe: An epidemiological assessment Measles in Europe: an epidemiological assessment Lancet 2009 373 383 389 10.1016/S0140-6736(08)61849-8 19131097 Muscat M Shefer A Ben Mamou M Spataru R Jankovic D Deshevoy S Butler R Pfeifer D The state of measles and rubella in the WHO European Region, 2013 Clin Microbiol Infect 2014 20 12 18 10.1111/1469-0691.12584 24520948 Nitsch-Osuch A Korzeniewski K Kuchar E Zielonka T Życińska K Wardyn K Epidemiological and immunological reasons for pertussis vaccination in adolescents and adults Respir Physiol Neurobiol 2013 187 99 103 10.1016/j.resp.2013.02.007 23419520 Oleckno WA Epidemiology: concepts and methods 2008 Long Grove Waveland Press Orenstein WA Pertussis in adults: epidemiology, signs, symptoms, and implications for vaccination Clin Infect Dis 1999 28 147 150 10.1086/515061 10028091 Otto W, Mankertz A, Santibanez S, Saygili H, Wenzel J, Jilg W, Wieland W, Borgmann S (2010) Ongoing outbreak of mumps affecting adolescents and young adults in Bavaria, Germany. Euro Surveill 15. [pii]:19748 Plass D Mangen MJ Kraemer A Pinheiro P Gilsdorf A Krause G Gibbons CL Van Lier A McDonald SA Brooke RJ Kramarz P Cassini A Kretzschmar ME The disease burden of hepatitis B, influenza, measles and salmonellosis in Germany: first results of the Burden of Communicable Diseases in Europe Study Epidemiol Infect 2014 142 2024 2035 10.1017/S0950268813003312 24480146 Riffelmann M Littmann M Hülsse C O’Brien J Wirsing von König CH Pertussis: incidence, symptoms and costs Dtsch Med Wochenschr 2006 131 2829 2834 10.1055/s-2006-957208 17160764 RKI Empfehlungen der Ständigen Impfkommission (STIKO) am Robert Koch-Institut Epidemiol Bull 2014 34 305 340 RKI (2015a) Falldefinitionen des Robert Koch-Instituts zur Übermittlung von Erkrankungs- oder Todesfällen und Nachweisen von Krankheitserregern, Ausgabe 2015. http://www.rki.de/DE/Content/Infekt/IfSG/Falldefinition/falldefinition_node.html. Accessed 4 Apr 2016 RKI Überblick über die Epidemiologie der Masern in 2014 und aktuelle Situation in 2015 in Deutschland Epidemiol Bull 2015 10 69 82 Roggendorf H Santibanez S Mankertz A van Treeck U Roggendorf M Two consecutive measles outbreaks with genotypes D8 and D4 in two mainly unvaccinated communities in Germany Med Microbiol Immunol 2012 201 349 355 10.1007/s00430-012-0240-7 22570014 Rothstein E Edwards K Health burden of pertussis in adolescents and adults Pediatr Infect Dis J 2005 24 44 47 10.1097/01.inf.0000160912.58660.87 Roush SW (2011) Chapter 19: Enhancing surveillance. In: Manual for the surveillance of vaccine-preventable diseases, Centers for Disease Control and Prevention. http://www.cdc.gov/vaccines/pubs/surv-manual/. Accessed 4 Apr 2016 Roush SW Murphy TV Historical comparisons of morbidity and mortality for vaccine-preventable diseases in the United States JAMA 2007 298 2155 2163 10.1001/jama.298.18.2155 18000199 Siedler A, Arndt U (2010) Impact of the routine varicella vaccination programme on varicella epidemiology in Germany. Euro Surveill 15. [pii]:19530 Siedler A Dettmann M Hospitalization with varicella and shingles before and after introduction of childhood varicella vaccination in Germany Hum Vaccin Immunother 2014 10 3594 3600 10.4161/hv.34426 25483695 Siedler A Leitmeyer K Die Bedeutung von Sentinels für die Implementierung und Evaluation von Impfstrategien Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz 2004 47 1136 1143 10.1007/s00103-004-0947-8 15583884 Siedler A Tischer A MankertzA Santibanez S Two outbreaks of measles in Germany 2005 Euro Surveill 2006 11 131 134 16645244 Siedler A Gruber A Mankertz A Measles surveillance in Germany: from sentinel to mandatory surveillance Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz 2013 56 1321 1328 10.1007/s00103-013-1787-1 23990096 Siedler A Hecht J Rieck T Tolksdorf K Hengel H Varicella vaccination in Germany: a provisional appraisal in the context of MMR vaccination Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz 2013 56 1313 1320 10.1007/s00103-013-1789-z 23990095 Sin MA Zenke R Rönckendorf R Littmann M Jorgensen P Hellenbrand W Pertussis outbreak in primary and secondary schools in Ludwigslust, Germany demonstrating the role of waning immunity Pediatr Infect Dis J 2009 28 242 244 10.1097/INF.0b013e31818a5d69 19209094 Spackova M Muehlen M Siedler A Complications of varicella after implementation of routine childhood varicella vaccination in Germany Pediatr Infect Dis J 2010 29 884 886 10.1097/INF.0b013e3181e2817f 20442685 Stojanov S Liese J Belohradsky B Hospitalization and complications in children under 2 years of age with Bordetella pertussis infection Infection 2000 28 106 110 10.1007/s150100050056 10782397 Streng A Grote V Carr D Hagemann C Liese JG Varicella routine vaccination and the effects on varicella epidemiology—results from the Bavarian Varicella Surveillance Project (BaVariPro), 2006–2011 BMC Infect Dis 2013 13 303 10.1186/1471-2334-13-303 23815523 Takla A, Wichmann O, Klinc C, Hautmann W, Rieck T, Koch J (2013) Mumps epidemiology in Germany 2007–2011. Euro Surveill 18. [pii]:20557 Takla A Wichmann O Rieck T Matysiak-Klose D Measles incidence and reporting trends in Germany, 2007–2011 Bull World Health Organ 2014 92 742 749 10.2471/BLT.13.135145 25378728 Tischer A Siedler A Rasch G Surveillance of measles in Germany Gesundheitswesen 2001 63 703 709 10.1055/s-2001-18413 11713702 Tischer A Siedler A Santibanez S Grüber A Rasch G Are measles still a problem in Germany? Results of a nation-wide laboratory-based sentinel Monatsschrift fur Kinderheilkunde 2002 150 1077 1086 10.1007/s00112-002-0513-0 Voigt K King NB Disability weights in the global burden of disease 2010 study; two steps forward, one step back? Bull World Health Organ 2014 92 226 228 10.2471/BLT.13.126227 24700983 Wadl M Siedler A Krämer W Haindl ME Gebrande S Krenn-Lanzl I Mankertz A Hautmann W Measles transmission from an anthroposophic community to the general population, Germany 2008 BMC Public Health 2011 11 474 10.1186/1471-2458-11-474 21676265 Wagenpfeil S Neiss A Banz K Wutzler P Empirical data on the varicella situation in Germany for vaccination decisions Clin Microbiol Infec 2004 10 425 430 10.1111/j.1469-0691.2004.00853.x 15113320 WHO (2010) Renewed commitment to elimination of measles and rubella and prevention of congenital rubella syndrome by 2015 and sustained support for polio-free status in the WHO European Region. (http://www.who.int/immunization/sage/3_Resolution_EURO_RC60_eRes12.pdf). Accessed 11 June 2015 Wichmann O Hellenbrand W Sagebiel D Santibanez S Ahlemeyer G Vogt G Siedler A van Treeck U Large measles outbreak at a German public school, 2006 Pediatr Infect Dis J 2007 26 782 786 10.1097/INF.0b013e318060aca1 17721371 Wichmann O Siedler A Sagebiel D Hellenbrand W Santibanez S Mankertz A Vogt G Uv Treeck Krause G Further efforts needed to achieve measles elimination in Germany: results of an outbreak investigation Bull World Health Organ 2009 87 108 115 10.2471/BLT.07.050187 19274362 Ziebold C von Kries R Lang R Weigl J Schmitt HJ Severe complications of varicella in previously healthy children in Germany: a 1-year survey Pediatrics 2001 108 E79 10.1542/peds.108.1.79 11694663

PMC005xxxxxx/PMC5002041.txt

==== Front Cell Mol Life SciCell. Mol. Life SciCellular and Molecular Life Sciences1420-682X1420-9071Springer International Publishing Cham 226410.1007/s00018-016-2264-4ReviewInhibitory interneurons in visual cortical plasticity van Versendaal Daniëlle 1Levelt Christiaan N. +31 20 5665359/64533c.levelt@nin.knaw.nl 121 Department of Molecular Visual Plasticity, Netherlands Institute for Neuroscience, Institute of the Royal Netherlands Academy of Arts and Sciences, Meibergdreef 47, 1105 BA Amsterdam, The Netherlands 2 Department of Molecular and Cellular Neurobiology, Center for Neurogenomics and Cognitive Research, VU University Amsterdam, de Boelelaan 1085, 1081HV Amsterdam, The Netherlands 18 5 2016 18 5 2016 2016 73 19 3677 3691 9 11 2015 30 4 2016 3 5 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.For proper maturation of the neocortex and acquisition of specific functions and skills, exposure to sensory stimuli is vital during critical periods of development when synaptic connectivity is highly malleable. To preserve reliable cortical processing, it is essential that these critical periods end after which learning becomes more conditional and active interaction with the environment becomes more important. How these age-dependent forms of plasticity are regulated has been studied extensively in the primary visual cortex. This has revealed that inhibitory innervation plays a crucial role and that a temporary decrease in inhibition is essential for plasticity to take place. Here, we discuss how different interneuron subsets regulate plasticity during different stages of cortical maturation. We propose a theory in which different interneuron subsets select the sources of neuronal input that undergo plasticity. Keywords Ocular dominance plasticityAdultPerceptual learningInhibition V1SomatostatinVasoactive intestinal peptideParvalbuminNeurogliaform cellshttp://dx.doi.org/10.13039/501100004963Seventh Framework Programme223326Levelt Christiaan N. Stiching BlindenhulpPraktijkgeneratorissue-copyright-statement© Springer International Publishing 2016 ==== Body The regulation of cortical plasticity The brain shows a tremendous ability to adapt to its ever-changing environment. The root of this adaptation is the formation and refinement of neural circuits, allowing our brains to develop, acquire knowledge, learn new skills, and recover from injuries. The way experience influences the structure and function of neuronal connections, referred to as experience-dependent plasticity, and changes during the course of our lives. During early development, passive exposure to input from the environment is important for proper maturation of the neocortex. In fact, for acquiring and retaining certain functions and skills, it is an absolute requirement that such exposure takes place during well-defined periods of development. These periods during which neural connectivity is especially malleable are called “critical periods.” Critical periods were first formally defined by Austrian biologist Konrad Lorenz who discovered that the first hours after hatching are crucial for graylag geese to bond with their mother [1]. In humans, the presence of critical periods in speech development is demonstrated by rare cases of so-called feral children who grow up isolated from human contact. Not being exposed to language vocalizations interferes with their ability to perceive and produce phonemes, the building blocks of language [2, 3]. A situation, which applies to all of us, is that if we are not exposed to the sounds of a particular language during the first years after birth, our auditory system has great difficulty distinguishing particular language-specific sounds [2, 3]. A proper development of the visual cortex also requires experience. This is exemplified by amblyopia (or “lazy eye”) in which low-quality input from one eye for an extended period of time causes its inputs to the cortex to become less effective, leading to lowered cortical acuity and reduced depth perception [4]. Its treatment, correction of the primary visual deficit in the affected eye and temporary occlusion of the dominant eye, has to occur before the age of 8 years when the critical period for this form of plasticity closes [5]. From these examples, it is clear that limited or erroneous experience during critical periods has lifelong consequences. This raises the question why critical periods close at all. Would it not be better if high levels of plasticity were retained throughout life? For several reasons, it is important for critical periods to end. First, while high plasticity levels improve function based on experience, they also cause vulnerability to deterioration of optimal function induced by incongruous inputs. Second, lower and higher brain regions are connected through feedforward and feedback connections [6, 7]. If the lower cortical areas continuously change the way they process information, the bidirectional communication with higher cortical areas would be severely hampered. Although critical periods close at a particular age, a certain level of plasticity is retained, albeit of a different nature. First, learning becomes more conditional and often requires instructions. The passive exposure to stimuli is much less efficient in driving plasticity, and active interaction with the environment involving various forms of reinforcement becomes the dominant way of learning. Second, the substrate of plasticity changes. During critical periods, feedforward connections undergo extensive changes [8], while later in life, associative inputs are the more malleable [9, 10]. How is this switch from critical period to adult forms of plasticity achieved? It has been known for quite some time now that the development of inhibitory innervation plays a crucial role in opening and closing critical periods [11–13]. More recently, evidence is accumulating, which suggests that different subsets of inhibitory interneurons regulate plasticity levels during critical periods and in adulthood [14, 15]. They may contribute to selecting different sources of neuronal input and regulate, which inputs undergo plasticity under specific circumstances. Here, we review the properties of different interneuron subsets and propose a hypothesis on how they may regulate different forms of plasticity during development of the visual cortex. Plasticity at different stages of cortical development The primary visual cortex (V1) has been used extensively to study cortical plasticity during development and adulthood. During the first weeks after mice are born, plasticity in V1 is driven by spontaneous activity originating in the thalamus and cortex [16], and later also by spontaneous retinal activity [17]. This spontaneous-activity-mediated plasticity is essential for setting up the thalamocortical and cortical circuitry of the visual system. Later on, visual input from the two eyes starts to drive plasticity and refines these circuits in V1. The most studied form of plasticity during this developmental stage is ocular dominance (OD) plasticity [18], which is important for the development of binocular vision and when misguided, can cause amblyopia. It can be induced experimentally by temporarily closing one eye. Visual responses of the two retinas are propagated to the visual system through the optic nerves [19]. These partially cross at the optical chiasm and project to the lateral geniculate nuclei (LGN) in the thalamus of both hemispheres. Visual information from LGN is relayed predominantly to layer 4 of the primary visual cortex (V1). Occlusion of one eye [monocular deprivation (MD)] during the critical period shifts the responsiveness of neurons to input from the non-deprived eye [20, 21]. This functional shift in OD is accompanied by extensive rearrangements of thalamocortical projections, with those serving the closed eye retracting and those of the open eye expanding [8, 22–24]. In addition to dendritic spines, the protrusions on excitatory neurons on which most excitatory synapses are located show structural plasticity during OD plasticity [25]. Making use of in vivo two-photon microscopy in mice in which a fraction of cortical neurons are expressing a green fluorescent protein (GFP), it was found that MD causes a rapid increase in the loss and gain of dendritic spines of layer 2/3 and 5 pyramidal neurons in V1 [25, 26]. In the first months after critical period closure, OD plasticity can still be induced but less efficiently so and does not involve rearrangement of thalamocortical projections [23, 27, 28]. While MD still increases spine turnover in pyramidal neurons in layer 5, this is no longer the case in layer 2/3 [29]. After critical period closure, additional forms of plasticity become more dominant in V1, most importantly perceptual learning. This is the improvement in the ability to detect or discriminate visual stimuli induced by repeated practice. Perceptual learning is the type of learning that allows experienced birdwatchers to spot the bird in a tree which an untrained person would overlook. It involves various visual cortical areas, including V1 [30–34], and is strongly influenced by reinforcement signals, such as reward or punishment [35]. Perceptual learning often requires the interaction of feedforward information with contextual information. Such contextual information is provided to V1 by feedback connections. These connections are, therefore, likely to be an important substrate of plasticity during perceptual learning instead of the feedforward connections that are fine-tuned during the critical period. Plasticity regulation How the critical period of OD plasticity is opened and closed and the transition to adult forms of plasticity are achieved is under intense investigation. The strong decline in structural plasticity after critical period closure suggests that cell-intrinsic mechanisms restricting structural plasticity are responsible for critical period closure. Indeed, inactivating certain signaling pathways, which inhibit structural plasticity, interferes with critical period closure [36–39]. However, inhibitory innervation has been found to be at least as important in this process and to represent a reversible and specific regulator of plasticity levels in the developing and adult cortex [40–42]. Mice deficient for one of the two isoforms of glutamatergic acid decarboxylase (GAD65), a γ-aminobutyric acid (GABA) synthesizing enzyme, have reduced GABA release and show no OD plasticity. This can be rescued by increasing inhibition pharmacologically with diazepam [11]. Furthermore, increasing the level of GABAergic transmission by benzodiazepine infusion in very young mice promotes the early onset of the critical period of OD plasticity [12]. A precocious critical period can also be induced by accelerating the maturation of inhibitory innervation. This can be achieved by genetically increasing cortical BDNF levels or enzymatic removal of polysialic acid, which is mostly associated with neural cell adhesion molecules [13, 43, 44]. The maturation of inhibitory innervation is thus an important factor in critical period onset. Further increasing inhibition during development closes the critical period [13, 44–47]. These discoveries suggest that a gradual increase in inhibition defines the critical period. However, it was recently discovered that a temporary suppression of inhibition occurs during plasticity in the visual cortex, both during the critical period [47], and adulthood [14, 48–50] and increases plasticity levels [47]. Depending on the age or behavioral state during which plasticity is induced and what the substrates of cortical plasticity are, this disinhibition involves different interneuron subsets and underlying mechanisms. To understand how this may work, knowledge on the various cortical interneuron subsets and their connectivity and functions is essential. The main interneuron subtypes GABAergic interneurons make up for only 10–20 % of the neuronal population in the cortex, yet their function is vital for shaping cortical activity. The high diversity of interneuron subsets in terms of gene expression profiles, physiological properties, and connectivity patterns is reflected in their specialized functional roles in cortical processing, such as balancing network activity, tuning width sharpening, and controlling the flow of information and synchronization at the circuit level [51, 52]. In recent years, we obtained a much better understanding of the developmental origins, genetic factors, and activity-dependent events that shape interneuron development and differentiation. In contrast to excitatory pyramidal cells, which originate from the subventricular zone lining the developing cortex, inhibitory interneurons are derived from a more distant source: the ganglionic eminences in the ventral portion of the telencephalon [53] (Fig. 1). In mice, cortical interneurons are first generated within the medial ganglionic eminence (MGE) with a peak production at around embryonic day 14 (E14), followed by the interneurons that are derived from the caudal ganglionic eminence (CGE) around E16 [54, 55]. Notably, different interneuron subtypes are generated within the MGE and CGE (Fig. 1).Fig. 1 Anatomy of the embryonic telencephalon showing the two main structures from which inhibitory interneurons are derived: the medial ganglionic eminence (MGE) and the caudal ganglionic eminence (CGE), as a 3D structure in the intact brain as well as in two sections. The MGE and CGE give rise to different interneuron subtypes: 5HT3aR expressing interneurons are derived from the CGE and PV and SST expressing interneurons are derived from the MGE. Progenitor cells tangentially migrate to the appropriate cortical area before they radially position themselves via the ventricular zone (VZ), intermediate zone (IZ) and subplate (SP) to their final laminar position in the cortical plate (CP) Around birth, postmitotic interneuron progenitors migrate tangentially to the appropriate cortical area before they migrate radially via the ventricular zone (VZ), intermediate zone (IZ), and subplate (SP) to their final laminar position in the cortical plate (CP) [56, 57], (Fig. 1). MGE-derived interneurons populate the cortical layers in an inside-out order as do pyramidal cells. CGE-derived interneurons do not follow this sequence and accumulate predominantly in the top layers [54, 57]. During the first postnatal week, the progenitor cells specify into different subclasses of interneurons during which they acquire their mature morphologies, neurochemical expression patterns, and electrical properties, and form stereotypical cortical circuits [52]. Here, we focus on four interneuron subtypes that make up for the majority of cortical interneurons: two MGE-derived subtypes that express either the Ca2+ binding protein parvalbumin (PV) or the neuropeptides somatostatin (SST), and two CGE-derived subtypes both expressing the serotonin receptor 5HT3aR together with either vasoactive intestinal peptide (VIP), or reelin [58–60] (Fig. 2).Fig. 2 Schematic representation of the main projections to and from pyramidal cells and interneurons within the six layers of the primary visual cortex (V1). Shown are rough estimates of densities (black circles) from local, thalamic (lateral geniculate nucleus (LGN) and the lateral posterior nucleus (LPN) of the thalamus), feedback and callosal projections to the different layers of V1 (left panel) and to different subtypes of interneurons (right panel). Estimates are based on the literature and Allen Mouse Brain Connectivity Atlas [158]. Layers 5 and 2/3 mainly receive local inputs, whereas layer 4 mostly receives thalamic input from the LGN. Conversely, layer 1 mostly receives thalamic input from the LPN, callosal inputs, and feedback projections. The subtypes of interneurons discussed in this article (middle right panel) express either a combination of the serotonin receptor 5HT3aR with reelin or vasoactive intestinal peptide (VIP) or are positive for parvalbumin (PV) or somatostatin (SST). Neurogliaform cells (NGF) express 5HT3aR and reelin and are indicated in green, 5HT3aR positive interneurons that express VIP are indicated in blue, chandelier and basket cells express PV and are indicated in purple, and finally, Martinotti cells that express SST are indicated in red. Both NGF cells and VIP+ interneurons are strongly responsive to nicotinergic and serotonergic neuromodulatory inputs and inputs from higher brain regions (feedback and callosal). NGF cells provide strong local inhibition through volume release of GABA mainly in the upper layers, but also in deeper layers. They inhibit all types of local excitatory and inhibitory neurons (not shown in figure). VIP interneurons mainly innervate other interneurons (SST+ and to a lesser extent PV+ interneurons). Basket cells are mainly innervated by thalamic (LGN) and local excitatory axons. They innervate the proximal dendrites and somata of pyramidal cells with a bias to layer 2/3 and layer 4. They receive inhibitory inputs from SST+ and VIP+ interneurons and other basket cells. Chandelier cells are special in the sense that they form inhibitory synapses on the axon initial segment of pyramidal cells (not shown in figure). Finally, Martinotti cells predominantly receive local inputs and preferentially form inhibitory synapses on distal dendrites and tufts of pyramidal cells PV-expressing interneurons Interneurons expressing PV are MGE derived and are the largest group of interneurons in the cortex, accounting for 40 % of the total GABAergic population [55, 58]. Of the PV-expressing interneurons, a small proportion is constituted by the chandelier cells that target the axon initial segments of principle neurons. Evidence suggests that chandelier cells depolarize or hyperpolarize principal cells depending on whether these cells are quiescent or whether their membrane potential fluctuates, as is often observed in vivo [61, 62]. The actual function of chandelier cells is not yet understood. Most PV+ interneurons are fast-spiking basket cells that predominantly innervate proximal dendrites and somata of their targets, and provide the main source of somatic inhibition [63–65]. Their cell bodies are found in all cortical layers with the exception of layer 1, and they are most numerous in layers 4 and 5 [65–68]. Most basket cells project locally, but in some cases, their axons can cross different layers [58, 66]. Basket cells receive the bulk of the thalamic input to interneurons and are the dominant interneuron subset exerting control of pyramidal cell firing [69–71] (Fig. 2). They, thus, provide strong feedforward inhibition and may gate sensory input from the thalamus. Fast-spiking basket cells also receive pooled input from local cortical neurons with different tuning properties [72, 73]. This causes them to be only weakly tuned but highly suited for regulating the dynamic range of cortical responses. This is a crucial function in highly recurrent networks. While such networks enable the cortex to selectively amplify relevant information, they carry the risk of runaway activity. Optogenetic approaches have shown that PV+ basket cells reduce the activity of cortical excitatory neurons by both thresholding and scaling their responses, thus keeping the system within its optimal dynamic range [74, 75]. PV+ basket cells are also responsible for ensuring that the timing of sensory stimuli is accurately represented in sensory systems [76]. Cortical neurons summate sensory inputs that occur within a set period of time, thus triggering a response only if they coincide. PV+-basket-cell-mediated feedforward inhibition can narrow this window of integration and effectively regulate temporal summation by rapidly hyperpolarizing the neuron after receiving synaptic input. Finally, PV+ basket cells are also believed to orchestrate oscillatory activity in the gamma range (30–80 Hz) made possible by their fast and non-adapting firing properties and their extensive interconnectivity through inhibitory synapses and gap junctions [77–80]. SST-expressing interneurons The second group of MGE-derived interneurons expresses SST and makes up for 30 % of all cortical interneurons [58]. SST+ interneurons are typically Martinotti cells. The somata of these cells are most abundant in layers 2/3 and 5 and excluded from layer 1 [65–67, 81]. SST+ cells receive excitatory input from local pyramidal cells and form most of their inhibitory synapses on the dendritic tufts in layer 1 [65, 81, 82] but also on distal dendrites of neurons in other layers [68]. The distal dendrites mostly receive horizontal connections from other pyramidal neurons situated further away within V1, while the dendritic tufts receive associative and feedback connections from many different cortical areas and thalamic association nuclei, such as pulvinar (lateral posterior nucleus in rodents) (Fig. 2). Inhibitory synapses formed by SST+ interneurons are thus perfectly situated to gate these inputs. SST+ interneurons are also involved in feature coding, i.e., the sculpting of excitatory neuron responses. A classic example is surround suppression [83]. Neurons in V1 respond most strongly when a visual stimulus of a particular size is presented. When this visual stimulus is enlarged, the neuron will respond more weakly. This results in a relative enhancement of responses to borders of visual stimuli. Surround suppression thus enhances apparent contrast and underlies visual pop-out. This suppression by stimulation of the surrounding area of the classical receptive field in mouse V1 involves suppression by SST+ interneurons with much larger receptive fields [84]. However, SST+ interneuron-mediated inhibition is certainly not the only mechanism responsible for surround suppression, as it is only reduced but not absent under anesthesia when SST+ interneurons have little influence on visual responses in V1 [84–86] or when SST+ interneurons are optogenetically silenced [84]. In contrast to PV+ interneurons, SST+ Martinotti cells do not form inhibitory synapses onto each other, but extensively innervate other interneuron subsets, including PV+ basket cells [87, 88]. A subset of SST+ interneurons in layer 4 exclusively inhibits PV+ interneurons [89]. Activity of SST+ interneurons may, therefore, not only suppress horizontal and feedback connections, but also disinhibit feedforward connections. Finally, some SST+ interneuron subsets have been identified whose functions are not yet understood, including SST+ basket cells throughout the cortex and bitufted cells in layer 2/3 [90]. VIP-expressing interneurons The third largest group of interneurons (30 %) expresses the serotonin receptor (5HT3aR); a subset of these also expresses VIP [59]. These VIP+ interneurons are typically bipolar cells that are specialized in inhibiting other interneuron subtypes, especially SST+ interneurons and to a lesser extent PV+ basket cells [87, 91–93]. In addition, VIP+ bitufted cells have been identified, which also inhibit pyramidal neurons [90]. All VIP+ interneurons are activated by cholinergic and serotonergic inputs [94, 95], but also receive long-range intercortical and thalamic inputs (Fig. 2). Suppression of SST+ interneurons by VIP+ interneurons may enhance associative/feedback excitatory inputs. At the same time, it could cause suppression of feedforward connections as suppression of SST+ interneurons disinhibits PV+ interneurons. Because VIP+ interneurons express serotonin and nicotinic acetylcholine receptors, these neuromodulators may thus contribute to switching between feedforward and feedback input to V1 and provide reinforcement signals important for perceptual learning. Neurogliaform cells A second subset of 5HT3aR expressing interneurons consists of the VIP negative, reelin positive neurogliaform (NGF) cells [59, 96]. These interneurons have characteristic spider web morphology and provide inhibition by volume transmission of GABA that acts on postsynaptic targets through the slower metabotropic GABAB- and possibly extrasynaptic GABAA receptors [97]. These postsynaptic targets include all excitatory and inhibitory neurons with dendrites in the proximity of the NGF cell. NGF cells also form gap junctions with various other interneuron subsets through which they can synchronize them and exert powerful cortical inhibition [98, 99]. NGF cells can be found mainly in layer 1, where they also express neuron-derived neurotrophic factor (NDNF) [100] but also in layer 2/3 and to a lesser extent in layer 5. NGF cells in layer 2/3 receive thalamic and local cortical input. Layer 1 NGF cells also receive strong callosal and feedback inputs and evidence suggests that they regulate the dendritic integration of feedforward and callosal/feedback inputs [101]. Interestingly, layer 2/3 NGF cells have also been found to inhibit feedforward inhibition by PV+ basket cells [102] showing that like SST+ interneurons, they may have a role in switching between feedforward and feedback inputs. Selecting the substrate of plasticity by selective disinhibition How may this diverse set of interneurons with their various functions work together in regulating plasticity levels during development and in adulthood? We propose that interneurons select the different sources of neuronal input that may be relevant for learning at particular developmental stages and/or under specific circumstances. We will argue that different interneuron subsets are involved in regulating plasticity at the different stages of development: early on, PV+ basket cells are the main regulators of plasticity, while during adulthood, SST+ interneurons, controlled by VIP+ interneurons, appear to become the more dominant regulators of cortical plasticity. Regulation of critical period plasticity by PV+ basket cells Early during development, inhibitory innervation in V1 is weak, and cortical neurons are spontaneously active with high synchronicity [103]. This cortex- and later retina-derived spontaneous activity drives the precise wiring of cortical neurons [16, 104, 105]. It was recently discovered that synapses that exhibit low synchronicity with nearby synapses are more likely to be depressed [104]. This “out-of-sync, lose-your-link” mechanism is believed to underlie the clustering of co-active synapses. The time window for optimal desynchronization-induced plasticity is very broad (1.5–2 s) and matches the duration of spontaneous waves originating in the retina [104]. Once the eyes open, visual inputs start contributing to activity in V1. Based on the precise timing of these inputs, experience-dependent plasticity will optimize cortical neuronal circuits. This improves visual processing leading to increased acuity and fine tunes binocular vision important for depth perception. Thus, plasticity mechanisms must now be adjusted to a new source of information that also has different temporal properties. The development of inhibitory synapses formed by PV+ basket cells is thought to both adjust the temporal aspects of cortical processing and suppress spontaneous activity, thus optimizing conditions for plasticity based on visual input. PV+ basket cells suppress spontaneous activity and decrease the window of spike-timing-dependent plasticity Switching to visually driven neuronal activity as the substrate of plasticity may require the active suppression of spontaneous activity in V1 [106–108] (Fig. 3). During the critical period, PV+ interneurons have been shown to decrease spontaneous activity while leaving visual response strength unchanged [85], possibly favoring visual inputs over spontaneous activity as the substrate of cortical plasticity.Fig. 3 Proposed model of plasticity substrate selection by different interneuron subsets during the critical period and in adulthood. When visual input is altered by monocular deprivation during the critical period, net inhibition provided by PV+ interneurons decreases, so that feedforward connections can undergo plastic changes (indicated by the red spot), which is sufficient for learning. In the adult visual system, perceptual learning is reinforcement dependent and may involve plasticity of feedback connections providing contextual information about the feedforward inputs that are reinforced (in this example, the black bar with the retinotopy and orientation matching the bird’s black wing). This plasticity is facilitated by reduced inhibition of SST+ interneurons that innervate the dendritic tufts. Suppression of SST+ interneuron activity is mediated through inhibition by VIP+ interneurons whose activity depends on the behavioral state of the animal Furthermore, PV+ basket cell-mediated inhibition can control the timing precision of neuronal responses. Increasing their influence reduces the time window of temporal integration and spike-timing-dependent plasticity [76, 109]. The slowly progressing rise in inhibition during the critical period may thus gradually increase the stringency of plasticity and the temporal resolution of cortical activity in V1 while at the same time suppressing spontaneous activity and weak inputs. This eventually results in a stable, well-tuned, and fast network with limited noise. Regulation of PV+-basket-cell-mediated inhibition is crucial for critical period plasticity Interestingly, PV+-basket-cell-mediated inhibition does not simply increase during the critical period, but is strongly influenced by visual input. Like excitatory neurons, they shift their ocular preference upon monocular deprivation [85, 110–112]. Furthermore, PV+ interneurons become temporarily suppressed upon a brief period of MD [47]. This rapid downregulation of PV+ interneuron activity is essential for inducing OD plasticity and disappears with critical period closure [47]. It has been suggested that plasticity of interneurons may cause selective suppression of deprived eye responses after MD [110, 113–115]. However, optogenetic reduction of PV+-, SST+-, or VIP+-interneuron-mediated inhibition after induction of OD plasticity does not cause any recovery of the OD shift, implying that such an instructive role of inhibition is improbable [85]. More likely, the temporary suppression of PV+ interneurons upon MD is essential for disinhibiting weak inputs from the open eye and widening the time window for synaptic integration. This reduction in the stringency of plasticity may help to recruit and strengthen new synaptic inputs after MD, allowing reoptimization of visual processing in V1. As mentioned earlier, critical period closure can be interfered with by inactivating specific signaling cascades involving extracellular matrix- or myelin-based factors limiting axon growth. Recent studies show that inactivating some of these signaling cascades specifically in PV+ interneurons is sufficient to interfere with critical period closure [36, 39, 116–118]. This suggests that critical period closure involves mechanisms intrinsic to PV+ interneurons that limit their potential to temporarily reduce their activity. This idea is also supported by the finding that transplantation of immature interneurons into V1 enhances plasticity in adult mice [119–121]. Taken together, the function of PV+ basket cells in regulating the dynamic range and gating feedforward inputs may contribute to selecting visually driven inputs for cortical plasticity (Fig. 3). The control of PV+ basket cells over the window of temporal integration of synaptic inputs could at the same time define the timing on which the plasticity is based. Because the responses of PV+ basket cells are adjustable during the critical period, the stringency of these plasticity rules can be altered. This allows for the rewiring of V1 connectivity based on changes in visual input as long as the critical period lasts. Plasticity during adulthood With the decline of critical period plasticity, there is an overall change in the main substrate of cortical plasticity. While during the critical period, feedforward connections, such as the thalamocortical projections, undergo extensive rearrangements, most types of plasticity that take place during adolescence and adulthood typically involve horizontal and feedback connections in V1. Their synapses are predominantly formed on distal dendrites and dendritic tufts in layer 1. These dendritic compartments are strongly innervated by SST+ interneurons and layer 1 NGF cells, which may underlie their dominant role in regulating plasticity during adulthood. Various forms of plasticity can be induced in adult V1. These include adult ocular dominance plasticity, retinal-lesion-induced plasticity, and perceptual learning. Another type of adult plasticity in rodent V1 is stimulus-selective response plasticity. When a visual stimulus is presented repeatedly, V1 will become more responsive to this stimulus but not to others [122]. Surprisingly, this type of plasticity can result in eye-specific changes in cortical responsiveness and may well involve plasticity at thalamocortical connections [122]. In line with the idea that PV+ interneurons are involved in regulating plasticity of feedforward connections, stimulus-selective response plasticity has recently been found to involve changes in PV+-interneuron-mediated inhibition [123]. As little is known about the exact nature of stimulus-selective response plasticity and the excitatory and inhibitory connectivity that is involved [124], we will not discuss it further. However, the fact that it is induced by passive viewing and may alter feedforward connections means that the separation of the substrates of plasticity with age is not absolute. Plasticity induced by monocular deprivation or retinal lesions Despite critical period closure, OD plasticity can still take place in the young adult cortex, though in a less efficient and permanent fashion than during the critical period [125–128]. In mice, a low level of OD plasticity can be induced up to 6 months of age [28]. While this phenomenon is particularly pronounced in mice, some OD plasticity after critical period closure is also observed in other species. In cats, for example, the critical period closes around 8 weeks of age, but some levels of OD plasticity can still be induced up to many months after birth [129]. Interestingly, OD plasticity in older cats does not involve layer 4, but is restricted to layers 2/3 and 5 [129]. A related form of cortical plasticity that can be readily induced in adult V1 across species is retinal-lesion-induced plasticity. Initially, V1 becomes unresponsive to the lesioned part of the retina. Over time, however, the lesion projection zone starts to respond to stimuli in neighboring visual-field positions [130–133]. This form of plasticity is also thought to involve the reorganization of horizontal connections in V1 [9]. Together, these findings support the idea that only during the critical period, feedforward connections undergo extensive plasticity, while horizontal and feedback connections are the main substrate of plasticity after critical period closure. Perceptual learning One of the dominant types of plasticity that occurs in sensory systems after critical period closure is perceptual learning. It can be induced experimentally by instructing the subject, and/or by rewarding or punishing certain behaviors in response to a specific visual stimulus. Depending on the specific task, perceptual learning can result in changes in the responses of V1 neurons [30–34] as well as in higher visual areas [134, 135]. Studies in macaque monkeys, for example, have shown that when monkeys are taught to discriminate between visual stimuli with slightly different orientations, changes in orientation tuning occur in the pyramidal layers of V1. Interestingly, no such changes are observed in the input layer, layer 4 [33]. A similar observation has been made in mice learning an active avoidance task. Mice learned to initiate running on a treadmill when a visual stimulus of a defined orientation was presented. Failure to do so resulted in a mild shock. In these mice, anticipatory responses to the punishment could be recorded in layer 2/3 neurons of V1, but not in layer 4 [10]. It was also observed that in layer 4, neuronal responses became sparser. However, this also occurred when mice were not trained but passively viewed the same visual stimuli, suggesting some type of habituation occurred that was unrelated to perceptual learning. In some tasks, perceptual learning in primates has its strongest influence in V1 during task execution, which suggests that in these cases, alterations occurred in higher visual areas or in feedback connections from higher to lower areas, but not in the feedforward connections from the LGN to V1 [32]. Moreover, anesthesia, which suppresses feedback inputs, typically also suppresses learned changes in V1 responses [136]. Together, these findings illustrate that perceptual learning in V1 typically involves plasticity in the extragranular layers receiving feedback connections from other brain regions rather than the input layers receiving feedforward sensory information. Disinhibition during adult cortical plasticity These forms of postcritical period plasticity are all associated with disinhibition. Chronic in vivo two-photon microscopy revealed that spines and boutons of interneurons are lost during retinal-lesion-induced plasticity in mouse V1 [50]. Other studies used gephyrin-GFP to label the postsynaptic side of inhibitory synapses. Using chronic in vivo imaging, these studies showed that inhibitory synapses formed onto pyramidal cell dendrites and spines in the top layers of V1 are rapidly eliminated when OD plasticity is induced in young adult mice [48, 49]. It is not yet clear what the identity is of the interneurons whose synapses are lost in these paradigms. Since volume release of GABA by NGF cells in layer 1 can strongly suppress the influence of callosal and possibly other layer 1 inputs on the dendritic tufts of layer 5 pyramidal cells [101], they are an interesting candidate. There is more evidence suggesting that SST+ interneurons are the main cell type involved. For one, they form most of the inhibitory synapses in the top layers. In addition, it was recently found that in mice learning a motor task, inhibitory synapses on pyramidal neurons in motor cortex were also lost, specifically those formed by SST+ interneurons. Inhibitory synapses formed by PV+ interneurons persisted [137], although it should be mentioned that only PV+ boutons forming synapses close to the cell body were assessed, while it is known that PV+ basket cells also form inhibitory synapses on distal dendrites and even spines [138]. Enhancing or decreasing the activity of SST+ interneurons using optogenetics interfered with the learned behavior. A more direct line of evidence suggesting the involvement of SST+ interneurons in regulating plasticity in adult V1 comes from studies analyzing the activity of SST+ interneurons during visual learning. In the previously mentioned active avoidance task in which mice learned to run in response to a visual stimulus to avoid a mild shock, it was observed that SST+ interneurons became less active. Increasing their activity interfered with the learned task [10]. Another series of studies found that when adult mice were running on a treadmill, while visual stimuli were presented, this resulted in the suppression of SST+ interneurons and a facilitation of OD plasticity [139, 140]. Others did not find evidence for SST+ interneuron suppression during running [141], and the cause of this apparent discrepancy still needs to be resolved. This notwithstanding, virally mediated expression of tetanus toxin in SST+ interneurons, which suppressed GABA release, also enhanced adult OD plasticity [14]. Together, these studies suggest that release from SST+-interneuron-mediated inhibition enhances adult plasticity. It thus appears that the change in the substrate of plasticity matches the interneuron subsets involved in regulating plasticity. While PV+ interneurons gating feedforward inputs regulate critical period plasticity, SST+ interneurons forming most inhibitory synapses in layer 1 and gating horizontal and feedback connections appear to be important regulators of adult cortical plasticity (Fig. 3). Disinhibitory circuits How can reduction of SST+-interneuron-mediated inhibition be achieved specifically during learning? Connectivity studies have found that SST+ Martinotti cells are innervated by VIP+ interneurons [87, 90, 92]. VIP+ interneurons, in turn, are extensively innervated by neurons in other brain areas [140] and express nicotinic acetylcholine receptors and serotonin receptors making them highly sensitive to neuromodulatory inputs [58, 142]. Through these long-range connections, VIP+ interneurons can thus be activated during behavioral states in which learning is required. In various brain regions of the mouse, VIP+-interneuron-mediated disinhibition has been found to involve modulatory inputs that signal reward, punishment, or arousal [92, 93, 143], consequently suppressing SST+ interneuron activity and releasing inhibition of horizontal or feedback connections. This disinhibition, in turn, stimulates plasticity. Similarly, in V1, the enhancement of adult OD plasticity in mice running on a treadmill required the activation of VIP+ interneurons [140]. Optogenetically activating VIP+ interneurons also enhanced adult OD plasticity even when the mice did not run [14]. How reducing SST+ interneuron mediated inhibition may enhance plasticity is under intense investigation. It is, however, tempting to speculate that disinhibition permits the potentiation of relevant feedback connections: those that provide the contextual information about feedforward inputs, which is relevant for making a choice leading to reward. To illustrate this, imagine a situation in which a subject needs to learn to recognize the image of a particular bird (Fig. 3). Every time the image of this bird is shown, and the subject can correctly differentiate it from other bird images; a reward is given. Neurons in V1 will respond to the image of the bird. However, feedforward inputs to a particular neuron in V1 could be just the same when a picture of another bird is shown. Only feedback connections to this neuron can provide contextual inputs differentiating between the various images. When these contextual inputs are disinhibited through reinforcement signals, their inputs may become strengthened causing the neuron in V1 to become more responsive to the feedforward input, but only when it is presented as part of the rewarded bird image. This may significantly improve the efficiency by which the image is recognized. An important question is whether disinhibition is in any way specific for the feedforward or feedback inputs that are being reinforced. For example, different VIP+ interneurons may be selectively activated by different contextual- or behavioral state-dependent inputs. Some may be activated by running, others by reward, punishment, or specific contextual feedback signals. Moreover, SST+ interneurons may be tuned to the feedforward inputs that are reinforced, or selectively innervate dendritic branches or spines that receive relevant contextual inputs. Evidence for these conditions has been found. For example, stimulus-specific disinhibition has been observed in V1, mediated through long-range connections from cingulate cortex onto VIP+ interneurons [144]. It is also known that SST+ interneurons have well-defined receptive fields, and show orientation tuning [72, 85, 144, 145]. Interestingly, SST+ interneurons often form inhibitory synapses onto dendritic spines [49, 146]. This wiring allows for the selective inhibition, and thus also disinhibition, of synaptic inputs. In support of such selective disinhibition, it was recently shown that when mice learn two different motor tasks, different dendritic branches of layer 5 pyramidal neurons show Ca2+ spikes [15]. Silencing SST+ interneurons caused a loss in the branch specificity of these Ca2+ spikes. SST+ interneuron silencing did not reduce learning of a single task, but did cause decreased performance in a previously learned task, once a second task was learned. Thus, SST+ interneurons appear to gate specific inputs to dendritic tufts and their suppression by VIP+ interneurons may allow the strengthening of selective inputs relevant to the task to be learned (Fig. 3). Unraveling the connectivity rules between interneurons of different subtypes and with different selectivity for visual or behavioral stimuli will be crucial for understanding how cortical plasticity is regulated, but may turn out to be a daunting task. Conclusions and future directions In summary, during different stages of brain maturation, plasticity is boosted by temporary disinhibition. However, the interneurons involved, the underlying mechanisms, and the inputs that undergo plasticity differ depending on the developmental stage and condition under which plasticity occurs. What is the use of this temporary disinhibition? We hypothesize that information processing is more efficient when signal-to-noise ratios are high. However, this comes at the cost of suppressing information that may be essential to execute tasks that are not routine and require learning. Thus, the downregulation of interneuron activity under the right circumstances may help to serve these opposing needs. Many fundamental questions remain to be answered. During development, critical periods in different brain areas occur at different stages of postnatal development. Generally speaking, higher cortical areas undergo plasticity at a later stage than lower areas. An important question is how this timing is regulated. The timing of the critical period in V1 is partially regulated through retinal input, which drives the development of the extracellular matrix and inhibitory innervation [147]. Do high cortical areas wait for a particular type of input from lower areas? Or is a strict genetic program followed? It is also unknown to what extent the development of feedback connections awaits closure of the critical period. It would be especially important to understand whether inhibitory inputs in layer 1 hold off the development or plasticity of these feedback inputs. This could be regulated through an initially strong influence of NGF cells on dendritic tufts or low influence of VIP+ interneurons over SST+ interneurons during the critical period. However, until now, it remains unknown what role VIP+-, SST+- or NGF interneurons play during the critical period. Similarly, it has been noticed that after the critical period, OD plasticity can only be induced in young adult but not in older mice [28]. Possibly, SST+ interneurons may also become less controllable with age, thus further reducing plasticity of horizontal connections. This may explain why transplantation of embryonic SST+ interneurons enhances adult OD plasticity [121]. It remains unclear whether the connectivity of different interneuron subsets as described in V1 is the same in other cortical areas. Studies on disinhibition in auditory, sensory, visual, prefrontal, and motor cortices have already provided some apparently contradicting results [87, 92, 93, 143]. Most likely, general connectivity rules between interneuron subtypes exist across the neocortex. However, at the same time, many different subtypes of SST+, PV+, VIP+, and NGF interneurons may exist with diverse connectivity patterns and properties. These patterns may well be dependent on the function of the cortical area and the specific responsiveness of interneurons to various behavioral conditions and sensory inputs. The identification of additional genetic markers to further subdivide the various interneuron populations may help understanding the connectivity rules of cortical inhibition [100]. In addition, extensive connectivity studies [87, 90, 98, 148] of interneurons whose functional properties have been determined in vivo, as done for excitatory neurons, [149] will be required for solving this complex puzzle. Interneuron dysfunction has been implicated in many neurodevelopmental disorders, including autism, schizophrenia, and intellectual disability [150, 151]. The increasing knowledge on the role of inhibition in the regulation of critical periods during development and reinforcement learning later on is likely to open up new avenues to treat these disorders. This may involve extending or reactivating critical periods to correct or prevent maladaptation of the developing networks, or altering inhibitory tone to improve the learning ability of people suffering from these disorders [152]. In rodents, several approaches that alter cortical inhibition have proved effective in increasing plasticity in V1, including environmental enrichment [153, 154], housing animals in the dark [155, 156], degrading the extracellular matrix [39], treatment with serotonin reuptake inhibitors [157], and opto- or pharmacogenetically altering interneuron activity [14, 47]. To develop selective and powerful approaches to enhance cortical plasticity in human patients, it is crucial that we identify the exact working mechanisms and targets of these treatments. A better understanding of how inhibition and disinhibition regulate cortical plasticity is, therefore, indispensable. The authors would like to thank Drs. J. Alexander Heimel, Christian Lohmann, Matthew W. Self and Pieter R. Roelfsema for the critical reading of the manuscript. This research was made possible through funding from the European Community’s Seventh Framework Programme (FP2007-2013) under grant agreement no 223326. This work was also supported by a donation from Praktijkgenerator b.v. and a grant from Stichting Blindenhulp. ==== Refs References 1. Lorenz K Der Kumpan in der Umwelt des Vogels J Ornithol 1935 83 137 213 10.1007/BF01905355 2. Friedmann N Rusou D Critical period for first language: the crucial role of language input during the first year of life Curr Opin Neurobiol 2015 35 27 34 10.1016/j.conb.2015.06.003 26111432 3. Kuhl PK Brain mechanisms in early language acquisition Neuron 2010 67 713 727 10.1016/j.neuron.2010.08.038 20826304 4. Levi DM Pathophysiology of binocular vision and amblyopia Curr Opin Ophthalmol 1994 5 3 10 10.1097/00055735-199410000-00002 10150812 5. Vaegan Taylor D Critical period for deprivation amblyopia in children Trans Ophthalmol Soc UK 1979 99 432 439 298827 6. Hochstein S Ahissar M View from the top: hierarchies and reverse hierarchies in the visual system Neuron 2002 36 791 804 10.1016/S0896-6273(02)01091-7 12467584 7. Lamme VA Roelfsema PR The distinct modes of vision offered by feedforward and recurrent processing Trends Neurosci 2000 23 571 579 10.1016/S0166-2236(00)01657-X 11074267 8. Hubel DH Wiesel TN LeVay S Plasticity of ocular dominance columns in monkey striate cortex Philos Trans R Soc Lond B Biol Sci 1977 278 377 409 10.1098/rstb.1977.0050 19791 9. Gilbert CD Li W Adult visual cortical plasticity Neuron 2012 75 250 264 10.1016/j.neuron.2012.06.030 22841310 10. Makino H Komiyama T Learning enhances the relative impact of top-down processing in the visual cortex Nat Neurosci 2015 18 1116 1122 10.1038/nn.4061 26167904 11. Hensch TK Fagiolini M Mataga N Stryker MP Baekkeskov S Kash SF Local GABA circuit control of experience-dependent plasticity in developing visual cortex Science 1998 282 1504 1508 10.1126/science.282.5393.1504 9822384 12. Fagiolini M Hensch TK Inhibitory threshold for critical-period activation in primary visual cortex Nature 2000 404 183 186 10.1038/35004582 10724170 13. Huang ZJ Kirkwood A Pizzorusso T Porciatti V Morales B Bear MF Maffei L Tonegawa S BDNF regulates the maturation of inhibition and the critical period of plasticity in mouse visual cortex Cell 1999 98 739 755 10.1016/S0092-8674(00)81509-3 10499792 14. Fu Y Kaneko M Tang Y Alvarez-Buylla A Stryker MP A cortical disinhibitory circuit for enhancing adult plasticity Elife 2015 4 e05558 10.7554/eLife.05558 25626167 15. Cichon J Gan WB Branch-specific dendritic Ca(2+) spikes cause persistent synaptic plasticity Nature 2015 520 180 185 10.1038/nature14251 25822789 16. Siegel F Heimel JA Peters J Lohmann C Peripheral and central inputs shape network dynamics in the developing visual cortex in vivo Curr Biol 2012 22 253 258 10.1016/j.cub.2011.12.026 22264606 17. Mooney R Penn AA Gallego R Shatz CJ Thalamic relay of spontaneous retinal activity prior to vision Neuron 1996 17 863 874 10.1016/S0896-6273(00)80218-4 8938119 18. Levelt CN Hubener M Critical-period plasticity in the visual cortex Annu Rev Neurosci 2012 35 309 330 10.1146/annurev-neuro-061010-113813 22462544 19. Polyak SL The vertebrate visual system; its origin, structure, and function and its manifestations in disease 1957 Chicago University of Chicago Press 20. Wiesel TN Hubel DH Single-cell responses in striate cortex of kittens deprived of vision in one eye J Neurophysiol 1963 26 1003 1017 14084161 21. Drager UC Observations on monocular deprivation in mice J Neurophysiol 1978 41 28 42 621544 22. Antonini A Stryker MP Plasticity of geniculocortical afferents following brief or prolonged monocular occlusion in the cat J Comp Neurol. 1996 369 64 82 10.1002/(SICI)1096-9861(19960520)369:1<64::AID-CNE5>3.0.CO;2-I 8723703 23. Antonini A Fagiolini M Stryker MP Anatomical correlates of functional plasticity in mouse visual cortex J Neurosci 1999 19 4388 4406 10341241 24. Shatz CJ Stryker MP Ocular dominance in layer IV of the cat’s visual cortex and the effects of monocular deprivation J Physiol 1978 281 267 283 10.1113/jphysiol.1978.sp012421 702379 25. Mataga N Mizuguchi Y Hensch TK Experience-dependent pruning of dendritic spines in visual cortex by tissue plasminogen activator Neuron 2004 44 1031 1041 10.1016/j.neuron.2004.11.028 15603745 26. Tropea D Majewska AK Garcia R Sur M Structural dynamics of synapses in vivo correlate with functional changes during experience-dependent plasticity in visual cortex J Neurosci 2010 30 11086 11095 10.1523/JNEUROSCI.1661-10.2010 20720116 27. Holtmaat AJ Trachtenberg JT Wilbrecht L Shepherd GM Zhang X Knott GW Svoboda K Transient and persistent dendritic spines in the neocortex in vivo Neuron 2005 45 279 291 10.1016/j.neuron.2005.01.003 15664179 28. Lehmann K Lowel S Age-dependent ocular dominance plasticity in adult mice PLoS One 2008 3 e3120 10.1371/journal.pone.0003120 18769674 29. Hofer SB Mrsic-Flogel TD Bonhoeffer T Hubener M Experience leaves a lasting structural trace in cortical circuits Nature 2009 457 313 317 10.1038/nature07487 19005470 30. De Weerd P Reithler J van de Ven V Been M Jacobs C Sack AT Posttraining transcranial magnetic stimulation of striate cortex disrupts consolidation early in visual skill learning J Neurosci 2012 32 1981 1988 10.1523/JNEUROSCI.3712-11.2011 22323712 31. Ghose D Barnett ZP Wallace MT Impact of response duration on multisensory integration J Neurophysiol 2012 108 2534 2544 10.1152/jn.00286.2012 22896723 32. Li W Piech V Gilbert CD Perceptual learning and top-down influences in primary visual cortex Nat Neurosci 2004 7 651 657 10.1038/nn1255 15156149 33. Schoups A Vogels R Qian N Orban G Practising orientation identification improves orientation coding in V1 neurons Nature 2001 412 549 553 10.1038/35087601 11484056 34. Shibata K Watanabe T Sasaki Y Kawato M Perceptual learning incepted by decoded fMRI neurofeedback without stimulus presentation Science 2011 334 1413 1415 10.1126/science.1212003 22158821 35. Roelfsema PR van Ooyen A Watanabe T Perceptual learning rules based on reinforcers and attention Trends Cogn Sci 2010 14 64 71 10.1016/j.tics.2009.11.005 20060771 36. McGee AW Yang Y Fischer QS Daw NW Strittmatter SM Experience-driven plasticity of visual cortex limited by myelin and Nogo receptor Science 2005 309 2222 2226 10.1126/science.1114362 16195464 37. Syken J Grandpre T Kanold PO Shatz CJ PirB restricts ocular-dominance plasticity in visual cortex Science 2006 313 1795 1800 10.1126/science.1128232 16917027 38. Huang X Stodieck SK Goetze B Cui L Wong MH Wenzel C Hosang L Dong Y Lowel S Schluter OM Progressive maturation of silent synapses governs the duration of a critical period Proc Natl Acad Sci USA 2015 112 E3131 E3140 10.1073/pnas.1506488112 26015564 39. Pizzorusso T Medini P Berardi N Chierzi S Fawcett JW Maffei L Reactivation of ocular dominance plasticity in the adult visual cortex Science 2002 298 1248 1251 10.1126/science.1072699 12424383 40. Morishita H Hensch TK Critical period revisited: impact on vision Curr Opin Neurobiol 2008 18 101 107 10.1016/j.conb.2008.05.009 18534841 41. Heimel JA van Versendaal D Levelt CN The role of GABAergic inhibition in ocular dominance plasticity Neural Plast. 2011 2011 391763 21826276 42. Trachtenberg JT Competition, inhibition, and critical periods of cortical plasticity Curr Opin Neurobiol 2015 35 44 48 10.1016/j.conb.2015.06.006 26126153 43. Di CG Chattopadhyaya B Kuhlman SJ Fu Y Belanger MC Wu CZ Rutishauser U Maffei L Huang ZJ Activity-dependent PSA expression regulates inhibitory maturation and onset of critical period plasticity Nat Neurosci 2007 10 1569 1577 10.1038/nn2008 18026099 44. Hanover JL Huang ZJ Tonegawa S Stryker MP Brain-derived neurotrophic factor overexpression induces precocious critical period in mouse visual cortex J Neurosci 1999 19 RC40 10559430 45. Baroncelli L Sale A Viegi A Maya Vetencourt JF De PR Baldini S Maffei L Experience-dependent reactivation of ocular dominance plasticity in the adult visual cortex Exp Neurol 2010 226 100 109 10.1016/j.expneurol.2010.08.009 20713044 46. Harauzov A Spolidoro M DiCristo G De PR Cancedda L Pizzorusso T Viegi A Berardi N Maffei L Reducing intracortical inhibition in the adult visual cortex promotes ocular dominance plasticity J Neurosci 2010 30 361 371 10.1523/JNEUROSCI.2233-09.2010 20053917 47. Kuhlman SJ Olivas ND Tring E Ikrar T Xu X Trachtenberg JT A disinhibitory microcircuit initiates critical-period plasticity in the visual cortex Nature 2013 501 543 546 10.1038/nature12485 23975100 48. Chen JL Villa KL Cha JW So PT Kubota Y Nedivi E Clustered dynamics of inhibitory synapses and dendritic spines in the adult neocortex Neuron 2012 74 361 373 10.1016/j.neuron.2012.02.030 22542188 49. van Versendaal D Rajendran R Saiepour MH Klooster J Smit-Rigter L Sommeijer JP De Zeeuw CI Hofer SB Heimel JA Levelt CN Elimination of inhibitory synapses is a major component of adult ocular dominance plasticity Neuron 2012 74 374 383 10.1016/j.neuron.2012.03.015 22542189 50. Keck T Scheuss V Jacobsen RI Wierenga CJ Eysel UT Bonhoeffer T Hubener M Loss of sensory input causes rapid structural changes of inhibitory neurons in adult mouse visual cortex Neuron 2011 71 869 882 10.1016/j.neuron.2011.06.034 21903080 51. Ascoli GA Alonso-Nanclares L Anderson SA Barrionuevo G Benavides-Piccione R Burkhalter A Buzsaki G Cauli B DeFelipe J Fairen A Feldmeyer D Fishell G Fregnac Y Freund TF Gardner D Gardner EP Goldberg JH Helmstaedter M Hestrin S Karube F Kisvarday ZF Lambolez B Lewis DA Marin O Markram H Munoz A Packer A Petersen CC Rockland KS Rossier J Rudy B Somogyi P Staiger JF Tamas G Thomson AM Toledo-Rodriguez M Wang Y West DC Yuste R Petilla terminology: nomenclature of features of GABAergic interneurons of the cerebral cortex Nat Rev Neurosci 2008 9 557 568 10.1038/nrn2402 18568015 52. Kepecs A Fishell G Interneuron cell types are fit to function Nature 2014 505 318 326 10.1038/nature12983 24429630 53. Anderson SA Eisenstat DD Shi L Rubenstein JL Interneuron migration from basal forebrain to neocortex: dependence on Dlx genes Science 1997 278 474 476 10.1126/science.278.5337.474 9334308 54. Miyoshi G Fishell G GABAergic interneuron lineages selectively sort into specific cortical layers during early postnatal development Cereb Cortex 2011 21 845 852 10.1093/cercor/bhq155 20732898 55. Butt SJ Fuccillo M Nery S Noctor S Kriegstein A Corbin JG Fishell G The temporal and spatial origins of cortical interneurons predict their physiological subtype Neuron 2005 48 591 604 10.1016/j.neuron.2005.09.034 16301176 56. Anderson SA Marin O Horn C Jennings K Rubenstein JL Distinct cortical migrations from the medial and lateral ganglionic eminences Development 2001 128 353 363 11152634 57. Miyoshi G Hjerling-Leffler J Karayannis T Sousa VH Butt SJ Battiste J Johnson JE Machold RP Fishell G Genetic fate mapping reveals that the caudal ganglionic eminence produces a large and diverse population of superficial cortical interneurons J Neurosci 2010 30 1582 1594 10.1523/JNEUROSCI.4515-09.2010 20130169 58. Rudy B Fishell G Lee S Hjerling-Leffler J Three groups of interneurons account for nearly 100 % of neocortical GABAergic neurons Dev Neurobiol 2011 71 45 61 10.1002/dneu.20853 21154909 59. Lee S Hjerling-Leffler J Zagha E Fishell G Rudy B The largest group of superficial neocortical GABAergic interneurons expresses ionotropic serotonin receptors J Neurosci 2010 30 16796 16808 10.1523/JNEUROSCI.1869-10.2010 21159951 60. Xu X Roby KD Callaway EM Immunochemical characterization of inhibitory mouse cortical neurons: three chemically distinct classes of inhibitory cells J Comp Neurol 2010 518 389 404 10.1002/cne.22229 19950390 61. Woodruff AR McGarry LM Vogels TP Inan M Anderson SA Yuste R State-dependent function of neocortical chandelier cells J Neurosci 2011 31 17872 17886 10.1523/JNEUROSCI.3894-11.2011 22159102 62. Szabadics J Varga C Molnar G Olah S Barzo P Tamas G Excitatory effect of GABAergic axo-axonic cells in cortical microcircuits Science 2006 311 233 235 10.1126/science.1121325 16410524 63. Somogyi P Kisvarday ZF Martin KA Whitteridge D Synaptic connections of morphologically identified and physiologically characterized large basket cells in the striate cortex of cat Neuroscience 1983 10 261 294 10.1016/0306-4522(83)90133-1 6633861 64. Kubota Y Kawaguchi Y Dependence of GABAergic synaptic areas on the interneuron type and target size J Neurosci 2000 20 375 386 10627614 65. Kawaguchi Y Kubota Y Neurochemical features and synaptic connections of large physiologically-identified GABAergic cells in the rat frontal cortex Neuroscience 1998 85 677 701 10.1016/S0306-4522(97)00685-4 9639265 66. Markram H Toledo-Rodriguez M Wang Y Gupta A Silberberg G Wu C Interneurons of the neocortical inhibitory system Nat Rev Neurosci 2004 5 793 807 10.1038/nrn1519 15378039 67. Gonchar Y Wang Q Burkhalter A Multiple distinct subtypes of GABAergic neurons in mouse visual cortex identified by triple immunostaining Front Neuroanat 2007 1 3 18958197 68. Kawaguchi Y Kubota Y GABAergic cell subtypes and their synaptic connections in rat frontal cortex Cereb Cortex 1997 7 476 486 10.1093/cercor/7.6.476 9276173 69. DeFelipe J Jones EG Parvalbumin immunoreactivity reveals layer IV of monkey cerebral cortex as a mosaic of microzones of thalamic afferent terminations Brain Res 1991 562 39 47 10.1016/0006-8993(91)91184-3 1799871 70. Del Rio MR DeFelipe J A study of SMI 32-stained pyramidal cells, parvalbumin-immunoreactive chandelier cells, and presumptive thalamocortical axons in the human temporal neocortex J Comp Neurol 1994 342 389 408 10.1002/cne.903420307 7517410 71. Cruikshank SJ Lewis TJ Connors BW Synaptic basis for intense thalamocortical activation of feedforward inhibitory cells in neocortex Nat Neurosci 2007 10 462 468 17334362 72. Kerlin AM Andermann ML Berezovskii VK Reid RC Broadly tuned response properties of diverse inhibitory neuron subtypes in mouse visual cortex Neuron 2010 67 858 871 10.1016/j.neuron.2010.08.002 20826316 73. Hofer SB Ko H Pichler B Vogelstein J Ros H Zeng H Lein E Lesica NA Mrsic-Flogel TD Differential connectivity and response dynamics of excitatory and inhibitory neurons in visual cortex Nat Neurosci 2011 14 1045 1052 10.1038/nn.2876 21765421 74. Atallah BV Bruns W Carandini M Scanziani M Parvalbumin-expressing interneurons linearly transform cortical responses to visual stimuli Neuron 2012 73 159 170 10.1016/j.neuron.2011.12.013 22243754 75. Runyan CA Schummers J Van WA Kuhlman SJ Wilson NR Huang ZJ Sur M Response features of parvalbumin-expressing interneurons suggest precise roles for subtypes of inhibition in visual cortex Neuron 2010 67 847 857 10.1016/j.neuron.2010.08.006 20826315 76. Pouille F Scanziani M Enforcement of temporal fidelity in pyramidal cells by somatic feed-forward inhibition Science 2001 293 1159 1163 10.1126/science.1060342 11498596 77. Cardin JA Carlen M Meletis K Knoblich U Zhang F Deisseroth K Tsai LH Moore CI Driving fast-spiking cells induces gamma rhythm and controls sensory responses Nature 2009 459 663 667 10.1038/nature08002 19396156 78. Galarreta M Hestrin S Electrical and chemical synapses among parvalbumin fast-spiking GABAergic interneurons in adult mouse neocortex Proc Natl Acad Sci USA 2002 99 12438 12443 10.1073/pnas.192159599 12213962 79. Sohal VS Zhang F Yizhar O Deisseroth K Parvalbumin neurons and gamma rhythms enhance cortical circuit performance Nature 2009 459 698 702 10.1038/nature07991 19396159 80. Tamas G Buhl EH Lorincz A Somogyi P Proximally targeted GABAergic synapses and gap junctions synchronize cortical interneurons Nat Neurosci 2000 3 366 371 10.1038/73936 10725926 81. Wang Y Toledo-Rodriguez M Gupta A Wu C Silberberg G Luo J Markram H Anatomical, physiological and molecular properties of Martinotti cells in the somatosensory cortex of the juvenile rat J Physiol 2004 561 65 90 10.1113/jphysiol.2004.073353 15331670 82. Fanselow EE Richardson KA Connors BW Selective, state-dependent activation of somatostatin-expressing inhibitory interneurons in mouse neocortex J Neurophysiol 2008 100 2640 2652 10.1152/jn.90691.2008 18799598 83. Blakemore C Tobin EA Lateral inhibition between orientation detectors in the cat’s visual cortex Exp Brain Res 1972 15 439 440 10.1007/BF00234129 5079475 84. Adesnik H Bruns W Taniguchi H Huang ZJ Scanziani M A neural circuit for spatial summation in visual cortex Nature 2012 490 226 231 10.1038/nature11526 23060193 85. Saiepour MH Rajendran R Omrani A Ma WP Tao HW Heimel JA Levelt CN Ocular dominance plasticity disrupts binocular inhibition-excitation matching in visual cortex Curr Biol 2015 25 713 721 10.1016/j.cub.2015.01.024 25754642 86. Self MW Lorteije JA Vangeneugden J van Beest EH Grigore ME Levelt CN Heimel JA Roelfsema PR Orientation-tuned surround suppression in mouse visual cortex J Neurosci 2014 34 9290 9304 10.1523/JNEUROSCI.5051-13.2014 25009262 87. Pfeffer CK Xue M He M Huang ZJ Scanziani M Inhibition of inhibition in visual cortex: the logic of connections between molecularly distinct interneurons Nat Neurosci 2013 16 1068 1076 10.1038/nn.3446 23817549 88. Cottam JC Smith SL Hausser M Target-specific effects of somatostatin-expressing interneurons on neocortical visual processing J Neurosci 2013 33 19567 19578 10.1523/JNEUROSCI.2624-13.2013 24336721 89. Xu H Jeong HY Tremblay R Rudy B Neocortical somatostatin-expressing GABAergic interneurons disinhibit the thalamorecipient layer 4 Neuron 2013 77 155 167 10.1016/j.neuron.2012.11.004 23312523 90. Jiang X Shen S Cadwell CR Berens P Sinz F Ecker AS Patel S Tolias AS Principles of connectivity among morphologically defined cell types in adult neocortex Science 2015 350 aac9462 10.1126/science.aac9462 26612957 91. David C Schleicher A Zuschratter W Staiger JF The innervation of parvalbumin-containing interneurons by VIP-immunopositive interneurons in the primary somatosensory cortex of the adult rat Eur J Neurosci 2007 25 2329 2340 10.1111/j.1460-9568.2007.05496.x 17445231 92. Lee S Kruglikov I Huang ZJ Fishell G Rudy B A disinhibitory circuit mediates motor integration in the somatosensory cortex Nat Neurosci 2013 16 1662 1670 10.1038/nn.3544 24097044 93. Pi HJ Hangya B Kvitsiani D Sanders JI Huang ZJ Kepecs A Cortical interneurons that specialize in disinhibitory control Nature 2013 503 521 524 10.1038/nature12676 24097352 94. Paspalas CD Papadopoulos GC Serotoninergic afferents preferentially innervate distinct subclasses of peptidergic interneurons in the rat visual cortex Brain Res 2001 891 158 167 10.1016/S0006-8993(00)03193-0 11164819 95. Kawaguchi Y Selective cholinergic modulation of cortical GABAergic cell subtypes J Neurophysiol 1997 78 1743 1747 9310461 96. Armstrong C Krook-Magnuson E Soltesz I Neurogliaform and Ivy cells: a major family of nNOS expressing GABAergic neurons Front Neural Circuits 2012 6 23 10.3389/fncir.2012.00023 22623913 97. Szabadics J Tamas G Soltesz I Different transmitter transients underlie presynaptic cell type specificity of GABAA, slow and GABAA, fast Proc Natl Acad Sci USA 2007 104 14831 14836 10.1073/pnas.0707204104 17785408 98. Jiang X Wang G Lee AJ Stornetta RL Zhu JJ The organization of two new cortical interneuronal circuits Nat Neurosci 2013 16 210 218 10.1038/nn.3305 23313910 99. Simon A Olah S Molnar G Szabadics J Tamas G Gap-junctional coupling between neurogliaform cells and various interneuron types in the neocortex J Neurosci 2005 25 6278 6285 10.1523/JNEUROSCI.1431-05.2005 16000617 100. Tasic B Menon V Nguyen TN Kim TK Jarsky T Yao Z Levi B Gray LT Sorensen SA Dolbeare T Bertagnolli D Goldy J Shapovalova N Parry S Lee C Smith K Bernard A Madisen L Sunkin SM Hawrylycz M Koch C Zeng H Adult mouse cortical cell taxonomy revealed by single cell transcriptomics Nat Neurosci 2016 19 2 335 346 10.1038/nn.4216 26727548 101. Palmer LM Schulz JM Murphy SC Ledergerber D Murayama M Larkum ME The cellular basis of GABA(B)-mediated interhemispheric inhibition Science 2012 335 989 993 10.1126/science.1217276 22363012 102. Chittajallu R Pelkey KA McBain CJ Neurogliaform cells dynamically regulate somatosensory integration via synapse-specific modulation Nat Neurosci 2013 16 13 15 10.1038/nn.3284 23222912 103. Ackman JB Crair MC Role of emergent neural activity in visual map development Curr Opin Neurobiol 2014 24 166 175 10.1016/j.conb.2013.11.011 24492092 104. Winnubst J Cheyne JE Niculescu D Lohmann C Spontaneous activity drives local synaptic plasticity in vivo Neuron 2015 87 399 410 10.1016/j.neuron.2015.06.029 26182421 105. Shatz CJ Stryker MP Prenatal tetrodotoxin infusion blocks segregation of retinogeniculate afferents Science 1988 242 87 89 10.1126/science.3175636 3175636 106. Gogolla N Takesian AE Feng G Fagiolini M Hensch TK Sensory integration in mouse insular cortex reflects GABA circuit maturation Neuron 2014 83 894 905 10.1016/j.neuron.2014.06.033 25088363 107. Toyoizumi T Miyamoto H Yazaki-Sugiyama Y Atapour N Hensch TK Miller KD A theory of the transition to critical period plasticity: inhibition selectively suppresses spontaneous activity Neuron 2013 80 51 63 10.1016/j.neuron.2013.07.022 24094102 108. Tagawa Y Kanold PO Majdan M Shatz CJ Multiple periods of functional ocular dominance plasticity in mouse visual cortex Nat Neurosci 2005 8 380 388 10.1038/nn1410 15723060 109. Kuhlman SJ Lu J Lazarus MS Huang ZJ Maturation of GABAergic inhibition promotes strengthening of temporally coherent inputs among convergent pathways PLoS Comput Biol 2010 6 e1000797 10.1371/journal.pcbi.1000797 20532211 110. Yazaki-Sugiyama Y Kang S Cateau H Fukai T Hensch TK Bidirectional plasticity in fast-spiking GABA circuits by visual experience Nature 2009 462 218 221 10.1038/nature08485 19907494 111. Gandhi SP Yanagawa Y Stryker MP Delayed plasticity of inhibitory neurons in developing visual cortex Proc Natl Acad Sci USA 2008 105 16797 16802 10.1073/pnas.0806159105 18940923 112. Kameyama K Sohya K Ebina T Fukuda A Yanagawa Y Tsumoto T Difference in binocularity and ocular dominance plasticity between GABAergic and excitatory cortical neurons J Neurosci 2010 30 1551 1559 10.1523/JNEUROSCI.5025-09.2010 20107082 113. Duffy FH Burchfiel JL Conway JL Bicuculline reversal of deprivation amblyopia in the cat Nature 1976 260 256 257 10.1038/260256a0 1256565 114. Sillito AM Kemp JA Blakemore C The role of GABAergic inhibition in the cortical effects of monocular deprivation Nature 1981 291 318 320 10.1038/291318a0 7231550 115. Sillito AM Kemp JA Patel H Inhibitory interactions contributing to the ocular dominance of monocularly dominated cells in the normal cat striate cortex Exp Brain Res 1980 41 1 10 10.1007/BF00236673 7461064 116. Ye Q Miao QL Experience-dependent development of perineuronal nets and chondroitin sulfate proteoglycan receptors in mouse visual cortex Matrix Biol 2013 32 352 363 10.1016/j.matbio.2013.04.001 23597636 117. Stephany CE Chan LL Parivash SN Dorton HM Piechowicz M Qiu S McGee AW Plasticity of binocularity and visual acuity are differentially limited by nogo receptor J Neurosci 2014 34 11631 11640 10.1523/JNEUROSCI.0545-14.2014 25164659 118. Carulli D Pizzorusso T Kwok JC Putignano E Poli A Forostyak S Andrews MR Deepa SS Glant TT Fawcett JW Animals lacking link protein have attenuated perineuronal nets and persistent plasticity Brain 2010 133 2331 2347 10.1093/brain/awq145 20566484 119. Southwell DG Froemke RC Alvarez-Buylla A Stryker MP Gandhi SP Cortical plasticity induced by inhibitory neuron transplantation Science 2010 327 1145 1148 10.1126/science.1183962 20185728 120. Davis MF Figueroa Velez DX Guevarra RP Yang MC Habeeb M Carathedathu MC Gandhi SP inhibitory neuron transplantation into adult visual cortex creates a new critical period that rescues impaired vision Neuron 2015 86 1055 1066 10.1016/j.neuron.2015.03.062 25937171 121. Tang Y Stryker MP Alvarez-Buylla A Espinosa JS Cortical plasticity induced by transplantation of embryonic somatostatin or parvalbumin interneurons Proc Natl Acad Sci USA 2014 111 18339 18344 10.1073/pnas.1421844112 25489113 122. Frenkel MY Sawtell NB Diogo AC Yoon B Neve RL Bear MF Instructive effect of visual experience in mouse visual cortex Neuron 2006 51 339 349 10.1016/j.neuron.2006.06.026 16880128 123. Kaplan ES Cooke SF Komorowski RW Chubykin AA Thomazeau A Khibnik LA Gavornik JP Bear MF Contrasting roles for parvalbumin-expressing inhibitory neurons in two forms of adult visual cortical plasticity Elife 2016 5 e11450 10.7554/eLife.11450 26943618 124. Cooke SF Bear MF How the mechanisms of long-term synaptic potentiation and depression serve experience-dependent plasticity in primary visual cortex Philos Trans R Soc Lond B Biol Sci 2014 369 20130284 10.1098/rstb.2013.0284 24298166 125. Heimel JA Hartman RJ Hermans JM Levelt CN Screening mouse vision with intrinsic signal optical imaging Eur J Neurosci 2007 25 795 804 10.1111/j.1460-9568.2007.05333.x 17328775 126. Sawtell NB Frenkel MY Philpot BD Nakazawa K Tonegawa S Bear MF NMDA receptor-dependent ocular dominance plasticity in adult visual cortex Neuron 2003 38 977 985 10.1016/S0896-6273(03)00323-4 12818182 127. Hofer SB Mrsic-Flogel TD Bonhoeffer T Hubener M Lifelong learning: ocular dominance plasticity in mouse visual cortex Curr Opin Neurobiol 2006 16 451 459 10.1016/j.conb.2006.06.007 16837188 128. Pham TA Graham SJ Suzuki S Barco A Kandel ER Gordon B Lickey ME A semi-persistent adult ocular dominance plasticity in visual cortex is stabilized by activated CREB Learn Mem 2004 11 738 747 10.1101/lm.75304 15537732 129. Daw NW Fox K Sato H Czepita D Critical period for monocular deprivation in the cat visual cortex J Neurophysiol 1992 67 197 202 1552319 130. Gilbert CD Wiesel TN Receptive field dynamics in adult primary visual cortex Nature 1992 356 150 152 10.1038/356150a0 1545866 131. Giannikopoulos DV Eysel UT Dynamics and specificity of cortical map reorganization after retinal lesions Proc Natl Acad Sci USA 2006 103 10805 10810 10.1073/pnas.0604539103 16818873 132. Kaas JH Krubitzer LA Chino YM Langston AL Polley EH Blair N Reorganization of retinotopic cortical maps in adult mammals after lesions of the retina Science 1990 248 229 231 10.1126/science.2326637 2326637 133. Keck T Mrsic-Flogel TD Vaz AM Eysel UT Bonhoeffer T Hubener M Massive restructuring of neuronal circuits during functional reorganization of adult visual cortex Nat Neurosci 2008 11 1162 1167 10.1038/nn.2181 18758460 134. Freedman DJ Assad JA Experience-dependent representation of visual categories in parietal cortex Nature 2006 443 85 88 10.1038/nature05078 16936716 135. Yang T Maunsell JH The effect of perceptual learning on neuronal responses in monkey visual area V4 J Neurosci 2004 24 1617 1626 10.1523/JNEUROSCI.4442-03.2004 14973244 136. Li W Piech V Gilbert CD Learning to link visual contours Neuron 2008 57 442 451 10.1016/j.neuron.2007.12.011 18255036 137. Chen SX Kim AN Peters AJ Komiyama T Subtype-specific plasticity of inhibitory circuits in motor cortex during motor learning Nat Neurosci 2015 18 1109 1115 10.1038/nn.4049 26098758 138. Wang Y Gupta A Toledo-Rodriguez M Wu CZ Markram H Anatomical, physiological, molecular and circuit properties of nest basket cells in the developing somatosensory cortex Cereb Cortex 2002 12 395 410 10.1093/cercor/12.4.395 11884355 139. Kaneko M Stryker MP Sensory experience during locomotion promotes recovery of function in adult visual cortex Elife 2014 3 e02798 10.7554/eLife.02798 24970838 140. Fu Y Tucciarone JM Espinosa JS Sheng N Darcy DP Nicoll RA Huang ZJ Stryker MP A cortical circuit for gain control by behavioral state Cell 2014 156 1139 1152 10.1016/j.cell.2014.01.050 24630718 141. Polack PO Friedman J Golshani P Cellular mechanisms of brain state-dependent gain modulation in visual cortex Nat Neurosci 2013 16 1331 1339 10.1038/nn.3464 23872595 142. Alitto HJ Dan Y Cell-type-specific modulation of neocortical activity by basal forebrain input Front Syst Neurosci 2012 6 79 23316142 143. Letzkus JJ Wolff SB Meyer EM Tovote P Courtin J Herry C Luthi A A disinhibitory microcircuit for associative fear learning in the auditory cortex Nature 2011 480 331 335 10.1038/nature10674 22158104 144. Zhang S Xu M Kamigaki T Hoang Do JP Chang WC Jenvay S Miyamichi K Luo L Dan Y Selective attention. Long-range and local circuits for top-down modulation of visual cortex processing Science 2014 345 660 665 10.1126/science.1254126 25104383 145. Ma WP Liu BH Li YT Huang ZJ Zhang LI Tao HW Visual representations by cortical somatostatin inhibitory neurons-selective but with weak and delayed responses J Neurosci 2010 30 14371 14379 10.1523/JNEUROSCI.3248-10.2010 20980594 146. Chiu CQ Lur G Morse TM Carnevale NT Ellis-Davies GC Higley MJ Compartmentalization of GABAergic inhibition by dendritic spines Science 2013 340 759 762 10.1126/science.1234274 23661763 147. Berardi N Pizzorusso T Maffei L Extracellular matrix and visual cortical plasticity: freeing the synapse Neuron 2004 44 905 908 15603733 148. Markram H Muller E Ramaswamy S Reimann MW Abdellah M Sanchez CA Ailamaki A Alonso-Nanclares L Antille N Arsever S Kahou GA Berger TK Bilgili A Buncic N Chalimourda A Chindemi G Courcol JD Delalondre F Delattre V Druckmann S Dumusc R Dynes J Eilemann S Gal E Gevaert ME Ghobril JP Gidon A Graham JW Gupta A Haenel V Hay E Heinis T Hernando JB Hines M Kanari L Keller D Kenyon J Khazen G Kim Y King JG Kisvarday Z Kumbhar P Lasserre S Le Be JV Magalhaes BR Merchan-Perez A Meystre J Morrice BR Muller J Munoz-Cespedes A Muralidhar S Muthurasa K Nachbaur D Newton TH Nolte M Ovcharenko A Palacios J Pastor L Perin R Ranjan R Riachi I Rodriguez JR Riquelme JL Rossert C Sfyrakis K Shi Y Shillcock JC Silberberg G Silva R Tauheed F Telefont M Toledo-Rodriguez M Trankler T Van GW Diaz JV Walker R Wang Y Zaninetta SM DeFelipe J Hill SL Segev I Schurmann F Reconstruction and simulation of neocortical microcircuitry Cell 2015 163 456 492 10.1016/j.cell.2015.09.029 26451489 149. Ko H Hofer SB Pichler B Buchanan KA Sjostrom PJ Mrsic-Flogel TD Functional specificity of local synaptic connections in neocortical networks Nature 2011 473 87 91 10.1038/nature09880 21478872 150. Marin O Interneuron dysfunction in psychiatric disorders Nat Rev Neurosci 2012 13 107 120 22251963 151. Chattopadhyaya B Cristo GD GABAergic circuit dysfunctions in neurodevelopmental disorders Front Psychiatry 2012 3 51 10.3389/fpsyt.2012.00051 22666213 152. Meredith RM Dawitz J Kramvis I Sensitive time-windows for susceptibility in neurodevelopmental disorders Trends Neurosci 2012 35 335 344 10.1016/j.tins.2012.03.005 22542246 153. Sale A Maya Vetencourt JF Medini P Cenni MC Baroncelli L De PR Maffei L Environmental enrichment in adulthood promotes amblyopia recovery through a reduction of intracortical inhibition Nat Neurosci 2007 10 679 681 10.1038/nn1899 17468749 154. Sale A Putignano E Cancedda L Landi S Cirulli F Berardi N Maffei L Enriched environment and acceleration of visual system development Neuropharmacology 2004 47 649 660 10.1016/j.neuropharm.2004.07.008 15458836 155. Huang S Gu Y Quinlan EM Kirkwood A A refractory period for rejuvenating GABAergic synaptic transmission and ocular dominance plasticity with dark exposure J Neurosci 2010 30 16636 16642 10.1523/JNEUROSCI.4384-10.2010 21148002 156. He HY Hodos W Quinlan EM Visual deprivation reactivates rapid ocular dominance plasticity in adult visual cortex J Neurosci 2006 26 2951 2955 10.1523/JNEUROSCI.5554-05.2006 16540572 157. Maya Vetencourt JF Sale A Viegi A Baroncelli L De PR O’Leary OF Castren E Maffei L The antidepressant fluoxetine restores plasticity in the adult visual cortex Science 2008 320 385 388 10.1126/science.1150516 18420937 158. Oh SW Harris JA Ng L Winslow B Cain N Mihalas S Wang Q Lau C Kuan L Henry AM Mortrud MT Ouellette B Nguyen TN Sorensen SA Slaughterbeck CR Wakeman W Li Y Feng D Ho A Nicholas E Hirokawa KE Bohn P Joines KM Peng H Hawrylycz MJ Phillips JW Hohmann JG Wohnoutka P Gerfen CR Koch C Bernard A Dang C Jones AR Zeng H A mesoscale connectome of the mouse brain Nature 2014 508 207 214 10.1038/nature13186 24695228

PMC005xxxxxx/PMC5002042.txt

==== Front Mamm GenomeMamm. GenomeMammalian Genome0938-89901432-1777Springer US New York 965610.1007/s00335-016-9656-5ArticleUncovering the liver’s role in immunity through RNA co-expression networks Harrall Kylie K. 12Kechris Katerina J. 2Tabakoff Boris 1Hoffman Paula L. 3Hines Lisa M. 4Tsukamoto Hidekazu 56Pravenec Michal 7Printz Morton 8Saba Laura M. (303) 724-9697laura.saba@ucdenver.edu 11 Department of Pharmaceutical Sciences, University of Colorado Skaggs School of Pharmacy and Pharmaceutical Sciences, Aurora, CO 80045 USA 2 Department of Biostatistics and Informatics, University of Colorado School of Public Health, Aurora, CO 80045 USA 3 Department of Pharmacology, University of Colorado School of Medicine, Aurora, CO 80045 USA 4 Department of Biology, University of Colorado at Colorado Springs, Colorado Springs, CO 80918 USA 5 Department of Pathology, Southern California Research Center for ALPD and Cirrhosis, Keck School of Medicine of USC, University of Southern California, Los Angeles, CA 90089 USA 6 Department of Veterans Affairs, Great Los Angeles Healthcare System, Los Angeles, CA 90089 USA 7 Department of Model Diseases, Institute of Physiology, Czech Academy of Sciences, Prague, Czech Republic 8 Department of Pharmacology, University of California, San Diego, La Jolla CA, 92093 USA 11 7 2016 11 7 2016 2016 27 9 469 484 25 4 2016 27 6 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Gene co-expression analysis has proven to be a powerful tool for ascertaining the organization of gene products into networks that are important for organ function. An organ, such as the liver, engages in a multitude of functions important for the survival of humans, rats, and other animals; these liver functions include energy metabolism, metabolism of xenobiotics, immune system function, and hormonal homeostasis. With the availability of organ-specific transcriptomes, we can now examine the role of RNA transcripts (both protein-coding and non-coding) in these functions. A systems genetic approach for identifying and characterizing liver gene networks within a recombinant inbred panel of rats was used to identify genetically regulated transcriptional networks (modules). For these modules, biological consensus was found between functional enrichment analysis and publicly available phenotypic quantitative trait loci (QTL). In particular, the biological function of two liver modules could be linked to immune response. The eigengene QTLs for these co-expression modules were located at genomic regions coincident with highly significant phenotypic QTLs; these phenotypes were related to rheumatoid arthritis, food preference, and basal corticosterone levels in rats. Our analysis illustrates that genetically and biologically driven RNA-based networks, such as the ones identified as part of this research, provide insight into the genetic influences on organ functions. These networks can pinpoint phenotypes that manifest through the interaction of many organs/tissues and can identify unannotated or under-annotated RNA transcripts that play a role in these phenotypes. Electronic supplementary material The online version of this article (doi:10.1007/s00335-016-9656-5) contains supplementary material, which is available to authorized users. 10.13039/100000027National Institute on Alcohol Abuse and AlcoholismAAU01 Developmental Grant-INIA ProjectR24AA01316210.13039/100000050National Heart, Lung, and Blood InstituteHL35018Banbury Fundissue-copyright-statement© Springer Science+Business Media New York 2016 ==== Body Introduction With the emergence of the fields of precision medicine and systems genetics, the need for animal models and statistical methods to facilitate the translation of research from bench to bedside has grown (Malaney et al. 2014; Ritchie 2012). Identification of determinants of genetic susceptibility to disease and environmental toxins, and the integration of pharmacogenomic information are critical components of the NIH initiative to develop precision human medicine (Riley et al. 2015). However, genetic studies in humans are very costly in both time and resources; some of these costs can be attributed to the requirement of large sample sizes and the lack of a priori control over genetic and environmental variables. In addition, many human studies cannot be performed for ethical reasons. Rodent models not only provide opportunities to control and manipulate genetics and environmental factors, but they also provide the opportunity to investigate biological processes that cannot ethically be studied in humans. Rats are a prominent model for studying the mechanisms of disease, the impact of environmental factors, and new drug technologies (Aitman et al. 2008). Genetic models, rapid advances in sequencing technologies, and recent gene-editing techniques have allowed rats to remain a major resource for functional and mechanistic studies in medicine (Parker et al. 2014). Historically, investigators have often used a single inbred strain or rodents from an outbred stock to examine biological mechanisms of disease (Festing 2014). Another approach, which has distinct advantages for systems genetic studies, is to use a panel of inbred or recombinant inbred (RI) rat strains (Printz et al. 2003). In such a panel, the varied, but known, genetic composition of the panel is relatively static and retained over generations. This ‘reproducibility’ of genetic background allows for the identification of interactions among multiple phenotypes, and their relationships to genetic variation. The accumulation of multiple behavioral, physiological, and molecular phenotypes over generations and across laboratories is essential for truly integrative systems genetics research. One type of molecular phenotype that is of particular interest in systems genetics research is RNA transcript abundance. By quantitatively measuring RNA expression levels across an inbred rat panel, one can attempt to describe a functional relationship between genes through co-expression. The general theory of co-expression studies is that if the expression levels of two transcripts react to different genetic backgrounds in a similar manner, then the two transcripts are likely involved in a similar biological process (Allocco et al. 2004). One can use graph theory to describe the relationships among genes based on co-expression across different genetic backgrounds (such as in RI panels). For example, these co-expression networks can be modeled as robust scale-free gene networks (Ravasz et al. 2002; Weiss et al. 2012). Transcripts within a scale-free network are not connected at random, rather these networks are composed of many nodes with few connections, and a few hub nodes, which are connected to many other transcripts. The crux of our work is based on the proposition that if a genetic locus influences both a behavioral/physiological trait and RNA expression levels in an identified transcriptional network, then the transcriptional network is likely to influence the behavioral/physiological trait (Saba et al. 2015). In the current work, we sought to answer one question: can one use information about the genetic locus of expression variation in a module (eigengene QTL), from a newly described co-expression network, to search phenotypic quantitative trait loci (QTL) databases for overlapping behavioral/physiologic QTLs and arrive at an informative relationship between a physiological/behavioral phenotype and the transcriptome. We utilized RI rats and quantitative systems genetics methods (network analysis of RNA expression levels) to group both well-annotated, under-annotated, and unannotated RNA transcripts into biologically relevant networks, and to link these networks to phenotypic traits through shared genetic influences (i.e., overlapping phenotypic and expression QTL). We focused on the liver transcriptome of the HXB/BXH RI rat panel. Our approach, outlined in Fig. 1, contained five steps: (1) define the liver transcriptome from the progenitor strains of the HXB/BXH RI rat panel (SHR and BN-Lx), (2) measure expression levels of the liver transcriptome in the HXB/BXH RI rat panel, (3) identify co-expression modules within the liver, (4) identify modules exhibiting high levels of genetic control, and (5) identify the biological relevance of genetically driven modules.Fig. 1 An outline of systems genetics approach. This workflow was used to identify genetically driven and biologically relevant liver modules This approach led to the discovery of two genetically driven co-expression liver modules that were associated with the immune system. Although the liver is often associated with metabolism, it has many roles in the body (Higuchi and Gores 2003). This research has uncovered a genetic pathway associated with the lesser known function of liver, innate immunity (Gao et al. 2008; Racanelli and Rehermann 2006), and demonstrates how system genetics approaches, such as the one discussed in this manuscript, have the potential to aid in the discovery of genetic influences on biological functions and to expand current research directions. Materials and methods Animals RI panels are produced by first crossing two genetically and phenotypically diverse inbred strains to create an F1 generation with one chromosome copy from each parent. At the F2 generation, as a result of recombinant events, each animal’s DNA sequence becomes a mosaic of the original DNA sequences of the two parental strains. After many generations of brother/sister mating, an RI panel is produced that contains a large number of inbred strains. All rats within a strain are genetically identical (similar to monozygotic twins), and that genetic identity is retained over generations. Rats from different strains within the panel share approximately 50 % of the genetic sequence of the parental strains, much like dizygotic twins or siblings (Printz et al. 2003). In this study, we assessed rats from the HXB/BXH RI panel, which was originally developed in the Czech Republic by Vladimir Kren (Institute of Biology of Charles University) and Michal Pravenec (Institute of Physiology of the Czech Academy of Sciences). This panel was generated using gender reciprocal crossing between the congenic Brown Norway strain with polydactyly-luxate syndrome (BN-Lx/Cub) and the spontaneous hypertensive rat strain (SHR/OlaIpcv), with sixty generations of brother/sister mating after the F2 generation (Printz et al. 2003). Although this panel was originally developed to examine differences in cardiovascular traits, many other traits vary across the strains and have been studied extensively (Bielavská et al. 2002, 2002; Conti et al. 2004; Kunes et al. 1994; Pravenec et al. 2004; Vanderlinden et al. 2014). Identification of RNA transcripts in the liver Whole liver RNA sequencing Male rats from substrains of the original parental strains of the HXB/BXH RI panel, BN-Lx/CubPrin (BN-Lx) and SHR/OlaIpcvPrin (SHR) were used for RNA-Seq analyses. Rats were maintained and bred at the University of California, San Diego. Three male rats from each parental strain (70–90 days old) were quickly anesthetized with isofluorane/air and decapitated according to a protocol approved by the UCSD IACUC. Livers were quickly frozen in liquid nitrogen, and they were shipped to the University of Colorado for RNA extraction and library preparation. For each of the six liver samples, total RNA was extracted with the RNeasy Midi Kit; additional cleanup was performed using the RNeasy Mini Kit (Qiagen, Valencia, CA, USA). External RNA Controls Consortium (ERCC) Synthetic Spike-Ins (ThermoFisher Scientific, Wilmington, DE, USA) were added to the extracted RNA; 4 µl was added from a 1:100 dilution of either Mix 1 or 2. Sequencing libraries were constructed using the Illumina TruSeq Stranded RNA Sample Preparation kit (Illumina, San Diego, CA, USA), in accordance with the manufacturer’s protocol, and library quality was assessed using the Agilent Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA). Ribosomal RNA depletion was carried out as part of the Illumina TruSeq Stranded Total RNA Library Prep Kit, which includes the Ribo-Zero ribosomal RNA reduction chemistry. Samples were sequenced (2x100 paired-end reads) on the Illumina HiSeq2000 (Illumina, San Diego, CA, USA) with two samples multiplexed per lane. For each of the livers of the rats from the BN-Lx and SHR parental strains, the raw reads were trimmed for quality and adaptors using trim-galore (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/, version 0.4.0). The trimmed reads were aligned using Tophat2 (Kim et al. 2013) (Version 2.1.0) with default parameters. The reads were initially aligned to rRNA from the RepeatMasker database (Smit et al. 1996), which was downloaded from the UCSC Genome Browser (Karolchik et al. 2004); all of the paired-end reads with either end aligned to those sequences were eliminated from the analysis. If reads did not align to rRNA, they were aligned to the rn5 version of their respective strain-specific genomes. The strain-specific genomes were constructed with DNA-Seq data from male brains of the BN-Lx/CubPrin and SHR/OlaIpcvPrin strains (Hermsen et al. 2015), and they are publicly available on the PhenoGen website [http://phenogen.ucdenver.edu (Saba et al. 2015)]. Transcriptome reconstruction in progenitor strains For reconstructing the rat liver transcriptome, reads from all biological replicates within an inbred strain were combined. Strain-specific transcriptomes were then reconstructed using the Cufflinks algorithm (Trapnell et al. 2012) (Version 2.2.1), with the rn5 Ensembl Rat Transcriptome (Roberts et al. 2011; Trapnell et al. 2010) as a guide. Mitochondrial chromosomes were initially masked due to the depth of reads within this small chromosome, which would dramatically increase computational burden. Within the strain-specific transcriptomes, “high-confidence transcripts” (defined below) were identified using estimated expression levels and transcript length. Fragments per kilobase of transcript per million mapped reads (FPKM) were compared across transcripts, and a threshold was set to filter out low confidence transcripts. FPKM values were calculated from the combined reads within a strain. The focus was to qualitatively describe the transcriptome without the restriction that the transcript be expressed in multiple animals. Along with the FPKM threshold, transcripts had to exceed a transcribed length of 300 base pairs (bp); this was the length required to survive size selection during RNeasy RNA extraction and the gel-size selection step within the Illumina TruSeq protocol. For mapping and annotation, the resulting high-confidence BN-Lx and SHR transcriptomes were merged to create a single combined transcriptome. Ensembl-annotated mitochondrial transcripts were reintroduced at this stage, so that these transcripts were included in the final mapped transcriptome. Overlapping transcripts between the two strain-specific transcriptomes were identified and merged into a single transcript if they occurred on the same strand (or one strand was not designated), and they met at least one of the following requirements: (1) all of the splice junctions matched, or (2) both transcripts contained only one exon, and when the two transcripts were compared, their transcription start and stop sites were within 100 bp of each other. After merging transcripts that overlapped, transcripts (i.e., splice variants) were grouped into genes. We determined that two transcripts were from the same gene if their transcription start sites matched exactly, their transcription stop sites matched exactly, or at least one splice junction matched exactly. Because the Cufflinks algorithm accommodates reads aligned to multiple transcripts by splitting their read count among the different transcripts (often multiple splice variants of the same gene) in a probabilistic manner, the addition of an alternative transcript to a transcriptome may alter the expression estimate of the original transcript(s). To identify the most robust set of transcripts, the combined transcriptome of high-confidence transcripts was quantitated separately in each strain. These transcripts were quantitated using coverage, i.e., the average number of reads per nucleotide across the whole transcript. If a transcript’s coverage dropped below an average of 50 reads per nucleotide in each of the strains when considering all transcripts in the combined transcriptome, it was dropped from the combined transcriptome. This iteration of quantitation and reduction of the transcriptome was repeated until less than 5 % of the remaining transcripts fell below our coverage threshold. Quantitation of RNA expression in the livers of the HXB/BXH recombinant inbred panel Microarray expression measurements Male rats from 21 strains of the HXB/BXH RI panel were used for exon array analyses. Like the parental rat strains used for RNA-Seq, rats from the RI panel were maintained and bred at the University of California, San Diego (UCSD). Three to four male rats from each strain (70–90 days old) were quickly anesthetized with isofluorane/air and decapitated according to a protocol approved by the UCSD IACUC. Livers were quickly extracted and frozen on dry ice or in liquid nitrogen. Livers were shipped to the University of Colorado for RNA extraction and microarray processing. RNA was extracted from each whole liver using the RNeasy Midi Kit (Qiagen) and the RNeasy Mini kit (Qiagen) for cleanup. cDNA from the liver of each individual rat was hybridized to a separate Affymetrix Rat Exon 1.0 ST array following the manufacturer’s protocol. All processed array data were examined for quality using the tools outlined in detail on the PhenoGen website. Probe sequences for the Affymetrix Rat Exon 1.0 ST Array were obtained from the Affymetrix website (http://www.affymetrix.com/). High-integrity probes were identified based on their alignment to the genome and known genomic variants in the BN-Lx and SHR strains (Saba et al. 2015). For example, probes were removed from the analysis if they did not align perfectly and singularly to the rn5 rat genome. Probes were also removed if they aligned to a region that harbored an SNP, small insert, or small deletion between the BN-Lx or SHR strains, when compared to the Brown Norway (BN) reference genome (Hermsen et al. 2015). Entire probe sets were eliminated if less than three probes remained after filtering. This array mask is publicly available on the PhenoGen website. The full set of probe sets on the Affymetrix array targets many different regions of the genome; this include annotated genes, and unannotated regions with sequence characteristics that predict possible transcription. This breadth of coverage allows for the quantitation of both annotated and unannotated transcripts in the reconstruction. To do this, probe sets from the exon array were mapped to transcripts in the reconstructed liver transcriptome. Probe sets were retained if their targeted region was exclusively and entirely within a gene. This allowed for the selection of probe sets specific for genes that are expressed in the liver of naïve BN-Lx and SHR rats, including probe sets that targeted unannotated genes. All of the probes that targeted a particular gene were grouped into a single expression estimate, which is called a gene cluster throughout this manuscript. The mapping of probe sets to gene clusters is available on the PhenoGen website. Genes with at least one “high-confidence” transcript in the reconstruction, which precisely matched a transcript annotated in the rn5 Ensembl Rat Transcriptome, were associated with gene information for that annotated transcript. Affymetrix Power Tools (http://www.affymetrix.com/) was used to implement an RMA-sketch algorithm to estimate expression levels for individual gene clusters. Correction for batch effects was executed using the ComBat algorithm (Johnson et al. 2007). We required that at least 5 % of samples had expression above background (the detection above background (DABG) p value <0.0001) for a gene cluster to be retained for further analyses. Network analysis Weighted gene correlation network analysis (WGCNA) Strain means of gene clusters were used for co-expression analysis. Each gene cluster represents a singular gene product, which is defined as all of the RNA transcripts derived from a particular gene. An unsigned weighted gene correlation network was constructed using the R package WGCNA (Langfelder and Horvath 2008) (version 1.49). A co-expression similarity matrix was derived by calculating pairwise Pearson correlation coefficients between the expression levels of genes across strains. This similarity matrix was subsequently transformed into a scale-free connectivity network by raising each correlation coefficient to a power of 7; this threshold was chosen using criteria and methods from Langfelder, Horvath, and Zhang (Langfelder and Horvath 2008; Zhang and Horvath 2005) for modeling a scale-free network. In an effort to avoid spurious connections within our networks, a topological overlap map (TOM) was used to assess both the direct connection between two gene products and the indirect connection between the two gene products based on the similarity of their relationships with other gene products (Langfelder 2013). The minimum module size was set at 5 to allow for smaller modules, compared to the default setting of 30; subsequently, modules were merged if their eigengenes were highly correlated using a threshold of 0.75. Gene product co-expression modules were identified with dynamic tree cutting (Langfelder et al. 2008); this method identifies large clusters of gene products with an initial static cut, then refines the large clusters by recursively splitting them into subclusters. Principal Components Analysis was used to calculate an eigengene (the first principal component) for each co-expression module. Each eigengene represents the expression pattern across strains for the module, and the accuracy of these representations can be quantified as the proportion of variance of gene expression across strains explained by the eigengene. Within each module, we calculated the intramodular connectivity and the number of connections for each gene product (Zhang and Horvath 2005). Intramodular connectivity was defined as the sum of connection strengths with gene products within a module associated with a gene product of interest, and the gene product with highest connectivity within each module was identified as the hub gene product. The number of intramodular connections of a gene product represents the number of gene products within the module that are correlated with that gene product (correlation coefficient > |0.5|). Selection of genetically driven modules To identify genetically driven modules in the liver, modules were required to have: (1) an eigengene that explained over 50 % of module variance (Langfelder et al. 2011), and (2) a significant module eigengene quantitative trait loci (meQTL), i.e., genome-wide p value <0.05. meQTLs were identified by testing for an association between a module eigengene and a set of genomic markers (i.e., strain distribution patterns) for the HXB/BXH panel. This set of markers was downloaded from PhenoGen, and further details are available in the study by Saba et al. (2015). Associations were calculated with R/qtl (Broman et al. 2003) (Version 1.28–4) using marker regression with a genome-wide LOD significance threshold based on 1000 permutations (Churchill and Doerge 1994). Identifying biological relevance of modules We used two independent public databases to uncover the biological function of each genetically driven module. First, we used PANTHER (http://pantherdb.org/) to find significant enrichment for pathways or gene ontology (GO) terms among the annotated gene products of the module (Mi et al. 2016). If a module was significantly enriched for a GO term or PANTHER pathway, we used the Rat Genome Database’s (RGD) QTL search (Shimoyama et al. 2015) (http://rgd.mcw.edu/) to find phenotypic QTLs in the same genomic location as the meQTL. PANTHER gene ontology and enrichment PANTHER’s statistical overrepresentation test (Mi et al. 2013) was used to identify enrichment (adjusted p < 0.05) of PANTHER pathways and GO terms, which were categorized as either related to biological processes, molecular functions, or cellular components. Because this type of analysis increases in power as more gene products are included, we required that each module contain at least ten Ensembl annotated gene products. This type of analysis identifies terms or pathways that are statistically overrepresented by gene products represented within a module. PANTHER uses the binomial distribution to compare a set of reference genes to the genes that occur in each module; in this case, our reference set was constrained to the genes included in the network analysis. A Bonferroni adjustment was applied within PANTHER to each GO term or pathway that occurred in the overrepresentation analysis. When many gene products within a module share a common GO term or are from the same annotated pathway, one can postulate that the module’s collective function is related to the ontology term or pathway. The statistical test for enrichment can produce a significant result if only one gene product from the module is associated with a GO term or pathway when the probability of even one gene product from the term/pathway appearing in the module by chance is extremely low. Because of this, we implemented an additional criterion that more than one gene product within the module needed to be associated with the term/pathway. Phenotypic QTLs To investigate the etiologic relationship between liver modules and genetic factors that may predispose animals to specific phenotypes, phenotypic QTLs were identified using the RGD’s QTL search. RGD contains a repository of genetic information for the rat species; as of 2015, the website contained records for 53,345 genes, 108,875 transcripts, and 2163 QTLs (Shimoyama et al. 2015). We considered all QTLs that were documented to overlap the peak of each module’s significant meQTL. We set a high standard for potentially relevant QTLs with a LOD score ≥10. Since this QTL data source is curated from published rat QTL papers, precise genome-wide p values are often not reported. Additional characterization of modules Partial correlation In the event that a module of interest had many gene products physically located within the same region of the genome, we assessed the biological nature of our findings with partial correlation. Associations between two co-localized gene products are more likely to share a biological function/process, if the partial correlation remains significant when accounting for an associated cis-locus (Supplementary Fig. S1). Pairwise correlations were recalculated between gene clusters, while controlling for each module’s meQTL. Connectivity and the number of relevant connections (r > |0.5|) were re-assessed under the partial correlation model. Liver cell type-specific transcriptome analysis Alcohol-naïve adult male BN-Lx/CubPrin and SHR/OlaPrin rats were shipped to the Integrative Liver Cell Core at University of Southern California for separation of liver cell types. Kupffer cells (KC) and hepatic stellate cells (HSC) were isolated by sequential digestion of rat liver with pronase and collagenase followed by low-speed centrifugation and discontinuous arabinogalactan gradient ultracentrifugation. This procedure has been described in detail by Kamimura and Tsukamoto (1995), and by Tsukamoto et al. (1995). Sinusoidal endothelial cells (SEC) were isolated by collagenase perfusion, density gradient centrifugation, and centrifugal elutriation as previously described by Deleve (1994). Primary hepatocyte cells (HC) were isolated aseptically according to the method of Moldéus et al. (1978). This method is based on collagenase digestion and separation of liver parenchymal cells. Fresh cells were shipped to the University of Colorado Denver for RNA extraction and microarray processing. RNA from four rats per strain and cell type was extracted and hybridized to separate Affymetrix Rat Exon Arrays 1.0 ST (Affymetrix) in the same manner as described previously for the whole liver in the HXB/BXH panel. To make gene clusters comparable to our earlier analyses, the same methods used in the whole liver analysis were applied to the cell type-specific arrays; specifically, arrays were processed using the same probe mask and normalized with the same algorithm and batch corrections. A two-way ANOVA model was used to estimate expression means for each cell type and strain, and differential expression between cell types (either strain dependent or independent) was determined using an F-statistic and a false discovery rate (FDR) to correct for multiple comparisons. Further, genes were subset by module for pairwise cell type comparisons. Results Identification of RNA transcripts in the liver Whole liver RNA sequencing RNA isolated from each of the three livers of each strain (BN-Lx and SHR) was prepared and separately sequenced. We obtained 311 million paired-end reads for BN-Lx and 314 million for SHR. After trimming and the removal of reads that aligned to ribosomal RNA (rRNA), 224 million paired-end reads (72.0 %) from the BN-Lx rats aligned to the BN-Lx strain-specific genome and the synthetic spike-ins. 256 million paired-end reads (71.5 %) from the SHR aligned to the SHR strain-specific genome and the synthetic spike-ins. Synthetic spike-ins were included to improve normalization and batch correction. Transcriptome reconstruction in progenitor strains Within each progenitor strain transcriptome, an FPKM (based on combined reads within a strain) threshold of 1 was set to define a high-confidence transcript (Supplementary Fig. S2). For BN-Lx, we retained 17,343 transcripts (14 % of the total transcripts), with FPKM ranging from 1 to 11,952. Similarly in SHR, we retained 14,023 transcripts (16 % of the total transcripts), with FPKMs ranging from 1 to 5293. After the high-confidence transcripts of the BN-Lx and SHR transcriptomes were combined and iteratively quantitated with CuffLinks, 18,260 transcripts (14,201 genes) remained; 6671 transcripts were annotated in the Ensembl database (Birney et al. 2004) and 11,589 were not included in the rn5 version of the Ensembl database (Table 1). It is important to note that using an RNA-Seq-derived transcriptome to select transcripts, we were able to identify many unannotated isoforms of annotated genes with respect to the Ensembl database,Table 1 Summary of rat liver transcriptome reconstruction and RNA-Seq-guided microarray mask Strain Data type Unique gene IDs Total transcripts Ensembl transcripts Transcripts not annotated in the Ensembl database BN-Lx Complete transcriptome 123,143 143,107 28,836 114,271 High-confidence transcripts 13,593 17,343 6998 10,345 SHR Complete transcriptome 83,864 102,825 28,822 74,003 High-confidence transcripts 10,321 14,023 7136 6887 BN-Lx & SHR combined Transcriptome 14,201 18,260 6671 11,589 rn5 Affymetrix exon array gene clusters included in network analysis 9223 13,111 5994 7117 RNA-Seq high-confidence transcripts were required to have FPKM > 1 and length > 300 bp Quantitation of RNA expression in the livers of the HXB/BXH recombinant inbred panel Microarray expression measurements The same probe mask described by Saba et al. (2015) was used to filter the 4.1 million probes on the Affymetrix Rat Exon 1.0 ST Array to 3.8 million ‘high-integrity’ probes (890,607 probe sets). When these probe sets were mapped to the combined transcriptome of both the BN-Lx and SHR rats, 146,473 probe sets were contained entirely within liver transcripts that were identified from the RNA-Seq data. Specifically, 9847 gene products from the liver transcriptome were targeted by at least 1 probe set. All probe sets targeting a singular gene product were summarized into a gene cluster. After filtering out gene clusters that did not meet our detection above background criteria, 9223 gene clusters remained in our analysis (Table 1). Selection of genetically driven modules Using the 9223 gene clusters, we identified 50 co-expression modules: each contained between 5 and 3369 gene products (median module size = 22.5). We calculated an eigengene for each of the modules to represent the expression of genes within each module. The eigengenes accounted for between 43 and 81 % of the variance within each module (median 58 %). Fourteen modules had significant meQTL genome-wide LOD scores, based on a genome-wide p value <0.05, and all of the eigengenes for these modules explained at least 57 % of the respective module variance (Table 2).Table 2 Genetically driven co-expression modules in rat liver Module Genes Eigengene meQTL Number of genes in module Number of Ensembl annotated genes Proportion of variance in module explained by eigengene Location LOD Genome-wide p value Thistle 19 14 0.56 Chr1: 177 Mb 8.47 0.004 Yellow3 15 8 0.60 Chr1: 282 Mb 8.65 0.004 Plum 23 11 0.60 Chr2: 223 Mb 6.67 0.016 Palevioletred2 18 10 0.58 Chr2: 248 Mb 5.69 0.008 Indianred3 10 6 0.58 Chr5: 155 Mb 4.54 0.048 Lightcyan 119 67 0.58 Chr6: 6.6 Mb 3.28 0.041 Firebrick4 45 17 0.59 Chr8: 7.4 Mb 5.06 0.005 Plum2 38 22 0.67 Chr10: 26 Mb 4.47 0.017 Lightpink2 5 3 0.81 Chr10: 67.2 Mb 8.50 0.009 Tan4 9 3 0.67 Chr10: 87 Mb 7.51 <0.001 Magenta3 5 3 0.75 Chr15: 107 Mb 8.48 0.004 Pink4 13 6 0.57 Chr16: 36 Mb 6.11 0.005 Lightblue4 11 7 0.64 Chr18: 53.8 Mb 7.85 0.007 Thistle2 37 21 0.58 Chr20: 8.0 Mb 6.39 0.004 Fourteen modules had significant module eigengene QTLs (meQTLs; genome-wide p < 0.05), and a proportion of variance in the module explained by the eigengene greater than 0.5. Although partial correlation was performed for the thistle2 module; these numbers reflect the number of genes that occurred prior to partial correlation Identifying biological relevance of modules Seven of the modules contained at least ten Ensembl annotated gene products and, of these, four of the genetically driven modules were significantly enriched (adjusted p value <0.05) for at least one GO term or PANTHER pathway: thistle2, plum2, firebrick4, and lightcyan. To find physiological or behavioral traits associated with these modules, we used the RGD database to identify physiological and behavioral QTLs that overlapped the meQTLs of our significantly enriched modules (Table 3).Table 3 Genetically driven liver co-expression modules with inferred biological relevance Enrichment Module GO term/PANTHER pathway Smallest p value Name LOD References Thistle2 Antigen processing and presentationa, antigen processing and presentation of peptide or polysaccharide via MHC-class II, cellular defense response, MHC protein complex, Antigen binding 9.80 × 10−17 CIA autoantibody 39.9 Furuya et al. (2000) Food consumption 20.7 Marissal-Arvy et al. (2014a) Serum corticosterone level 20.5 Marissal-Arvy et al. (2014b) Adjuvant-induced arthritis 18.0 Kawahito et al. (1998), Joe 2006 Plum2 Immune responsea 4.16 × 10−2 Food consumption 18.7 Marissal-Arvy et al. (2014a) Insulin level 10.5 Marissal-Arvy et al. (2014b) Lightcyan Receptor bindinga, G protein-coupled receptor activity, actin binding 1.67 × 10−3 Bone mineral density 11.7 Koller et al. (2008) Bone mineral density 11.2 Koller et al. (2008) Firebrick4 Vesicle-mediated transporta, protein transport 2.88 × 10−3 Insulin/glucose ratio 18.5 Marissal-Arvy et al. (2014b) Summary of the results obtained from two public resources: PANTHER Gene List Analysis for PANTHER pathway and GO term enrichment, and the Rat Genome Database QTL search aMarks the enrichment category that corresponds to the smallest p value Thistle2 Twenty-three of the thistle2 module’s 37 gene products (Supplementary Table 1) were associated with aspects of the immune system according to gene ontology, and the module was highly enriched for several GO terms that are related to the immune system. Ten gene products from thistle2 were associated with antigen processing and presentation (GO: 0019882); this represents a 111-fold enrichment for this category (adjusted p value = 9.80 × 10−17). Of those ten gene products, two were specifically associated with antigen processing and presentation of peptide or polysaccharide antigen via MHC-class II (GO: 0002504) representing 100-fold enrichment for the category (adjusted p value = 3.13 × 10−2). The thistle2 module was also enriched for several other biological, cellular, and molecular gene ontology terms. Under biological processes, four genes were associated with cellular defense response (GO: 0006968), representing 22-fold enrichment for the category (adjusted p value = 5.50 × 10−3). For the cellular component, two genes were associated with the MHC protein complex (GO0042611), representing a 200-fold enrichment (adjusted p value = 3.64 × 10−3). Finally, with respect to molecular function, eight genes were associated with antigen binding (GO: 0003823), representing a 133-fold enrichment (adjusted p value = 2.17 × 10−13). Gene products that are involved in these processes are represented as yellow nodes (circles) in Fig. 2.Fig. 2 Connectivity within the thistle2 module. a Initial pairwise connections between gene products in the thistle2 co-expression module. b Pairwise connections that remained after adjustment for the module eigengene QTL, i.e., potential shared cis eQTL among genes. A red edge represents a negative correlation between the two nodes and a blue edge represents a positive correlation between two nodes. The thickness of each edge represents the strength of each correlation (i.e., a correlation with a larger magnitude has a thicker line), and edges are only visible if their associated correlation is greater than |0.60|. Unannotated genes are represented by chromosomal location. Genes are marked with an asterisk if none of its associated high-confidence transcripts were annotated in the Ensembl database, but at least one is similar to an annotated transcript from the rat Ensembl database, the rat RefSeq database, or the RefSeq database from other species. Yellow nodes represent gene products that were involved in module enrichment (immune system, antigen processing and presentation, cellular defense response, MHC protein complex, and antigen binding). RT-CE16, a Class I Major Histocompatibility Complex, is the hub gene product both before (a) and after (b) adjustment for shared cis eQTL The top five most highly connected gene products in this module function within the immune system. The hub is an RT1-class 1 gene product (RT1-CE16) that functions as an MHC-class I molecule. The remaining gene products include 2 other MHC-class I-related molecules, a gene product that processes MHC-class I molecules, and the lymphotoxin beta gene (TNF superfamily, member 3). A total of 24 phenotypic QTLs overlapped thistle2’s meQTL; four of these met our criteria: collagen-induced arthritis autoantibody (Ciaa1), adjuvant-induced arthritis (Aia1), food consumption (Foco23), and serum corticosterone levels (Scort12). Two of the overlapping phenotypic QTLs are associated with autoimmunity. The collagen-induced arthritis autoantibody (Ciaa1) phenotype is the IgG autoantibody titer against native rat type II collagen (Furuya et al. 2000). This phenotype is a model for rheumatoid arthritis; rats with higher susceptibility to collagen-induced arthritis have higher levels of the IgG autoantibody, which is a biomarker for the autoimmune disease. Adjuvant-induced arthritis is another model for rheumatoid arthritis. The adjuvant-induced arthritis phenotype (Aia1) is associated with disease severity; measures of the arthritis index were used to determine severity of the disease (Joe 2006; Kawahito et al. 1998). The remaining QTLs were associated with food preference and basal corticosterone levels. The food consumption phenotype (Foco23) is a reflection of macronutrient selection and weight gain. Marissal-Atvy et al. (2014a) allowed rats to self-select from different macronutrient-driven diets (i.e., protein-, carbohydrate-, and fat-rich diets), and documented diet preference and the resulting weight gain. This phenotypic QTL is associated with carbohydrate preference in female rats that have differential macronutrient preferences. Marissal-Atvy et al. (2014b) were also interested in phenotypes that were related to genetic variability within the hypothalamic–pituitary–adrenocortical (HPA) axis, so they investigated the overlap between QTLs associated with the function of the HPA and phenotypes that were associated with carbohydrate metabolism. One of these metabolism-related phenotypes was basal state serum corticosterone levels (Scort12). Plum2 There were 38 gene products in the plum2 module (Supplementary Table 2), and 22 of those gene products were Ensembl annotated. The module hub gene product was interferon regulatory factor 7 (Irf7), and at least 14 gene products within the plum2 module were related to interferon regulation, response to various types of interferon, defense to virus, and immune response. The module is significantly enriched for the immune response (GO: 0006955); it had a 13-fold enrichment for the category (adjusted p value = 4.16 × 10−2). Gene products that are involved in these processes are represented as yellow nodes (circles) in Fig. 3.Fig. 3 Connectivity within the plum2 module. Initial pairwise connections between gene products in the plum2 co-expression module. A red edge represents a negative correlation between two nodes and a blue edge represents a positive correlation between two nodes. The thickness of each edge represents the strength of each correlation (i.e., a correlation with a larger magnitude has a thicker line), and edges are only visible if their associated correlation is greater than |0.60|. Genes are marked with an asterisk if none of its associated high-confidence transcripts were annotated in the Ensembl database, but at least one is similar to an annotated transcript from the rat Ensembl database, the rat RefSeq database, or the RefSeq database from other species. Yellow nodes represent gene products that were involved in module enrichment (immune response). Irf7, Interferon regulatory factor 7, is the hub gene product We identified 35 phenotypic QTLs that overlapped the plum2 module’s meQTL; 2 of these phenotypic QTLs met our criteria: food consumption (Foco17) and plasma insulin level (Insul27). Like Foco23 (food consumption phenotype in thistle2), the food consumption phenotype overlapping plum2 (Foco17) is also associated with carbohydrate preference in female rats. While investigating the association between the HPA and metabolism, Marissal-Atvy et al. (2014b) found the plasma insulin level QTL (Insul27). Lightcyan There are 119 gene products in the lightcyan module; 67 gene products were Ensembl annotated. The hub gene product is prostaglandin G/H synthase 1 (Ptgs1), an oxygenase involved in inflammation mediated by chemokine and cytokine signaling pathways. The module was enriched for several GO terms that involved binding. Five gene products were associated with actin binding (GO: 0003779), representing a ninefold enrichment for the category (adjusted p value = 2.68 × 10−2). Ten gene products were associated with receptor binding (GO: 0005102), with sixfold enrichment for the category (adjusted p value = 1.67 × 10−3). Finally, four genes were associated with G protein-coupled receptor activity (GO: 000490), with 17-fold enrichment for the category (adjusted p value = 1.67 × 10−2). We identified 16 phenotypic QTLs that overlapped lightcyan’s meQTL; 2 of these phenotypic QTLs met our criteria, and both were related to bone mineral density (Bmd51 and Bmd52). The bone mineral density phenotypes Bmd51 and Bmd52 are associated with the variability in femur areal and volumetric bone mineral density, respectively (Koller et al. 2008). Firebrick4 There were 45 gene products in the firebrick4 module, and 17 of them were annotated in the Ensembl database. The hub was the coatomer protein complex subunit alpha gene product (Copa), which is a vesicle coat protein. Firebrick4 was enriched for several GO term transport categories. There were seven gene products associated with vesicle-mediated transport (GO: 0016192), representing an eightfold enrichment for the category (adjusted p value = 2.88 × 10−3), and seven were associated with protein transport (GO: 0015031), with sixfold enrichment for the category (adjusted p = 1.98 × 10−2). Five phenotypic QTLs overlapped the firebrick4 meQTL, but only one met our criteria: insulin/glucose ratio (Insglur5), which is associated with insulin resistance. In the same study that was referenced for the plasma insulin QTL that overlapped the plum2 module, Marissal-Atvy et al. (2014b) identified insulin resistance as a relevant phenotype for metabolism. Insulin resistance was calculated by finding the percent decrease in glycaemia after injecting animals with insulin. Agreement was found between the PANTHER enrichment and RGD QTL databases for the thistle2 and plum2 modules (Table 3). There were strong similarities between gene product enrichment and overlapping phenotypic QTLs for the thistle2 module. In both databases, there was strong evidence that this module is related to the immune system. The strongest overlapping phenotypic QTL was related to the IgG autoantibody and autoimmune disease, while the module was enriched for the immune system, and antigen presentation and processing. The similarities between enrichment and overlapping phenotypic QTLs are not as obvious for the plum2 module; however, several of the proteins that directly impact food consumption and insulin regulation are associated with pro-inflammatory responses (Wensveen et al. 2015), and the module was enriched for genes related to immune response. With these similarities in mind, we chose to further analyze our results with the thistle2 and plum2 gene product networks. Additional characterization of modules Partial correlation Twenty-five of the gene products within the thistle2 module were co-localized to chromosome 20, so we performed partial correlation analysis to determine if the correlation among gene products was due to independent causal loci in linkage disequilibrium. After adjusting for the meQTL, the module retained the majority of its nodes; in fact, none of gene products associated with chromosome 20 lost all of their connections (Fig. 2b). This indicates that these gene products from chromosome 20 likely have a biological relationship, and are not simply expressed in similar manner because of linkage disequilibrium among causal loci. Liver cell type-specific transcriptome analysis Of the 9331 genes that remained after pre-processing of the cell type-specific arrays, 8447 (90 %) were differentially expressed among cell types (FDR < 0.05). The normalized expression data and the original CEL files are available on PhenoGen. Because the thistle2 and plum2 modules are related to the immune system and immune response, we compared the Kupffer cell expression of genes within the modules to each of the other three cell types in this cell-specific analysis (hepatocytes, hepatic stellate cells, and sinusoidal endothelial cells). Thistle2 Of the 37 gene products in the thistle2 module, Kupffer cells had the highest expression among the 4 cell types in 27 (57 %) of the gene products (Fig. 4). The hub gene, RT1-CE16, was up-regulated in the Kupffer cells; these cells showed 43 % higher expression of the hub gene compared to hepatocytes, 23 % higher expression compared to hepatic stellate cells, and 8 % higher expression compared sinusoidal epithelial cells. The five most highly connected genes of the thistle2 module were up-regulated in Kupffer cells, and of the top 15 most highly connected genes in the thistle2 module, 13 were up-regulated in Kupffer cells.Fig. 4 Cell type-specific expression of gene products in the thistle2 module. Cell type: HC hepatocyte cells, HSC hepatic stellate cells, SEC sinusoidal endothelial cells, KC Kupffer cells. Strain: BN-Lx Brown Norway with polydactyly-luxate mutation, SHR spontaneous hypertensive rat. The color and intensity of each cell represent its relative expression compared to the strain average in the Kupffer cells, e.g., a value of 2 indicates that the expression in that strain and cell type is twice the average expression in the Kupffer cells. Genes are in order of connectivity after partial correlation, with the highest connectivity at the top of the graphic and the lowest connectivity at the bottom Plum2 Of the 38 gene products in the plum2 module, Kuppfer cells had the highest expression among the four cell types in 17 (45 %) of the gene products (Fig. 5). The hub gene, Ifr7, was up-regulated in Kupffer cells; these cells had 520 % higher expression of the hub gene compared to hepatocytes, 54 % higher expression compared to hepatic stellate cells, and 19 % higher expression compared to sinusoidal epithelial cells.Fig. 5 Cell type-specific expression of gene products in the plum2 module. Cell type: HC hepatocyte cells, HSC hepatic stellate cells, SEC sinusoidal endothelial cells, KC Kupffer cells. Strain: BN-Lx Brown Norway with polydactyl luxate mutation, SHR spontaneous hypertensive rat. The color and intensity of each cell represent its relative expression compared to the strain average in the Kupffer cells, e.g., a value of 2 indicates that the expression in that strain and cell type is twice the average expression in the Kupffer cells. Genes are in order of connectivity, with the highest connectivity at the top of the graphic and the lowest connectivity at the bottom Although not discussed here, the cell type-specific expression values for gene products in the firebrick4 and lightcyan modules are displayed in Supplementary Figs. 3 and 4. Discussion The approach discussed in this paper included increasing the interpretability of microarray data using RNA-Seq and DNA-Seq data, constructing gene product networks with WGCNA, and applying stringent criteria for the identification of genetically regulated modules within the rat liver. This resulted in the thistle2 module with gene products that were related to the immune system, including the hub gene product RT1-CE16 (an MHC-class I molecule). Thistle2 was also enriched for the immune system process, and for antigen processing and presentation. Not surprisingly, the majority of gene products in this module had higher expression in Kupffer cells than in hepatocytes, hepatic stellate cells, or sinusoidal endothelial cells (Figs. 4, 5). Further, after searching through publicly available phenotypic QTLs that overlapped the thistle2 meQTL, we found two phenotypic QTLs that mapped to the same region and were associated with immune system function: collagen-induced arthritis (Ciaa1) and adjuvant-induced arthritis (Aia1). Interestingly, Ciaa1 is located at the marker for lymphotoxin alpha, and the thistle2 module contains the gene product for lymphotoxin beta; lymphotoxin alpha and beta specifically bind to each other. The collagen-induced and adjuvant-induced arthritis rat models are commonly used to study rheumatoid arthritis (Joe 2006). Cell type-specific analyses showed that the majority of components within the thistle2 module primarily reside in Kupffer cells. These cells are the monocyte and macrophage cells of the liver (Li et al. 2010), are responsible for cytokine production and the presentation of antigen to B and T cells (Li et al. 2010), and are involved in systemic immune tolerance (You et al. 2008; Xu et al. 2014). It has also been shown that these monocyte and macrophage systems are an essential component of autoimmune disease (Li et al. 2010). With all of these in mind, it can be postulated that the liver contributes to rheumatoid arthritis through a genetically influenced and Kupffer cell-mediated effect on systemic immune tolerance that predisposes to, or diminishes, the likelihood of autoimmune disease. Behavioral phenotypic QTLs for food consumption and corticosterone levels also overlapped the thistle2 meQTL. Interestingly, factors influencing food consumption have been associated with measures of immune function. Leptin released from adipose tissue regulates the behavior of food consumption by suppressing appetite, and low levels of leptin have been shown to also suppress the immune system (Fernández-Riejos et al. 2008). It has also been observed that patients with rheumatoid arthritis have higher levels of leptin, and that leptin-deficient mice are less susceptible to antigen-induced arthritis (Lago et al. 2007). This implies that leptin, which plays an important role in food consumption, may also play a role in immune suppression and autoimmunity. Finally, food consumption and circulating corticosterone levels have been linked. In rats, it has been shown that the daily pattern of corticosterone can be modified by restrictive feeding (Namvar et al. 2016), and that high levels of glucocorticoids increase the intake of foods with high fat and/or sucrose (Dallman et al. 2005). We also identified the plum2 module, which was enriched for immune response. It contained gene products associated with the pro-inflammatory response: peg interferon alpha-2a/b, chemokines, interferon regulatory factors, and response to interferon. Additionally, many of the gene products in this module showed higher expression in Kupffer cells, when compared to the other three cell types. Phenotypic QTLs for food consumption and insulin regulation overlapped the genetic control of expression levels in plum2. Cortisol provides an important link between food consumption and insulin regulation; high cortisol levels can promote insulin resistance at the cellular level, which increases the hunger signals that lead to overeating (Aronson 2009). Leptin also helps to explain the relationship between food consumption and inflammation, because it functions in part as a pro-inflammatory cytokine and its levels are increased by inflammatory stimuli such as IL-1 and IL-6 (Lago et al. 2007). It has also been postulated that there is a relationship between insulin resistance, interferon-gamma, and inflammation (Wensveen et al. 2015). Wensveen et al. showed that obesity stimulates the up-regulation of interferon-gamma and NK cell proliferation, which in turn increases the production of pro-inflammatory macrophages and insulin sensitivity. In this analysis, we identified two genetically driven modules with obvious biological relevance; however, this does not imply that the remaining 48 modules do not represent biological pathways. Our goal was to demonstrate how these methods and models, in conjunction with currently available public databases, could be used to uncover biological relevance and, indeed, we identified two liver modules associated with immune response. There are several possible reasons why more modules were not linked to biological pathways. Nearly 5000 of the RNA-Seq-derived liver genes were not interrogated by the array, and some of the genes that were interrogated by the array failed to be detected above background; consequently, these technology-related limitations could miss important genes. Thus, modules that are derived in this manner may not include every gene product that is relevant to the related pathway; instead, they represent a set of the measureable genes related to the pathway. Publication biases, associated with the public databases used to define biological relevance, are another possible limitation for identifying more biologically relevant modules. For example, the QTL database may be biased to phenotypes that have been researched with respect to genetic determinants. Some phenotypes may have a higher abundance of QTLs, because rats have been good models for studying a particular disease process (e.g., alcohol response, blood pressure, arthritis); consequently, QTLs are both overrepresented and underrepresented in the current databases. Further, regardless of the rat population used for mapping, all of the RGD QTLs were considered. Although this “inclusive” approach may identify phenotypic QTL that do not represent the same causal loci as the overlapping meQTL, because they were detected in two different populations, the two loci are likely in linkage disequilibrium and may have similar effects. It is true that without detailed QTL studies using rat populations derived from the same or similar strains as the progenitors of the HXB/BXH RI panel, we may miss relevant phenotypic QTL. Some may argue that using a heterogeneous tissue sample will ‘mask’ important transcription patterns. Although some connections between genes will be missed, the connections that are identified are robust strong signals that are important in a systems genetics analysis. When we examined cell type-specific expression in modules that were associated with immunity and inflammation, as predicted, both modules contained strong signals from Kupffer cells. Kupffer cells account for a small proportion of the total cells in healthy liver, but we were still able to recover relevant pathways that they likely influence. Using whole tissue is not an end-all approach, but rather a starting point to identify key components of a system/network. In conclusion, our systems genetics approach improved the interpretability of microarray data by mapping probes to progenitor strain-specific transcriptomes, and allowed for the identification and characterization of liver co-expression modules within our panel of rats. This combination of methods and data allowed us to characterize previously unannotated gene products, and to identify genetically and biologically relevant gene networks. RNA-based networks can serve many purposes in future research: (1) they can be viewed as an intermediate product between DNA sequence variants and a particular phenotype, (2) they can be used to “hypothesize” about the function of individual unannotated transcripts through associations within a module, and (3) they can be used to identify optimal animal strains for future studies of disease etiology by prediction of individual strain predisposition to, magnitude of, and progression of disease. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary material 1 (DOCX 354 kb) This work was supported in part by NIAAA/NIH [R24AA013162 (BT); AAU01 Developmental Grant-INI Project (MPP); R24AA012885 (HT); P50AA011999 (HT)]; NHLBI/NIH [HL35018 (MPP)]; the Banbury Fund (BT). The authors are grateful for the expert technical assistance with microarrays and RNA-Seq provided by Yinni Yu and Adam Chapman, and for the expert technical assistance in maintaining the BN-Lx, SHR, and HXB/BXH rat colonies at UCSC provided by Laura Breen and Joseph Gatewood. ==== Refs References Aitman TJ Critser JK Cuppen E Dominiczak A Fernandez-Suarez XM Flint J Gauguier D Geurts AM Gould M Harris PC Progress and prospects in rat genetics: a community view Nat Genet 2008 40 516 522 10.1038/ng.147 18443588 Allocco DJ Kohane IS Butte AJ Quantifying the relationship between co-expression, co-regulation and gene function BMC Bioinformatics 2004 5 18 10.1186/1471-2105-5-18 15053845 Aronson D Cortisol—its role in stress, inflammation, and indications for diet therapy Todays Dietit. 2009 11 38 Bielavská E Kren V Musilová A Zídek V Pravenec M Genome scanning of the HXB/BXH sets of recombinant inbred strains of the rat for quantitative trait loci associated with conditioned taste aversion Behav Genet 2002 32 51 56 10.1023/A:1014407928865 11958542 Birney E Andrews TD Bevan P Caccamo M Chen Y Clarke L Coates G Cuff J Curwen V Cutts T An overview of Ensembl. Genome Res. 2004 14 925 928 10.1101/gr.1860604 15078858 Broman KW Wu H Sen Ś Churchill GA R/qtl: QTL mapping in experimental crosses Bioinformatics 2003 19 889 890 10.1093/bioinformatics/btg112 12724300 Churchill GA Doerge RW Empirical threshold values for quantitative trait mapping Genetics 1994 138 963 971 7851788 Conti LH Jirout M Breen L Vanella JJ Schork NJ Printz MP Identification of quantitative trait loci for anxiety and locomotion phenotypes in rat recombinant inbred strains Behav Genet 2004 34 93 103 10.1023/B:BEGE.0000009479.02183.1f 14739700 Dallman MF Pecoraro NC la Fleur SE Chronic stress and comfort foods: self-medication and abdominal obesity Brain Behav Immun 2005 19 275 280 10.1016/j.bbi.2004.11.004 15944067 Deleve LD Dacarbazine toxicity in murine liver cells: a model of hepatic endothelial injury and glutathione defense J Pharmacol Exp Ther 1994 268 1261 1270 8138939 Fernández-Riejos P Goberna R Sánchez-Margalet V Leptin promotes cell survival and activates Jurkat T lymphocytes by stimulation of mitogen-activated protein kinase Clin Exp Immunol 2008 151 3 505 518 10.1111/j.1365-2249.2007.03563.x 18234059 Festing MFW Evidence should trump intuition by preferring inbred strains to outbred stocks in preclinical research ILAR J 2014 55 399 404 10.1093/ilar/ilu036 25541542 Furuya T Salstrom JL McCall-Vining S Cannon GW Joe B Remmers EF Griffiths MM Wilder RL Genetic dissection of a rat model for rheumatoid arthritis: significant gender influences on autosomal modifier loci Hum Mol Genet 2000 9 2241 2250 10.1093/oxfordjournals.hmg.a018915 11001927 Gao B Jeong W-I Tian Z Liver: An organ with predominant innate immunity Hepatol Baltim Md 2008 47 729 736 10.1002/hep.22034 Hermsen R de Ligt J Spee W Blokzijl F Schäfer S Adami E Boymans S Flink S van Boxtel R van der Weide RH Genomic landscape of rat strain and substrain variation BMC Genom 2015 16 357 10.1186/s12864-015-1594-1 Higuchi H Gores GJ Mechanisms of liver injury: an overview Curr Mol Med 2003 3 483 490 10.2174/1566524033479528 14527080 Joe B Quest for arthritis-causative genetic factors in the rat Physiol Genomics 2006 27 1 11 10.1152/physiolgenomics.00034.2005 16804090 Johnson WE Li C Rabinovic A Adjusting batch effects in microarray expression data using empirical Bayes methods Biostatistics 2007 8 118 127 10.1093/biostatistics/kxj037 16632515 Kamimura S Tsukamoto H Cytokine gene expression by Kupffer cells in experimental alcoholic liver disease Hepatol Baltim Md 1995 22 1304 1309 Karolchik D Hinrichs AS Furey TS Roskin KM Sugnet CW Haussler D Kent WJ The UCSC table browser data retrieval tool Nucleic Acids Res 2004 32 D493 D496 10.1093/nar/gkh103 14681465 Kawahito Y Cannon GW Gulko PS Remmers EF Longman RE Reese VR Wang J Griffiths MM Wilder RL Localization of quantitative trait loci regulating adjuvant-induced arthritis in rats: evidence for genetic factors common to multiple autoimmune diseases J Immunol Baltim Md 1998 1950 161 4411 4419 Kim D Pertea G Trapnell C Pimentel H Kelley R Salzberg SL TopHat2: accurate alignment of transcriptomes in the presence of insertions, deletions and gene fusions Genome Biol 2013 14 R36 10.1186/gb-2013-14-4-r36 23618408 Koller DL Liu L Alam I Sun Q Econs MJ Foroud T Turner CH Linkage Screen for BMD Phenotypes in Male and Female COP and DA Rat Strains J Bone Miner Res 2008 23 1382 1388 10.1359/jbmr.080401 18707222 Kunes J Kren V Pravenec M Zicha J Use of recombinant inbred strains for evaluation of intermediate phenotypes in spontaneous hypertension Clin Exp Pharmacol Physiol 1994 21 903 906 10.1111/j.1440-1681.1994.tb02463.x 7882582 Lago F Dieguez C Gómez-Reino J Gualillo O Adipokines as emerging mediators of immune response and inflammation Nat Clin Pract Rheumatol 2007 3 716 724 10.1038/ncprheum0674 18037931 Langfelder, P. (2013). Signed vs. unsigned topological overlap matrix. Technical report Langfelder P Horvath S WGCNA: an R package for weighted correlation network analysis BMC Bioinformatics 2008 9 559 10.1186/1471-2105-9-559 19114008 Langfelder P Zhang B Horvath S Defining clusters from a hierarchical cluster tree: the Dynamic Tree Cut package for R Bioinformatics 2008 24 719 720 10.1093/bioinformatics/btm563 18024473 Langfelder P Luo R Oldham MC Horvath S Is my network module preserved and reproducible? PLoS Comput Biol 2011 7 e1001057 10.1371/journal.pcbi.1001057 21283776 Li Y Lee PY Reeves WH Monocyte and macrophage abnormalities in systemic lupus erythematosus Arch Immunol Ther Exp (Warsz) 2010 58 355 364 10.1007/s00005-010-0093-y 20676786 Malaney P Nicosia SV Davé V One mouse, one patient paradigm: new avatars of personalized cancer therapy Cancer Lett 2014 344 1 12 10.1016/j.canlet.2013.10.010 24157811 Marissal-Arvy N Diane A Moisan M-P Larue-Achagiotis C Tridon C Tome D Fromentin G Mormède P QTLs influencing carbohydrate and fat choice in a LOU/CxFischer 344 F2 rat population Obes. Silver Spring Md 2014 22 565 575 10.1002/oby.20485 Marissal-Arvy N Heliès J-M Tridon C Moisan M-P Mormède P Quantitative trait Loci influencing abdominal fat deposition and functional variability of the HPA axis in the rat Horm Metab Res Horm Stoffwechs Horm Métab 2014 46 635 643 10.1055/s-0034-1383574 Mi H Muruganujan A Casagrande JT Thomas PD Large-scale gene function analysis with the PANTHER classification system Nat Protoc 2013 8 1551 1566 10.1038/nprot.2013.092 23868073 Mi H Poudel S Muruganujan A Casagrande JT Thomas PD PANTHER version 10: expanded protein families and functions, and analysis tools Nucleic Acids Res 2016 44 D336 D342 10.1093/nar/gkv1194 26578592 Moldéus P Högberg J Orrenius S Isolation and use of liver cells Methods Enzymol 1978 52 60 71 10.1016/S0076-6879(78)52006-5 672656 Namvar S Gyte A Denn M Leighton B Piggins HD Dietary fat and corticosterone levels are contributing factors to meal anticipation Am J Physiol Regul Integr Comput Physiol 2016 310 R711 R723 10.1152/ajpregu.00308.2015 Parker, C.C., Chen, H., Flagel, S.B., Geurts, A.M., Richards, J.B., Robinson, T.E., Solberg Woods, L.C., and Palmer, A.A. (2014). Rats are the smart choice: Rationale for a renewed focus on rats in behavioral genetics. Neuropharmacology 76 Pt B, 250–258 Pravenec M Zídek V Landa V Simáková M Mlejnek P Kazdová L Bílá V Krenová D Kren V Genetic analysis of “metabolic syndrome” in the spontaneously hypertensive rat Physiol Res Acad Sci Bohemoslov 2004 53 Suppl 1 S15 S22 Printz MP Jirout M Jaworski R Alemayehu A Kren V Invited Review: HXB/BXH rat recombinant inbred strain platform: a newly enhanced tool for cardiovascular, behavioral, and developmental genetics and genomics J Appl Physiol 2003 94 2510 2522 10.1152/japplphysiol.00064.2003 12736193 Racanelli V Rehermann B The liver as an immunological organ Hepatol. Baltim. Md 2006 43 S54 S62 10.1002/hep.21060 Ravasz E Somera AL Mongru DA Oltvai ZN Barabási A-L Hierarchical organization of modularity in metabolic networks Science 2002 297 1551 1555 10.1126/science.1073374 12202830 Riley WT Nilsen WJ Manolio TA Masys DR Lauer M News from the NIH: potential contributions of the behavioral and social sciences to the precision medicine initiative Transl Behav Med 2015 5 243 246 10.1007/s13142-015-0320-5 26327928 Ritchie MD The success of pharmacogenomics in moving genetic association studies from bench to bedside: study design and implementation of precision medicine in the post-GWAS era Hum Genet 2012 131 1615 1626 10.1007/s00439-012-1221-z 22923055 Roberts A Pimentel H Trapnell C Pachter L Identification of novel transcripts in annotated genomes using RNA-Seq Bioinformatics 2011 27 2325 2329 10.1093/bioinformatics/btr355 21697122 Saba LM Flink SC Vanderlinden LA Israel Y Tampier L Colombo G Kiianmaa K Bell RL Printz MP Flodman P The sequenced rat brain transcriptome-its use in identifying networks predisposing alcohol consumption FEBS J 2015 282 3556 3578 10.1111/febs.13358 26183165 Shimoyama M De Pons J Hayman GT Laulederkind SJF Liu W Nigam R Petri V Smith JR Tutaj M Wang S-J The Rat Genome Database 2015: genomic, phenotypic and environmental variations and disease Nucleic Acids Res 2015 43 D743 D750 10.1093/nar/gku1026 25355511 Smit A, Hubley R, Green P (1996) RepeatMasker Open 3.0 Trapnell C Williams BA Pertea G Mortazavi A Kwan G van Baren MJ Salzberg SL Wold BJ Pachter L Transcript assembly and quantification by RNA-Seq reveals unannotated transcripts and isoform switching during cell differentiation Nat Biotechnol 2010 28 511 515 10.1038/nbt.1621 20436464 Trapnell C Roberts A Goff L Pertea G Kim D Kelley DR Pimentel H Salzberg SL Rinn JL Pachter L Differential gene and transcript expression analysis of RNA-seq experiments with TopHat and Cufflinks Nat Protoc 2012 7 562 578 10.1038/nprot.2012.016 22383036 Tsukamoto H Horne W Kamimura S Niemelä O Parkkila S Ylä-Herttuala S Brittenham GM Experimental liver cirrhosis induced by alcohol and iron J Clin Invest 1995 96 620 630 10.1172/JCI118077 7615836 Vanderlinden LA Saba LM Printz MP Flodman P Koob G Richardson HN Hoffman PL Tabakoff B Is the alcohol deprivation effect genetically mediated? Studies with HXB/BXH recombinant inbred rat strains Alcohol Clin Exp Res 2014 38 2148 2157 10.1111/acer.12471 24961585 Weiss JN Karma A MacLellan WR Deng M Rau C Rees CM Wang J Wisniewski N Eskin E Horvath S “Good Enough Solutions” and the Genetics of Complex Diseases Circ Res 2012 111 493 504 10.1161/CIRCRESAHA.112.269084 22859671 Wensveen FM Jelenčić V Valentić S Šestan M Wensveen TT Theurich S Glasner A Mendrila D Štimac D Wunderlich FT NK cells link obesity-induced adipose stress to inflammation and insulin resistance Nat Immunol 2015 16 376 385 10.1038/ni.3120 25729921 Xu L Yin W Sun R Wei H Tian Z Kupffer cell-derived IL-10 plays a key role in maintaining humoral immune tolerance in hepatitis B virus-persistent mice Hepatol Baltim Md 2014 59 443 452 10.1002/hep.26668 You Q Cheng L Kedl RM Ju C Mechanism of T cell tolerance induction by murine hepatic Kupffer cells Hepatol Baltim Md 2008 48 978 990 10.1002/hep.22395 Zhang B, Horvath S (2005) A general framework for weighted gene co-expression network analysis. Stat Appl Genet Mol Biol 4

PMC005xxxxxx/PMC5002043.txt

==== Front Cell Mol Life SciCell. Mol. Life SciCellular and Molecular Life Sciences1420-682X1420-9071Springer International Publishing Cham 226510.1007/s00018-016-2265-3ReviewCommon pitfalls of stem cell differentiation: a guide to improving protocols for neurodegenerative disease models and research Engel Martin Do-Ha Dzung Muñoz Sonia Sanz Ooi Lezanne +61 2 4221 5865lezanne@uow.edu.au Illawarra Health and Medical Research Institute, School of Biological Sciences, Faculty of Science, Medicine and Health, University of Wollongong, Wollongong, NSW Australia 6 5 2016 6 5 2016 2016 73 19 3693 3709 21 1 2016 5 4 2016 3 5 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Induced pluripotent stem cells and embryonic stem cells have revolutionized cellular neuroscience, providing the opportunity to model neurological diseases and test potential therapeutics in a pre-clinical setting. The power of these models has been widely discussed, but the potential pitfalls of stem cell differentiation in this research are less well described. We have analyzed the literature that describes differentiation of human pluripotent stem cells into three neural cell types that are commonly used to study diseases, including forebrain cholinergic neurons for Alzheimer’s disease, midbrain dopaminergic neurons for Parkinson’s disease and cortical astrocytes for neurodegenerative and psychiatric disorders. Published protocols for differentiation vary widely in the reported efficiency of target cell generation. Additionally, characterization of the cells by expression profile and functionality differs between studies and is often insufficient, leading to highly variable protocol outcomes. We have synthesized this information into a simple methodology that can be followed when performing or assessing differentiation techniques. Finally we propose three considerations for future research, including the use of physiological O2 conditions, three-dimensional co-culture systems and microfluidics to control feeding cycles and growth factor gradients. Following these guidelines will help researchers to ensure that robust and meaningful data is generated, enabling the full potential of stem cell differentiation for disease modeling and regenerative medicine. Keywords iPS cellsDifferentiationNeurodegenerationAlzheimer’s diseaseDopaminergic neuronsCholinergic neuronsAstrocyteshttp://dx.doi.org/10.13039/501100000925National Health and Medical Research CouncilAPP1071250APP1079995Ooi Lezanne issue-copyright-statement© Springer International Publishing 2016 ==== Body Introduction Studying mechanisms and potential therapeutic targets of neurodegenerative diseases has historically been limited by the timely access to the affected tissue. Furthermore, the use of animal models for disorders without a clear genetic cause has shown to be of limited translational value to the clinical setting. The possibility of creating neuronal cultures from human stem cells, particularly from induced pluripotent stem cells (iPSC) of diagnosed individuals, has received wide attention for the potential of creating translatable disease-in-a-dish models [1]. Following the discovery of iPSCs, several high profile publications have fuelled the enthusiasm for their use in research into Parkinson’s disease [2], Alzheimer’s disease [3, 4], motor neurone disease [5] and mental disorders [6]. While optimized methods for reprogramming iPSCs have broadly been adopted (recently reviewed by Revilla et al. [7]), the protocols for differentiating iPSCs into neuronal cultures vary significantly for the same desired cell types. Strategies for converting stem cells into terminally differentiated cells predominantly follow observations from developmental mouse and rat studies, intending to model the in vivo progression of chemical signaling. However, variations in composition, concentration and timing of the signaling molecules can lead to marked differences in the resulting cultures and maturation stage of the desired cells. Furthermore, recent studies report variations in the differentiation efficiency between different stem cell lines, which is particularly relevant if the differences are diagnosis-specific [8, 9]. Finally, novel culturing methods, including three-dimensional cultures and hypoxic conditions, have been reported to influence the differentiation efficiency [10, 11]. To ensure the generation of meaningful results, stem cell derived cultures will therefore need to undergo thorough characterization to identify the composition and functionality of the differentiated cells. The development and maturation of neuronal cells depends on different types of chemical signaling in vivo. Initially, growth factors and chemoattractants released into the extracellular space by progenitor cells trigger location specific differentiation [12]. Upon neurite formation, local neurotransmitter release guides dendrite development and cell maturation to form the dynamic networks of the central nervous system [13, 14]. While growth factor signals can be recreated in vitro through time-dependent media supplementation, location and cell type-specific synaptic inputs cannot easily be modeled. It is therefore highly relevant to consider the developmental and maturation environment of the desired cell type when deciding to study disease processes in iPSC derived cultures. With the present review, we provide an overview of existing differentiation strategies for neurodegenerative disease-relevant cell types of forebrain cholinergic neurons, midbrain dopaminergic neurons and cortical astrocytes. Advances in motor neuron differentiation have been extensively reviewed, including recently [15]. We address the variation in protocol efficiencies by providing a checklist that can be used to evaluate the quality and reproducibility of in vitro differentiation. Finally, we discuss recent culturing method developments aiming to improve the quality of stem cell derived neural cultures. Human stem cell derived in vitro disease models have the potential to overcome the limitations of existing cell line work and can become a vital research stream next to animal modeling strategies. However, to progress in disease understanding and treatment development, the quality and biological relevance of stem cell-derived cultures need to be ensured. From in vivo development to in vitro differentiation Forebrain cholinergic neurons Cholinergic neurons in the mammalian brain, including basal forebrain cholinergic neurons (BFCNs) and interneurons of the striatum, are defined by the production of acetylcholine (ACh) and its use as a neurotransmitter. BFCNs play an important role in cognitive functions, such as learning, memory and attention, and are implicated in the rapid eye movement sleep phase. The loss of BFCNs in neurodegenerative disorders, including Alzheimer’s disease, thus leads to severe cognitive impairments and memory deficits [16, 17]. Mammalian development of cholinergic neurons During mammalian development, cholinergic neurons are derived from neural precursors of the ectoderm layer. After the formation of the neural tube, a selection of cells start to respond to high concentrations of sonic hedgehog (SHH), a morphogen that induces ventralization (differentiation towards the anterior) of the neural tube, and low concentrations of Wnt, a morphogen that induces caudalization (differentiation towards the posterior) [18]. The morphogens retinoic acid (RA), bone morphogenetic proteins (BMPs) and fibroblast growth factors (FGFs) are also known to play a key role in telencephalon development, the most anterior region of the developing brain [19]. In rodents and primates, BFCNs are formed in the medial ganglionic eminence (MGE), a ventral region of the telencephalon, with projections to the hippocampus and frontal cerebral cortex [20], areas implicated in the cognitive and psychological deficits of neurodegenerative diseases [21, 22]. The anterior/posterior patterning of the brain begins at embryonic day 8.5 (E8.5) in mice, when the telencephalic neuroepithelia expresses the transcription factors forkhead box G1 (FOXG1) and paired box 6 (PAX6) [23–25]. SHH together with FOXG1 induce the expression of FGFs, which activate downstream transcription factors characteristic of the ventral patterning. The expression of NK2 homeobox 1 (NKX2.1) at E9.5 defines the MGE region and by day E12.5 the expression of LIM homeobox 8 (LHX8) and insulin gene enhancer protein 1 (ISL1) determine a cholinergic fate. Choline acetyl transferase (ChAT) expression in neural precursor cells arises at postnatal day 7 (P7), with ChAT-positive cells increasing in number after P8 and remaining stable during adulthood [26]. While the full range of molecular drivers for BFCN development remains to be mapped, several essential factors for the differentiation, growth and survival of BFCNs have been identified. The cholinergic phenotype of embryonic BFCNs is induced and maintained by BMP9 [27], whilst FGF2 and brain-derived neurotrophic factor (BDNF) stimulate cholinergic differentiation [28, 29]. The maturation, neurotransmitter phenotype, arborisation and survival processes are finally controlled through nerve growth factor (NGF) signaling [30]. These growth factors drive the gene expression profile of BFCNs through specific transcription factors, such as LHX8 [31] and GBX1, a gastrulation box homeodomain protein, which can be used in the differentiation of BFCNs in vitro [32]. Characteristics of mature cholinergic neurons Basal forebrain cholinergic neurons are characterized by their neuroanatomical location and the expression of ACh related genes. The key markers used to identify mature cholinergic neurons are involved in ACh synthesis (choline acetyltransferase, ChAT), transport (e.g., the vesicular ACh transporter, vAChT and the choline transporter, Cht1) and hydrolysis (e.g., acetylcholinesterase). Since NGF levels are directly related to the health and function of BFCNs [30], the high affinity neurotrophin receptor, TrkA, and the low affinity receptor, p75NTR, are also useful identifiers. Due to their extraordinary complexity, the characterization of the complete morphology of cholinergic neurons has been difficult. In vitro studies from primary cultures showed that BFCNs have large cell bodies with two to four primary neurites [33]. However, a recent study in mice has demonstrated that individual cholinergic neurons from the basal forebrain region have axons that develop up to 50 cm in length, with approximately 1000 branches [34]. The electrophysiological profile of BFCNs has been characterized in rats, where BFCNs show regular spontaneous discharge patterns with mean spontaneous activity of 20 impulses/s [35]. Further key characteristics of the BFCNs are a slow spiking activity (4–10 Hz) and slow after potentials (400–700 ms) when compared with non-cholinergic neurons (3–60 ms) [36], which have also been shown in iPSC-derived BFCNs [37]. The addition of NGF can increase slow depolarization and enhance synaptic activity by upregulating ChAT activity [38]. Current differentiation strategies for cholinergic neurons Cholinergic neurons derived from human embryonic stem cells (ESCs) and iPSCs can become an important tool for modeling neurodegenerative diseases, such as Alzheimer’s disease. However, due to their complexity, few studies have successfully differentiated BFCNs. Most of these studies start the differentiation process with the generation of embryoid bodies, followed by the formation of neural rosettes and neurospheres. These neurospheres contain neural precursor cells (NPCs) from which the different protocols will generate the BFCNs (Fig. 1).Fig. 1 Schematic diagram of basal forebrain cholinergic neuron differentiation. The differentiation of basal forebrain cholinergic neurons from pluripotent stem cell colonies is driven by transitions between two- and three-dimensional culturing stages, as well as timed exposure to essential growth factors, such as NGF. The presence or absence of developmental and maturation markers are essential guides to monitor the differentiation progress at each culturing stage towards mature, functional forebrain cholinergic neurons Nilbratt et al. assessed growth factors that could induce forebrain identity [39]. BDNF, NGF, ciliary neurotrophic factor (CNTF) and neurotrophin 3 (NT-3) were tested as candidates, but only BDNF and NGF reliably induced the expression of both NKX2.1 and LHX8 [39]. Following this, Bissonnette et al. [32] published a comparison of two different protocols to generate functional BFCNs from hESC. In both cases cells were pre-treated with RA, SHH and FGF8 [32]. One option relied on diffusible ligands for differentiation (Table 1), while the other additionally transfected an expression plasmid encoding the transcription factors LHX8 and GBX1 and a fluorescent tag. Transfected cells could be purified by fluorescence activated cell sorting, a step that increased the ratio of BFCNs in the final culture to 94 %. This method has also allowed the successful differentiation of BFCNs derived from iPSCs, as described by the same group, showing that the iPSC-derived BFCNs can be used as a model for Alzheimer’s disease, producing disease-related pathological features [40].Table 1 Comparison of basal forebrain cholinergic neuron differentiation protocols Differentiation protocol Nilbratt et al. [39] Bissonnette et al. ([32], BMP9 treatment) Bissonnette et al. ([32], nucleofection), Duan et al. [40] Liu et al. [41] Crompton et al. [37] Duration (days) ND 34 34 45 90 Efficiency (ChAT+ cells) 69–78 % 85 % 65 %; 94 % after FACS purification 38 % >90 % Growth factors BDNF, CNTF, EGF, FGF2, NGF, NT-3 BMP9, EGF, FGF2, FGF8, NGF, RA, SHH EGF, FGF2, FGF8, NGF, RA, SHH BDNF, BMP9, cAMP, IGF-1, NGF, SHH EGF, FGF2, SMI Developmental markers (protein, mRNA) BF-1, DLX1, DLX2, GBX2, GSH2, ISL1, LHX8, MASH1, NKX2.1 FORSE1 FORSE1, FOXG1, MASH1, NKX2.1 FOXG1, ISL1, MASH1, NKX2.1, OLIG2 FOXG1, ISL1, LHX8, NKX2.1 Maturity markers (protein, mRNA) ChAT, nAChRs, NMDAR, mAChRs, MAP2, p75NTR, TrkA, β-III-tubulin AChE, Calbindin, ChAT, MAP2, p75NTR, TrkA, vAChT ChAT, MAP2, p75NTR, vAChT ChAT, p75NTR, SYN-1, β-III-tubulin, vAChT ChAT, MAP2, p75NTR, SYN-1, β-III-tubulin, vAChT Physiological function Ca2+ response to ACh ACh production and release ACh production and release Functional voltage-gated channels Spontaneous action potentials ACh production and release Functional voltage-gated channels and cholinergic receptors Spontaneous action potentials Using a different strategy, Crompton et al. published a protocol for non-adherent differentiation of iPSCs into BFCNs [37]. In this procedure, neurospheres were treated with Nodal/transforming growth factor beta (TGF-β) inhibitor (small molecule inhibitor, SMI) to induce the endogenous expression of SHH, instead of its direct addition, resulting in a 90 % efficiency of β-III-tubulin/ChAT-expressing cells after 90 days [37]. Overall, only two of the mentioned protocols successfully reached >90 % ChAT-expressing cells. The main differences between the protocols are in their way of culturing (i.e., adherent by Bissonnette et al. [32] versus non-adherent by Crompton et al. [37]). This highlights the need for independent replication of both protocols to provide evidence for the use of either strategy. One potential advantage of the protocol developed by Bissonnette et al. involves using plasmid transfection via electroporation to trigger BFCN differentiation [32]. While this step allows fluorescently tagged cell sorting for purified cultures, transfection efficiency likely differs between each stem cell line and thus requires thorough optimization and counting of viable cells after sorting to produce replicable cultures. In summary, the majority of published protocols for cholinergic differentiation are based on the initial addition of SHH or its endogenous induction to induce ventral forebrain fate and the expression of developmental markers of the MGE. While treatment with NGF has been shown to be highly important for the generation of mature ChAT-expressing BFCNs (Fig. 1; Table 1), the incomplete functional characterization of mature BFCNs limits us from recommending a particular protocol. This shortcoming can be addressed by transplanting BFCN precursors into rodents to compare the in vitro maturation with in vivo maturation of cells from the same origin. While three of the listed protocols show that engrafted BFCN precursors develop into integrated BFCNs [32, 37, 41], none of the studies compared the in vitro differentiated cells with their in vivo counterparts. We can, thus, not yet recommend a reliable BFCN differentiation protocol. Midbrain dopaminergic neurons Midbrain dopaminergic (mDA) neurons are predominantly expressed in the substantia nigra pars compacta (SNc) and the ventral tegmental area (VTA) in rodents and primates [42–44]. SNc mDA neurons are required for initiation and control of motor functions, while VTA mDA neurons are important for reward behavior and cognition. Both nuclei are implicated in severe disorders, with degeneration of SNc mDA neurons being a hallmark of Parkinson’s disease, and impaired signaling of VTA mDA neurons being implicated in psychiatric disorders, such as schizophrenia and bipolar disorder. There is thus strong interest in differentiating human mDA neurons in vitro to study mechanisms contributing to the onset and progression of these disorders. Mammalian development of mDA neurons Midbrain DA neurons arise from NPCs of the ventral mesencephalon in mammals. The expression of aldehyde dehydrogenase 1 by progenitor cells at embryonic day 9.5 (E9.5) in mice triggers the development of post-mitotic cells, which produce the dopamine synthesizing enzyme tyrosine hydroxylase (TH) [45]. These mDA neuron precursors express nuclear receptor related 1 protein (NURR1) at E10.5 and differentiate into dopamine producing, TH-expressing neurons at E11.5 in the mediobasal floor plate [45, 46]. Neurogenesis of mDA neurons peaks at E12.5 in mice and E80 in non-human primates [42, 47–49], with mDA neuronal development occurring earlier in the SNc than VTA [47]. While the timing of prenatal mDA neurogenesis is similar in mice and rats, neurite development and migration is completed in postnatal week one in mice and three in rats [46, 50]. This suggests a likely further extension to the developmental period in primates that has yet to be mapped in humans. Neurogenesis of mDA neurons is driven by intrinsic and extrinsic signaling factors with precise temporal release patterns. The specific events have recently been reviewed in detail [51, 52]. Briefly, mDA progenitor cells receive target-derived neurotrophic factors from the mesencephalon floor plate of the dorsoventral neural tube axis (SHH and FGF8) [53–55] and later from the striatum (glial cell-derived neurotrophic factor, GDNF and neurturin, NRTN) [56–58]. Additional factors, such as BDNF, TGF3β, RA and ligands for members of the frizzled family of seven transmembrane receptors (Wnt1 and Wnt5a), promote mDA neuronal development and midbrain organization [59–63]. The variations in developmental timing of mDA neurons between species suggest an importance for the scheduling of the different factors between stem cells of different organisms. Characteristics of mature mDA neurons Midbrain DA neurons develop into a distinct cell type throughout their differentiation and maturation process which can be identified through functional, morphological and protein expression characteristics. The functional profile of mDA neurons requires the production of several key proteins, which have therefore been used as markers to identify mDA neurons within brain regions and in mixed primary cultures. The dopaminergic profile ultimately relies on the synthesis of dopamine from l-3,4-dihydroxyphenylalanine (l-DOPA) by TH, the re-uptake of dopamine from the synaptic cleft by dopamine transporters (DAT1), as well as the auto feedback loop, regulating dopamine production through activation of presynaptic dopamine receptor 2 (DR2). Promoter-driven expression or antibody labeling of TH has consequently been used to identify DA neurons in post mortem tissue [64], primary cell cultures [65] and human stem cell derived differentiated neurons [66]. However, TH is required for the first synthesis step of catecholamines and therefore also present in noradrenergic cells [67]. In addition, TH expression is tightly regulated by neuronal activity via modulation of histone acetylation levels and is neither specific to DA neurons nor expressed at consistent levels in these cells [68]. Selection based on DAT expression has shown high specificity for mature mDA neurons from the VTA, with lower expression levels in the SNc [69]. Consequently, purifying embryonic mouse brain cultures for DAT resulted in a higher proportion of mature mDA neurons than selecting for TH [70, 71]. To address limitations of TH and DAT expression based selection, a recent study explored cell surface proteins with high specificity for mDA neurons. Ganat et al. identified the nicotinic ACh receptor subunit 3 and 6 to be highly colocalized with mDA neurons in the mouse [72], suggesting that these receptor subunits could be additional selection options for stem cell-derived mDA neurons [73], following confirmation in human samples. The morphological profile of mDA neurons in vivo is dominated by their long projections, connecting the VTA with the neocortex via the mesocortical pathway and the SNc with the striatum via the nigrostriatal pathway. Limitations of the in vitro environment restrict this structural characteristic, moving the focus onto cell body size and shape. Midbrain DA neurons in the VTA are reportedly smaller and rounder than their elongated SNc relatives (13 vs. 19 µm Ø) [74]. Additionally, VTA mDA neurons are predominantly multipolar with radial projections while SNc mDA neurons have largely lateral and ventral projecting dendrites [75]. Although the reported characteristics are based on rodent in vitro and in vivo studies, they are expected to be observed in stem cell derived mDA neurons [76]. The electrophysiological profile of mDA neurons has been characterized in vivo and in vitro, identifying specific action potential patterns. Midbrain DA neurons can express either high bursts (>15 Hz) or slow background (1–5 Hz) action potential discharges [64], with each discharge starting with a prominent hyperpolarizing pulse [77] and depend on the G-protein-regulated inward-rectifier potassium channel 2 (GIRK2). The burst pattern has been shown to depend on the morphology of the mDA neurons [78] and requires specific stimulation in vitro [79]. This suggests that the functional assessment of stem cell-derived mDA neurons by electrophysiological characterization is met with several challenges, as the full phenotype depends on the input of the surrounding network and extracellular concentration of dopamine [77]. However, as similar action potential patterns have been identified in iPSC-derived mDA neurons [73], the electrophysiological profile forms the third critical characteristic of in vitro differentiated mDA neurons after protein expression and dopamine handling. Current differentiation strategies for mDA neurons The strong disease relevance of mDA neurons and their distinct developmental pathway has driven the differentiation of mDA neurons in vitro. Following the availability of mouse NPCs, mouse and human ESCs and more recently iPSCs, different strategies have been developed to create in vitro cultures rich in mature mDA neurons (Table 2). The majority of differentiation protocols closely follow the in vivo developmental stages, progressing from stem cells to neural rosettes (sometimes via embryoid body formation) to create neural progenitor cells and begin the mDA neuron patterning (Fig. 2). Cells at each stage are supplemented with a cocktail of growth factors reportedly found during the relevant developmental stages, such as SHH, FGF8 and BDNF (Table 2). Despite extensive testing of variations in media compositions, including undefined commercial options, and differentiation timings, the reported efficiency ratios have remained low with 8–40 % of generated cells expressing TH (Table 2). Higher yields are promised by the use of SMI, specifically targeting Wnt and GSK3β signaling [80] and factors involved in floor plate development [81]. However, since GSK3β signaling is tightly regulated during neurodevelopment [82], the risk of unintended consequences on mature mDA neurons remains to be assessed. To ensure the formation of physiologically relevant cells, the majority of protocols already go beyond TH expression analysis to identify mature mDA neurons. Critically, the thoroughness of each protocol in identifying the derived cultures still varies widely (Table 2), with two protocols presenting TH expressing cells without stating their proportion within the differentiated cultures [80, 83]. By not reporting electrophysiological characteristics, for example, it remains unclear whether the created dopamine-producing cells are close representations of SNc, VTA or other dopamine neurons. The protocol by Kriks et al. [81] uses fully adherent differentiation to form a promising foundation for further development and optimization of mDA neuron differentiation, with the strong need for identifying the full spectrum of the generated cultures.Table 2 Comparison of a representative selection of midbrain dopaminergic neuron differentiation protocols Differentiation protocol Brennand et al. [6] Chambers et al. [80] Fathi et al. [85] Hartfield et al. [73] Kriks et al. [81] Perrier et al. [86] Petit et al. [83] Reinhardt et al. [84] Tan et al. [87] Duration (days) 110 19 30 66 50 62 90 28 28 Efficiency (TH+ cells) 8 % ND (80 % PAX6+) 35 % 20.31 % 78 % 40 % ND (75 % β-III-tubulin+) 35 % 33 % Growth factors AA, BDNF, dcAMP, FGF2, GDNF AA, BDNF, cAMP, FGF2, FGF8, GDNF, Noggin, TGF3β AA, BDFN, dcAMP, FGF2, GDNF, LMX1A, SHH, SMI AA, BDNF, dcAMP, FGF8, GDNF, Noggin, SHH, SMI AA, BDNF, CHIR, DATP, dbcAMP, FGF8, GDNF, Purmorphamine, SHH, SMI, TGF3β AA, BDFN, dcAMP, FGF2, FGF8, GDNF, SHH, SMI AA, BDNF, cAMP, FGF8, GDNF, Noggin, SHH, SMI, TGF3β AA, BDNF, CHIR, dcAMP, FGF8, GDNF, PMA, SMI, TGF3β FGF2, FGF8, SHH Developmental markers (protein, mRNA) Nestin CDX2, FGF5, Nestin,OTX2, PLZF, PAX6, SOX1, SOX17 EN1, FOXA2, LMX1A, LMX1B, Nestin, PITX3, TAT, WNT1 FOXA2, Nestin, NURR1, PAX6 FOXA1, FOXA2, HESS, LHX2, LMX1A, NURR1, OTX2, PAX6 CRIPTO, NCAM, Nestin, PAX6, SOX1 DNMT3B, FOXA2, Nestin, PAX6 Brachyury, FOXA2, Nestin, PAX6, PAX7, SOX1, SOX9, TFAP2A NURR1 Maturity markers (protein, mRNA) GEPH, GluR1, MAP2, PSD95, SYN-1, TH, β-III-tubulin, VGAT, VGLUT1 TH, β-III-tubulin MAP2, PITX3, TH AADC, Calbindin, DAT1, GBA, GIRK2, IP3R, LRRK2, PITX3, α-synuclein, tau, TH DAT, GIRK2, TH, β-III-tubulin AADC, EN1, GFAP, MAP2, O4, PAX2, PAX5, SV2, SYN-1, TH, VMAT2 TH, β-III-tubulin MAP2, PAX3, PAX7, TH, β-III-tubulin ALDH1A1, GFAP, MAP2, PITX3, SCN1A, TH, β-III-tubulin, VMAT2 Physiological function DA production, spontaneous Ca2+ oscillations, spontaneous EPSPs and IPSPs ND ND DA production and uptake, spontaneous Ca2+ oscillations, mDA-specific mitochondrial responses, slow action potential trains DA production, slow action potential trains and spontaneous spikes DA production, repetitive action potential trains DA production Spontaneous action potentials ND Fig. 2 Schematic diagram of midbrain dopaminergic neuron differentiation. The differentiation of midbrain dopaminergic neurons from pluripotent stem cell colonies is driven by the transition from spherical to adherent cultures and a staggered supplementation of growth factors, particularly SHH, FGF8 and later TGF3β. The presence or absence of developmental and maturation markers are essential guides to monitor the differentiation progress at each culturing stage towards mature, functional midbrain dopaminergic neurons An important shortcoming across all mDA neuron differentiation protocols still remains: the direct comparison of in vitro differentiated neurons with engrafted and in vivo differentiated neurons. Two of the discussed protocols engrafted differentiated NPCs [84] or mDA neurons [81] into rodents, showing that these cells integrate and mature in their host environment. Neither of the protocols, however, compared the matured in vitro cultured cells with their in vivo relatives. The extent to which stem cell derived mDA neurons differ to their in vivo counterparts thus remains unclear. This information is vital to assess whether stem cell derived mDA neurons are appropriate for the range of questions they have been promised to answer. Cortical astrocytes Once defined as the “nerve glue” (neuroglia) for neurons, astrocytes have emerged as one of the key players in maintaining cellular homeostasis in the brain and spinal cord. Astrocytes regulate the ion concentration and remove excess neurotransmitter and cellular debris from the extracellular fluid surrounding neurons. In neurodegenerative diseases, however, this vast support network for neurons is often altered or dysregulated [88]. The dysregulation of astrocytes can lead to excessive neuroinflammation, a common pathology in neurodegenerative diseases, and contributes to neuronal deterioration [89]. It is, therefore, important to study the contribution of astrocytes to neuronal degeneration and disease. Mammalian development of astrocytes Astrocyte generation and maturation in vivo relies on precise temporal and positional stimuli provided by the cellular environment and neighboring cells [90]. In mice, the differentiation of astrocytes and other neural cells begins at E8.5 with neurulation of the neuroectoderm [91]. Neuroepithelial cells from the neuroectoderm firstly differentiate into radial glia, the precursors of astrocytes and NPCs [92]. Although radial glia and astrocytes are overlapping in their marker expression of proteins, such as vimentin and glial fibrillary acidic protein (GFAP) [93], the morphology, function and time of development differ significantly. In the mammalian brain, the main morphological characteristic of radial glia is their long processes, which extend from the ventricular zone all the way to the marginal zone of the pial surface [94]. Through asymmetric horizontal division, radial glia give rise to NPCs and initiate neurogenesis. The newly formed NPCs then migrate along the long processes of radial glia to form the different layers of the cortex [95]. In the late prenatal stage and early postnatal stage (E18-P7) in mice, after neurogenesis and neuronal migration are completed, radial glia differentiate into astrocytes, a process called astrogliogenesis. The differentiation of radial glia and NPCs into astrocytes is dependent on the activation of several signaling pathways. Two of these pathways are the JAK/STAT and the BMP-SMAD pathway [96]. These pathways are extrinsically activated by growth factors, including CNTF, cardiotrophin 1 (CT-1) and leukemia inhibitory factor (LIF), which are released by early neurons and late NPCs in vivo. Activation of these mechanisms leads to downstream events including chromatin modification and induction of astrocyte specific gene expression [97]. Transcription factors, such as signal transducer and activator of transcription 3 (STAT3), mothers against decapentaplegic homolog 1 (SMAD1), SMAD4 and nuclear factor 1A (NF1A) are activated and initiate the expression of GFAP, S100 calcium binding protein B (S-100β) and glutamate aspartate transporter (GLAST), all of which are currently used as astrocyte specific markers [98–100]. Within the first week after birth, astrocytes develop branched processes and attain their star like structure in the postnatal brain and spinal cord in mice [101]. Moreover, expression of the mature astrocyte marker glutamate transporter 1 (GLT1), a membrane protein important for the protection of neurons from glutamate-mediated excitotoxicity, was observed 2–3 weeks after birth [102]. Mouse studies have further shown that developmental GLT1 expression is upregulated in vitro when astrocytes are co-cultured with neurons [103]. This indicates that neuronal signaling is important for astrocyte maturation. To successfully differentiate stem cells into functionally and physiologically relevant astrocytes, the molecular pathways involved in the cell fate determination of astrocytes in vivo need to be replicated in vitro. However, as most of the developmental studies are based on mouse models, the translation of this information to human stem cell differentiation requires careful consideration. Characteristics of mature astrocytes In the pursuit of generating functional astrocytes, it is essential to assess the differentiated cells for the presence of astrocyte-specific characteristics that are generally observed in vivo. Morphologically, early astrocytes have large cell bodies with few processes. However, as astrocytes mature, more processes develop, elongate and branch out, giving the typical ‘star’ shape. In mice, this branching and elongation of processes takes place during the late postnatal stages (P14–P27) [104], suggesting that the astrocyte networks are formed after the maturation of neighboring neurons. On the molecular level, the expression of a selection of proteins is commonly used to characterize cells, in combination with cellular morphology. The most widely used marker in the characterisation of astrocytes is GFAP. In vivo mouse studies have shown that GFAP expression during astrocyte maturation peaks between E16-P1, slowly decreasing after this time point [105]. Mature astrocytes only express low levels of GFAP; however, GFAP expression in mature astrocytes may be upregulated following inflammatory activation [106]. Hence rather than being a marker for mature and functional astrocytes, GFAP expression is more indicative of either early immature astrocytes or reactive astrocytes. Another early developmental astrocyte marker is NF1A, an astrocyte specific transcription factor. This protein is highly expressed during mouse embryonic development (E10–E12.5), but its expression decreases as astrocytes mature [100]. Conversely to GFAP and NF1A, the expression of the astrocyte specific proteins aldolase C (ALDOC), GLAST and GLT1 increase as astrocytes mature [102, 107, 108]. These proteins therefore provide more appropriate markers of mature astrocytes than the commonly used GFAP. In humans, expression of the GLAST and GLT1 homologues, excitatory amino acid transporter 1 (EAAT1) and EAAT2, respectively, also increases with gestation time, based on studies of post mortem fetal brain tissue [109]. As such, increases in EAAT1 and EAAT2, coupled with a decrease in the early astrocyte markers, GFAP and NF1A, could be used as a robust indication of astrocyte maturation. Other astrocyte specific protein markers commonly chosen are S100β, aldehyde dehydrogenase 1 family member L1 (ALDH1L1) and the surface marker protein, CD44. The expression of these proteins increases at approximately E13 in mice and the expression profile does not significantly change throughout development [110–113]. Thus, while being astrocyte specific markers, these proteins are unsuitable indicators of astrocyte maturity. Although the expression profile of astrocyte-specific proteins provides important information regarding their maturity, it does not offer information about their functionality. Astrocytes provide a support network for neurons by closely monitoring and responding to the extracellular environment. One of the main functions of astrocytes is protecting neurons from excess neurotransmitter stimulation, by taking up glutamate via GLT1 and GLAST [114]. Additionally, astrocytes are a key player of the innate immune system of the CNS. As such, they express toll-like receptors (TLRs), which are able to recognize foreign particles [115]. Activation of TLRs leads to the release of chemical signaling molecules, such as cytokines and chemokines [e.g., interleukin-6 (IL-6), interferon-β and fractalkine (CX3CL1)] [116]. These molecules recruit and activate other immune cells to the site of injury. Furthermore, activated astrocytes also release neurotrophic factors, such as NGF, CNTF and BDNF to aid neuronal survival and regeneration. The controlled release of these immune and neurotrophic factors is thus a critical aspect of astrocyte function, which needs to be confirmed during in vitro differentiation. Overall, a panel of marker proteins should be used to identify the presence of mature astrocytes. Verifying their functional profile, particularly immune response and neurotransmitter handling, will ensure the relevance of in vitro astrocytes for modeling disease-relevant processes. Current differentiation strategies for astrocytes The generation of mature astrocytes for the purpose of studying neurological diseases has led to the development of several astrocyte differentiation protocols from pluripotent stem cells. However, these protocols vary in their duration, growth factor conditions and efficiency (Table 3).Table 3 Comparison of astrocyte differentiation protocols Differentiation protocol Krencik et al. [117] Emdad et al. [118] Serio et al. [119] Shaltouki et al. [120] Roybon et al. [122] Mormone et al. [121] Duration (days) 120 35 49 42 90 35 Efficiency (GFAP+ cells) 90 % 70 % 90 % 70 % 70 % 55 % (without sorting) Growth factors CNTF (or LIF), EGF, FGF2 CNTF, EGF, FGF2 CNTF, EGF, FGF2 LIF CNTF, FGF2, NRG1β1 AA, BDNF, CNTF, FGF2, GDNF, IGF, RA; maturation induced by withdrawal of growth factors CNTF, EGF, FGF2 Developmental markers (protein, mRNA) CD44, GFAP, NF1A GFAP GFAP, NF1A, vimentin CD44, GFAP, NF1A AQP4, CD44, GFAP, NF1A, S100β, vimentin GFAP Maturity markers (protein, mRNA) S100β AQP4, EAAT1 EAAT1, S100β ALDOC, EAAT1, S100β ALDH1L1, EAAT1, EAAT2 ALDOC, EAAT2 Physiological function Propagation of Ca2+ waves, glutamate uptake Migratory properties Glutamate uptake, promotion of synaptogenesis in neuron co-cultures Glutamate uptake, promotion of synaptogenesis in neuron co-cultures Propagation of Ca2+, glutamate uptake, inflammatory response (IL-6 release) Migratory properties The generation of neural rosettes and NPCs via embryoid body formation is similar across all protocols, lasting approximately 21 days. However major variations between protocols arise during the differentiation and maturation of NPCs to functional astrocytes. Once NPCs are present, the most common treatment in the generation of astrocytes is supplementation with epidermal growth factor (EGF) and FGF2 (Fig. 3), leading to the generation of early astrocytes, which has been confirmed by the expression of CD44, NF1A and vimentin [117–121]. Once the presence of early astrocytes is confirmed, the maturation of these cells is generally triggered by the addition of CNTF (Fig. 3). Although the growth factor conditions are very similar for most of the protocols (Table 3), a major difference lies in the maturation duration between each method. While Krencik et al. [117] matured the differentiated astrocytes for 100 days, the maturation time described by Emdad et al. [118], Serio et al. [119], Shaltouki et al. [120] and Mormone et al. [121] vary between 14 and 21 days. The discrepancy in maturation time was reflected in the proportion of GFAP-expressing cells, which varied between 55 % [121] and 90 % [117, 119]. To validate the functionality of the generated cells, some of these studies also confirmed the ability of astrocytes to take up glutamate from the cell medium [117, 119, 120].Fig. 3 Schematic diagram of astrocyte differentiation. The differentiation of astrocytes from pluripotent stem cell colonies follows the early neuronal developmental progress through spherical and adherent culture stages. Glial progenitor formation is triggered by the supplementation of EGF and FGF2, with CNTF being required for transition to mature astrocytes and aided by the neurotrophic factor NRG1β. The presence or absence of developmental and maturation markers are essential guides to monitor the differentiation progress at each culturing stage towards mature, functional astrocytes Unlike the previously described astrocyte differentiation protocols, Roybon et al. [122] used a different approach in differentiating astrocytes. The generated NPCs were not directly differentiated into astrocytes; instead the cells were first caudalized using RA and ascorbic acid (AA), prior to treatment with neurotrophic factors, including CNTF, to generate neurons. The differentiation of astrocytes was induced by the withdrawal of neurotrophic factors and the introduction of foetal bovine serum. The generated astrocytes were matured for 50–90 days. While previous protocols assessed the maturation of cells with the expression of GFAP, Roybon et al. [122] used this protein as a marker for early astrocytes and EAAT1, EAAT2 and ALDH1L1 to identify mature astrocytes. Moreover, the differentiated astrocytes took up glutamate and released IL-6 upon stimulation with tumor necrosis factor-α and IL-1β [122]. This study provides initial evidence that the differentiated cells are able to mount an immune response; however, the complexity of this activation requires further characterisation. Further evaluations of astrocyte maturation and function were conducted by transplanting pluripotent stem cell derived astrocytes into mice brains [117, 120–122]. These in vivo experiments primarily demonstrated that transplanted cells retained their astrocyte identity as indicated by the expression of GFAP. Krencik et al. demonstrated that transplanted differentiated astrocytes were able to form direct contact with cerebral blood vessels after 6 months, suggesting that in vitro derived cells are able to mature in vivo towards functional astrocytes [117]. Conversely, however, Roybon et al. analyzed marker expression of engrafted astrocytes in vivo and found high expression of immature markers, such as NF1A [122]. Thus, even in vivo transplantation of in vitro generated astrocytes may not be able to promote full maturation of these cells. Further characterisation of in vitro differentiated astrocytes is therefore essential. In summary, current astrocyte differentiation protocols use growth factors aligned with reported in vivo development. The heavy reliance on GFAP as an indicator for differentiation success and the limited reporting of functional characteristics prevent a reliable assessment of the maturity of the created cells. Furthermore, the differentiation periods vary vastly between the protocols, highlighting the uncertainty about the required phenotype of the generated astrocytes for relevant experimentation. To date, the protocol described by Krencik et al. [117] (and Serio et al. [119]) has reported the highest proportion of GFAP+ cells (90 %) with the confirmation of physiological functions, such as the propagation of Ca2+ waves and glutamate handling, following a labor-intensive maturation duration of 4–6 months. Roybon et al. reduced the maturation time of pluripotent stem cell derived astrocytes to 90 days, while retaining a similar efficiency [122]. This protocol [122] relies on several markers to track the maturity of the generated astrocytes and reports the most extensive physiological confirmation by showing the propagation of Ca2+, glutamate handling and inflammatory response upon activation. For the purpose of accurate disease modeling, the protocol by Roybon et al. shows the essential maturity confirmation necessary, consisting of a thorough expression analysis of maturity markers (EAAT1, EAAT2 and ALDOC), in conjunction with functional characterisation (neurotransmitter processing and inflammation response) relevant to the studied disease. Assessing differentiation techniques Checklist for high quality and reproducible differentiated cultures in vitro Induced pluripotent stem cells offer much promise in developing in vitro models to understand neurodegenerative disease mechanisms and for testing potential therapeutics. However, it is essential to generate a meaningful cell population that is both physiologically and clinically relevant. Evidently, an understanding of the desired cell type and its key characteristics is necessary to deliver high quality and reproducible cultures. The following are common experimental considerations to promote cell culture purity and consistency when performing or assessing differentiation:Adjust culture environment for each differentiation stage. The extracellular matrix (ECM) directly and indirectly affects the maintenance and differentiation of stem cell and neural cultures [123]. Human ESCs and iPSCs have been historically cultured on living feeder cells for mechanical and chemical stimulation, but recent research has shown that complex ECM protein combinations are superior in promoting culture stability and proliferation [124, 125]. The ECM dynamically evolves during neurodevelopment to form separate compartments in the CNS [126], comprised of different ECM protein combinations [127, 128]. Recapitulating in vivo neurodevelopment in culture thus requires adjustment of the ECM environment for the desired cell type and developmental stage. Differentiation of neocortical cells has shown to benefit from combining ECM proteins such as collagen I and fibronectin [129], but new differentiation protocols might require performing an ECM microarray [130]. Advanced three-dimensional culture surface modification can additionally be used to further the maturation of the desired cell type [87]. Optimize cell counts and the maturation duration of the protocol. Cell plating density affects differentiation, potentially due to alterations in effective local growth factor concentrations. Careful assessment of the appropriate cell density for plating needs to be optimized since different cell lines and different cell types proliferate at different rates. Longer protocols may be required for mature phenotypes but can lead to greater differences in the development of the required cell type compared to contaminating (unwanted) cell types. For example, developing neurons exit the cell cycle, whereas contaminating cells may continue to divide unless removed or their growth is inhibited. Identify checkpoints along the differentiation pathway and use a panel of cell markers. Expression of appropriate markers at checkpoints should be confirmed; also consider the use of positive or negative selection of cells, based on cell surface marker expression. At the final stage of differentiation the expression of a robust panel of cell-specific markers should be assessed for each line. Develop reporters to assess differentiated cell types. Reporters using fluorescent protein expression driven by an appropriate promoter can be used to identify cell types of interest. For example, motor neurons differentiated from ESCs have been identified in mixed cultures via lentiviral delivery of a construct bearing a GFP driven by the homeobox-9 promoter [131] and stable GFAP-driven TagRFP has been used to select iPSC-derived astrocytes [132]. Similar methods can be employed for other cell types using the promoter sequence of an appropriate cell-specific marker. Automated imaging options, such as the Incucyte live cell imaging system, can be used to provide images that are free from operator bias. Furthermore, fluorescent protein expression can also be used to sort the cells following differentiation, increasing the purity of the cell type of interest. Monitor the quality of cultures by assessing cellular functionality. The expression of a panel of marker proteins provides some evidence that the appropriate cell type has been generated. However, there are very few proteins that are expressed in one cell type alone; many commonly used (and supposedly cell-specific) markers are often expressed in multiple cell types. Further tests are therefore required to confirm that functional cells have been generated. Functionality can be assessed using in vitro live cell assays and should focus on non-biased assessments, such as receptor ligand quantification in culture medium, prior to experimental use. The steps involved in developing and optimizing a differentiation protocol are outlined in Box 1. Increasing the quality of cultures—suggestions for improvement Cell culture experiments need to mimic the physiological environment as closely as possible, in order to yield biologically relevant conclusions. The following developments will allow the experimental set-up to more closely represent the in vivo cellular background:Use of a hypoxic chamber to prevent chronic oxidative stress. The majority of cell culture experiments are performed under atmospheric oxygen conditions, i.e. 21 % O2. However cells in the brain and much of the body experience far lower levels, thought to range between 1 and 11 % O2 [133]. Recent data suggests that the process of reprogramming somatic cells to iPSCs is more efficient when performed under hypoxic conditions [134, 135]. Reprogramming requires a shift in cellular metabolism from oxidative to glycolytic conditions [10] and the hypoxia-inducible factors play a role in the coordination of these metabolic changes [136]. Comparisons of differentiation protocols under atmospheric vs. physiological O2 conditions are still limited. However, in the case of differentiation of cardiomyocytes, hypoxic culture increased differentiation yields by up to 1000-fold [137]. Measurements throughout the brain suggest that local oxygen tensions are heterogeneous in nature, exhibiting spatial and temporal differences depending on the microenvironment [138]. For example, neurons of the cerebral cortex experience a low oxygen field compared to venous O2 [138]. Neurons in such an oxygen environment are more sensitive to changes in cerebral blood flow, in the provision of oxygen and nutrients and are potentially more sensitive to oxidative stress. Investigations using in vivo imaging, for example by two-photon microscopy, allow the mapping of the partial pressure of oxygen at the μm/s resolution [139]. A detailed understanding of the local oxygen environment will allow us to provide in vivo conditions during in vitro experiments, an important consideration for further development. Future studies need to consider the implications of performing differentiation, and experiments, at atmospheric O2. Problems with feeding cycles could be partially overcome with microfluidics or bioreactors. A consideration in cell differentiation is the consistent and appropriate provision of nutrients and growth factors. Traditional cell feeding cycles of every 24 or 48 h can lead to dramatic variations in nutrient delivery and gradients of trophic factors. During the course of a culture, the concentrations of glucose, growth factors and micronutrients can vary widely in the growth medium, along with the pH. The proliferation and differentiation of stem cells is drastically affected by growth medium supplements [140], with the outcome that under different nutrient conditions the same cell line could yield different phenotypes. Microfluidics can be used to manipulate the delivery of medium on the micrometer scale, allowing for more controlled temporal and spatial supply of nutrients [141]. Alternatively larger cell preparations could be cultured in bioreactors that stir the growth medium; agitation has led to higher yields of differentiated cells in some protocols [137]. The use of bioreactors that are engineered to cultivate three-dimensional cultures, under hypoxic conditions, provides a further promising development [142]. Development of accurate co-culture systems and three-dimensional cultures. The architecture of the brain is clearly a complex multicellular environment. The presence of diverse cell types in cultures can be detrimental when studying cell-specific effects. However, a mixture of cell types may be required to understand a particular cellular process, such as neuronal:glial interactions or neuroinflammation. In addition, other cell types may be required for appropriate maturation. For example, astrocytes are required for the maturation of functional synapses [143] and in organizing the neuronal extracellular matrix [144]. As differentiation protocols improve so does our capacity for generating the appropriate mix of cells that function together. A key aspect of many neurodegenerative disorders is that multiple cell types are affected. Being able to accurately model the system holistically would be a huge leap forward in understanding the role of cellular interactions in disease pathogenesis. The generation of specific neuronal and glial subtypes remains under-developed. To accurately model the brain for neurodegenerative disease research, future studies will need to address the issue of subtype specification, potentially incorporating single cell analysis. In the case of excitable cells, cell function can be analyzed by electrophysiological assessment, coupled with single cell reverse transcription PCR to characterize the molecular signature of the cells generated. Faithful recapitulation of the molecular profile of specific cell subtypes is an important goal for future research. Conclusions Creating in vitro models of central nervous system disorders with human stem cells provides the medical research community with a powerful new research tool. To ensure the use of this option to its full potential, differentiation strategies need to be carefully planned and executed depending on the cell type desired and the experimental read-out. Confirmation of a robust panel of cell-specific markers, coupled with functional assays, is further required to provide evidence that the appropriate developmental pathway has been effectively recapitulated. Future studies need to focus on cultivating cells in more physiologically relevant environments by manipulating oxygen tension, overcoming issues with growth factor and nutrient gradients and developing multicellular and three-dimensional culture systems. Employing these quality improvement and control measures will lead to more reliable and reproducible results with strong clinical relevance. This work was supported by the National Health and Medical Research Council (NHMRC) of Australia (APP1071250 and APP1079995). Compliance with ethical standards Conflict of interest The authors declare that they have no conflict of interest. ==== Refs References 1. Csete M Translational prospects for human induced pluripotent stem cells Regen Med 2010 5 509 519 10.2217/rme.10.39 20632855 2. Devine MJ Ryten M Vodicka P Parkinson’s disease induced pluripotent stem cells with triplication of the α-synuclein locus Nat Commun 2011 2 440 10.1038/ncomms1453 21863007 3. Israel MA Yuan SH Bardy C Probing sporadic and familial Alzheimer’s disease using induced pluripotent stem cells Nature 2012 482 216 220 22278060 4. Ooi L Sidhu K Poljak A Induced pluripotent stem cells as tools for disease modelling and drug discovery in Alzheimer’s disease J Neural Transm 2013 120 103 111 10.1007/s00702-012-0839-2 22695755 5. Egawa N Kitaoka S Tsukita K Drug screening for ALS using patient-specific induced pluripotent stem cells Sci Transl Med 2012 4 145ra104 10.1126/scitranslmed.3004052 22855461 6. Brennand KJ Simone A Jou J Modelling schizophrenia using human induced pluripotent stem cells Nature 2011 473 221 225 10.1038/nature09915 21490598 7. Revilla A González C Iriondo A Current advances in the generation of human iPS cells: implications in cell-based regenerative medicine J Tissue Eng Regen Med 2015 25758460 8. Brennand K Savas JN Kim Y Phenotypic differences in hiPSC NPCs derived from patients with schizophrenia Mol Psychiatry 2014 24686136 9. Hartley BJ Tran N Ladran I Dopaminergic differentiation of schizophrenia hiPSCs Mol Psychiatry 2015 20 549 550 10.1038/mp.2014.194 25623947 10. López-Iglesias P Alcaina Y Tapia N Hypoxia induces pluripotency in primordial germ cells by HIF1α stabilization and Oct4 deregulation Antioxid Redox Signal 2015 22 205 223 10.1089/ars.2014.5871 25226357 11. Lu HF Lim S-X Leong MF Efficient neuronal differentiation and maturation of human pluripotent stem cells encapsulated in 3D microfibrous scaffolds Biomaterials 2012 33 9179 9187 10.1016/j.biomaterials.2012.09.006 22998816 12. Huang EJ Reichardt LF Neurotrophins: roles in neuronal development and function Annu Rev Neurosci 2001 24 677 736 10.1146/annurev.neuro.24.1.677 11520916 13. Hua JY Smith SJ Neural activity and the dynamics of central nervous system development Nat Neurosci 2004 7 327 332 10.1038/nn1218 15048120 14. Luján R Shigemoto R López-Bendito G Glutamate and GABA receptor signalling in the developing brain Neuroscience 2005 130 567 580 10.1016/j.neuroscience.2004.09.042 15590141 15. Lee S Huang EJ Modeling ALS and FTD with iPSC-derived neurons Brain Res 2015 16. Mufson EJ Counts SE Perez SE Ginsberg SD Cholinergic system during the progression of Alzheimer’s disease: therapeutic implications Expert Rev Neurother 2008 8 1703 1718 10.1586/14737175.8.11.1703 18986241 17. Whitehouse PJ Price DL Struble RG Alzheimer’s disease and senile dementia: loss of neurons in the basal forebrain Science 1982 215 1237 1239 10.1126/science.7058341 7058341 18. Li X-J Zhang X Johnson MA Coordination of sonic hedgehog and Wnt signaling determines ventral and dorsal telencephalic neuron types from human embryonic stem cells Development 2009 136 4055 4063 10.1242/dev.036624 19906872 19. Takahashi H Liu F-C Genetic patterning of the mammalian telencephalon by morphogenetic molecules and transcription factors Birth Defects Res C Embryo Today 2006 78 256 266 10.1002/bdrc.20077 17061260 20. Woolf NJ Cholinergic systems in mammalian brain and spinal cord Prog Neurobiol 1991 37 475 524 10.1016/0301-0082(91)90006-M 1763188 21. Braskie MN Thompson PM Understanding cognitive deficits in Alzheimer’s disease based on neuroimaging findings Trends Cogn Sci 2013 17 510 516 10.1016/j.tics.2013.08.007 24029445 22. Casanova MF Starkstein SE Jellinger KA Clinicopathological correlates of behavioral and psychological symptoms of dementia Acta Neuropathol 2011 122 117 135 10.1007/s00401-011-0821-3 21455688 23. Haubst N Berger J Radjendirane V Molecular dissection of Pax6 function: the specific roles of the paired domain and homeodomain in brain development Development 2004 131 6131 6140 10.1242/dev.01524 15548580 24. Holm PC Mader MT Haubst N Loss- and gain-of-function analyses reveal targets of Pax6 in the developing mouse telencephalon Mol Cell Neurosci 2007 34 99 119 10.1016/j.mcn.2006.10.008 17158062 25. Tao W Lai E Telencephalon-restricted expression of BF-1, a new member of the HNF-3/fork head gene family, in the developing rat brain Neuron 1992 8 957 966 10.1016/0896-6273(92)90210-5 1350202 26. Consonni S Leone S Becchetti A Amadeo A Developmental and neurochemical features of cholinergic neurons in the murine cerebral cortex BMC Neurosci 2009 10 18 10.1186/1471-2202-10-18 19272148 27. Lopez-Coviella I Follettie MT Mellott TJ Bone morphogenetic protein 9 induces the transcriptome of basal forebrain cholinergic neurons Proc Natl Acad Sci USA 2005 102 6984 6989 10.1073/pnas.0502097102 15870197 28. Knüsel B Winslow JW Rosenthal A Promotion of central cholinergic and dopaminergic neuron differentiation by brain-derived neurotrophic factor but not neurotrophin 3 Proc Natl Acad Sci USA 1991 88 961 965 10.1073/pnas.88.3.961 1992488 29. Knusel B Michel PP Schwaber JS Hefti F Selective and nonselective stimulation of central cholinergic and dopaminergic development in vitro by nerve growth factor, basic fibroblast growth factor, epidermal growth factor, insulin and the insulin-like growth factors I and II J Neurosci 1990 10 558 570 2406380 30. Sofroniew MV Howe CL Mobley WC Nerve growth factor signaling, neuroprotection, and neural repair Annu Rev Neurosci 2001 24 1217 1281 10.1146/annurev.neuro.24.1.1217 11520933 31. Matsumoto K Tanaka T Furuyama T L3, a novel murine LIM-homeodomain transcription factor expressed in the ventral telencephalon and the mesenchyme surrounding the oral cavity Neurosci Lett 1996 204 113 116 10.1016/0304-3940(96)12341-7 8929991 32. Bissonnette CJ Lyass L Bhattacharyya BJ The controlled generation of functional basal forebrain cholinergic neurons from human embryonic stem cells Stem Cells 2011 29 802 811 10.1002/stem.626 21381151 33. Ha DH Robertson RT Ribak CE Weiss JH Cultured basal forebrain cholinergic neurons in contact with cortical cells display synapses, enhanced morphological features, and decreased dependence on nerve growth factor J Comp Neurol 1996 373 451 465 10.1002/(SICI)1096-9861(19960923)373:3<451::AID-CNE9>3.0.CO;2-1 8889938 34. Wu H Williams J Nathans J Complete morphologies of basal forebrain cholinergic neurons in the mouse Elife 2014 3 e02444 24894464 35. Lamour Y Dutar P Rascol O Jobert A Basal forebrain neurons projecting to the rat frontoparietal cortex: electrophysiological and pharmacological properties Brain Res 1986 362 122 131 10.1016/0006-8993(86)91405-8 3002548 36. Markram H Segal M Electrophysiological characteristics of cholinergic and non-cholinergic neurons in the rat medial septum-diagonal band complex Brain Res 1990 513 171 174 10.1016/0006-8993(90)91106-Q 2350680 37. Crompton LA Byrne ML Taylor H Stepwise, non-adherent differentiation of human pluripotent stem cells to generate basal forebrain cholinergic neurons via hedgehog signaling Stem Cell Res 2013 11 1206 1221 10.1016/j.scr.2013.08.002 24013066 38. Moises HC Womble MD Washburn MS Williams LR Nerve growth factor facilitates cholinergic neurotransmission between nucleus basalis and the amygdala in rat: an electrophysiological analysis J Neurosci 1995 15 8131 8142 8613748 39. Nilbratt M Porras O Marutle A Neurotrophic factors promote cholinergic differentiation in human embryonic stem cell-derived neurons J Cell Mol Med 2010 14 1476 1484 10.1111/j.1582-4934.2009.00916.x 19799651 40. Duan L Bhattacharyya BJ Belmadani A Stem cell derived basal forebrain cholinergic neurons from Alzheimer’s disease patients are more susceptible to cell death Mol Neurodegener 2014 9 3 10.1186/1750-1326-9-3 24401693 41. Liu Y Weick JP Liu H Medial ganglionic eminence-like cells derived from human embryonic stem cells correct learning and memory deficits Nat Biotechnol 2013 31 440 447 10.1038/nbt.2565 23604284 42. Morrow BA Redmond DE Roth RH Elsworth JD Development of A9/A10 dopamine neurons during the second and third trimesters in the African green monkey J Comp Neurol 2005 488 215 223 10.1002/cne.20599 15924344 43. Prakash N Wurst W Development of dopaminergic neurons in the mammalian brain Cell Mol Life Sci 2006 63 187 206 10.1007/s00018-005-5387-6 16389456 44. Vernier P Moret F Callier S The degeneration of dopamine neurons in Parkinson’s disease: insights from embryology and evolution of the mesostriatocortical system Ann N Y Acad Sci 2004 1035 231 249 10.1196/annals.1332.015 15681811 45. Wallén A Zetterström RH Solomin L Fate of mesencephalic AHD2-expressing dopamine progenitor cells in NURR1 mutant mice Exp Cell Res 1999 253 737 746 10.1006/excr.1999.4691 10585298 46. Kawano H Ohyama K Kawamura K Nagatsu I Migration of dopaminergic neurons in the embryonic mesencephalon of mice Brain Res Dev Brain Res 1995 86 101 113 10.1016/0165-3806(95)00018-9 7544698 47. Bayer SA Wills KV Triarhou LC Ghetti B Time of neuron origin and gradients of neurogenesis in midbrain dopaminergic neurons in the mouse Exp Brain Res 1995 105 191 199 7498372 48. Martí J Wills KV Ghetti B Bayer SA A combined immunohistochemical and autoradiographic method to detect midbrain dopaminergic neurons and determine their time of origin Brain Res Brain Res Protoc 2002 9 197 205 10.1016/S1385-299X(02)00145-9 12113779 49. Morrow BA Roth RH Redmond DE Apoptotic natural cell death in developing primate dopamine midbrain neurons occurs during a restricted period in the second trimester of gestation Exp Neurol 2007 204 802 807 10.1016/j.expneurol.2007.01.009 17313945 50. Voorn P Kalsbeek A Jorritsma-Byham B Groenewegen HJ The pre- and postnatal development of the dopaminergic cell groups in the ventral mesencephalon and the dopaminergic innervation of the striatum of the rat Neuroscience 1988 25 857 887 10.1016/0306-4522(88)90041-3 3405431 51. Abeliovich A Hammond R Midbrain dopamine neuron differentiation: factors and fates Dev Biol 2007 304 447 454 10.1016/j.ydbio.2007.01.032 17331494 52. Roussa E Krieglstein K Induction and specification of midbrain dopaminergic cells: focus on SHH, FGF8, and TGF-beta Cell Tissue Res 2004 318 23 33 10.1007/s00441-004-0916-4 15322912 53. Hayes L Ralls S Wang H Ahn S Duration of Shh signaling contributes to mDA neuron diversity Dev Biol 2013 374 115 126 10.1016/j.ydbio.2012.11.016 23201023 54. Joksimovic M Anderegg A Roy A Spatiotemporally separable Shh domains in the midbrain define distinct dopaminergic progenitor pools Proc Natl Acad Sci USA 2009 106 19185 19190 10.1073/pnas.0904285106 19850875 55. Ye W Shimamura K Rubenstein JL FGF and Shh signals control dopaminergic and serotonergic cell fate in the anterior neural plate Cell 1998 93 755 766 10.1016/S0092-8674(00)81437-3 9630220 56. Kholodilov N Yarygina O Oo TF Regulation of the development of mesencephalic dopaminergic systems by the selective expression of glial cell line-derived neurotrophic factor in their targets J Neurosci 2004 24 3136 3146 10.1523/JNEUROSCI.4506-03.2004 15044553 57. Oo TF Kholodilov N Burke RE Regulation of natural cell death in dopaminergic neurons of the substantia nigra by striatal glial cell line-derived neurotrophic factor in vivo J Neurosci 2003 23 5141 5148 12832538 58. Horger BA Nishimura MC Armanini MP Neurturin exerts potent actions on survival and function of midbrain dopaminergic neurons J Neurosci 1998 18 4929 4937 9634558 59. Farkas LM Dünker N Roussa E Transforming growth factor-beta(s) are essential for the development of midbrain dopaminergic neurons in vitro and in vivo J Neurosci 2003 23 5178 5186 12832542 60. Castelo-Branco G Wagner J Rodriguez FJ Differential regulation of midbrain dopaminergic neuron development by Wnt-1, Wnt-3a, and Wnt-5a Proc Natl Acad Sci USA 2003 100 12747 12752 10.1073/pnas.1534900100 14557550 61. Schulte G Bryja V Rawal N Purified Wnt-5a increases differentiation of midbrain dopaminergic cells and dishevelled phosphorylation J Neurochem 2005 92 1550 1553 10.1111/j.1471-4159.2004.03022.x 15748172 62. Avantaggiato V Acampora D Tuorto F Simeone A Retinoic acid induces stage-specific repatterning of the rostral central nervous system Dev Biol 1996 175 347 357 10.1006/dbio.1996.0120 8626038 63. Baquet ZC Bickford PC Jones KR Brain-derived neurotrophic factor is required for the establishment of the proper number of dopaminergic neurons in the substantia nigra pars compacta J Neurosci 2005 25 6251 6259 10.1523/JNEUROSCI.4601-04.2005 15987955 64. Roeper J Dissecting the diversity of midbrain dopamine neurons Trends Neurosci 2013 36 336 342 10.1016/j.tins.2013.03.003 23582338 65. Gaven F Marin P Claeysen S Primary culture of mouse dopaminergic neurons J Vis Exp 2014 91 e51751 25226064 66. Smidt MP Burbach JPH How to make a mesodiencephalic dopaminergic neuron Nat Rev Neurosci 2007 8 21 32 10.1038/nrn2039 17180160 67. Pau KY Ma YJ Yu JH Topographic comparison of the expression of norepinephrine transporter, tyrosine hydroxylase and neuropeptide Y mRNA in association with dopamine beta-hydroxylase neurons in the rabbit brainstem Brain Res Mol Brain Res 1997 48 367 381 10.1016/S0169-328X(97)00113-7 9332734 68. White RB Thomas MG Moving beyond tyrosine hydroxylase to define dopaminergic neurons for use in cell replacement therapies for Parkinson’s disease CNS Neurol Disord Drug Targets 2012 11 340 349 10.2174/187152712800792758 22483315 69. Haber SN Ryoo H Cox C Lu W Subsets of midbrain dopaminergic neurons in monkeys are distinguished by different levels of mRNA for the dopamine transporter: comparison with the mRNA for the D2 receptor, tyrosine hydroxylase and calbindin immunoreactivity J Comp Neurol 1995 362 400 410 10.1002/cne.903620308 8576447 70. Kessler MA Yang M Gollomp KL The human tyrosine hydroxylase gene promoter Brain Res Mol Brain Res 2003 112 8 23 10.1016/S0169-328X(02)00694-0 12670698 71. Zhou W Lee YM Guy VC Freed CR Embryonic stem cells with GFP knocked into the dopamine transporter yield purified dopamine neurons in vitro and from knock-in mice Stem Cells 2009 27 2952 2961 10.1002/stem.201 19750538 72. Ganat YM Calder EL Kriks S Identification of embryonic stem cell-derived midbrain dopaminergic neurons for engraftment J Clin Invest 2012 122 2928 2939 10.1172/JCI58767 22751106 73. Hartfield EM Yamasaki-Mann M Ribeiro Fernandes HJ Physiological characterisation of human iPS-derived dopaminergic neurons PLoS One 2014 9 e87388 10.1371/journal.pone.0087388 24586273 74. Thompson L Barraud P Andersson E Identification of dopaminergic neurons of nigral and ventral tegmental area subtypes in grafts of fetal ventral mesencephalon based on cell morphology, protein expression, and efferent projections J Neurosci 2005 25 6467 6477 10.1523/JNEUROSCI.1676-05.2005 16000637 75. Grace AA Onn SP Morphology and electrophysiological properties of immunocytochemically identified rat dopamine neurons recorded in vitro J Neurosci 1989 9 3463 3481 2795134 76. Yan Y Yang D Zarnowska ED Directed differentiation of dopaminergic neuronal subtypes from human embryonic stem cells Stem Cells 2005 23 781 790 10.1634/stemcells.2004-0365 15917474 77. Ungless MA Grace AA Are you or aren’t you? Challenges associated with physiologically identifying dopamine neurons Trends Neurosci 2012 35 422 430 10.1016/j.tins.2012.02.003 22459161 78. Kötter R Feizelmeier M Species-dependence and relationship of morphological and electrophysiological properties in nigral compacta neurons Prog Neurobiol 1998 54 619 632 10.1016/S0301-0082(97)00088-9 9550193 79. Kuznetsov AS Kopell NJ Wilson CJ Transient high-frequency firing in a coupled-oscillator model of the mesencephalic dopaminergic neuron J Neurophysiol 2006 95 932 947 10.1152/jn.00691.2004 16207783 80. Chambers SM Fasano CA Papapetrou EP Highly efficient neural conversion of human ES and iPS cells by dual inhibition of SMAD signaling Nat Biotechnol 2009 27 275 280 10.1038/nbt.1529 19252484 81. Kriks S Shim J-W Piao J Dopamine neurons derived from human ES cells efficiently engraft in animal models of Parkinson’s disease Nature 2011 480 547 551 22056989 82. Hur E-M Zhou F-Q GSK3 signalling in neural development Nat Rev Neurosci 2010 11 539 551 10.1038/nrn2870 20648061 83. Petit I Kesner NS Karry R Induced pluripotent stem cells from hair follicles as a cellular model for neurodevelopmental disorders Stem Cell Res 2012 8 134 140 10.1016/j.scr.2011.09.003 22099027 84. Reinhardt P Glatza M Hemmer K Derivation and expansion using only small molecules of human neural progenitors for neurodegenerative disease modeling PLoS One 2013 8 e59252 10.1371/journal.pone.0059252 23533608 85. Fathi A Rasouli H Yeganeh M Efficient differentiation of human embryonic stem cells toward dopaminergic neurons using recombinant LMX1A factor Mol Biotechnol 2015 57 184 194 10.1007/s12033-014-9814-5 25380985 86. Perrier AL Tabar V Barberi T Derivation of midbrain dopamine neurons from human embryonic stem cells Proc Natl Acad Sci USA 2004 101 12543 12548 10.1073/pnas.0404700101 15310843 87. Tan KKB Tann JY Sathe SR Enhanced differentiation of neural progenitor cells into neurons of the mesencephalic dopaminergic subtype on topographical patterns Biomaterials 2015 43 32 43 10.1016/j.biomaterials.2014.11.036 25591959 88. Lobsiger CS Cleveland DW Glial cells as intrinsic components of non-cell-autonomous neurodegenerative disease Nat Neurosci 2007 10 1355 1360 10.1038/nn1988 17965655 89. Hennessy E Griffin ÉW Cunningham C Astrocytes are primed by chronic neurodegeneration to produce exaggerated chemokine and cell infiltration responses to acute stimulation with the cytokines IL-1β and TNF-α J Neurosci 2015 35 8411 8422 10.1523/JNEUROSCI.2745-14.2015 26041910 90. Kessaris N Pringle N Richardson WD Specification of CNS glia from neural stem cells in the embryonic neuroepithelium Philos Trans R Soc Lond B Biol Sci 2008 363 71 85 10.1098/rstb.2006.2013 17282992 91. Ybot-Gonzalez P Gaston-Massuet C Girdler G Neural plate morphogenesis during mouse neurulation is regulated by antagonism of Bmp signalling Development 2007 134 3203 3211 10.1242/dev.008177 17693602 92. Anthony TE Klein C Fishell G Heintz N Radial glia serve as neuronal progenitors in all regions of the central nervous system Neuron 2004 41 881 890 10.1016/S0896-6273(04)00140-0 15046721 93. Sancho-Tello M Vallés S Montoliu C Developmental pattern of GFAP and vimentin gene expression in rat brain and in radial glial cultures Glia 1995 15 157 166 10.1002/glia.440150208 8567067 94. Sild M Ruthazer ES Radial glia: progenitor, pathway, and partner Neuroscientist 2011 17 288 302 10.1177/1073858410385870 21558559 95. Noctor SC Flint AC Weissman TA Neurons derived from radial glial cells establish radial units in neocortex Nature 2001 409 714 720 10.1038/35055553 11217860 96. Yang Y Higashimori H Morel L Developmental maturation of astrocytes and pathogenesis of neurodevelopmental disorders J Neurodev Disord 2013 5 22 10.1186/1866-1955-5-22 23988237 97. Nakashima K Wiese S Yanagisawa M Developmental requirement of gp130 signaling in neuronal survival and astrocyte differentiation J Neurosci 1999 19 5429 5434 10377352 98. Hong S Song M-R STAT3 but not STAT1 is required for astrocyte differentiation PLoS One 2014 9 e86851 10.1371/journal.pone.0086851 24466267 99. Sun Y Nadal-Vicens M Misono S Neurogenin promotes neurogenesis and inhibits glial differentiation by independent mechanisms Cell 2001 104 365 376 10.1016/S0092-8674(01)00224-0 11239394 100. Deneen B Ho R Lukaszewicz A The transcription factor NFIA controls the onset of gliogenesis in the developing spinal cord Neuron 2006 52 953 968 10.1016/j.neuron.2006.11.019 17178400 101. Takahashi T Misson J-P Caviness VS Glial process elongation and branching in the developing murine neocortex: a qualitative and quantitative immunohistochemical analysis J Comp Neurol 1990 302 15 28 10.1002/cne.903020103 2086612 102. Kugler P Schleyer V Developmental expression of glutamate transporters and glutamate dehydrogenase in astrocytes of the postnatal rat hippocampus Hippocampus 2004 14 975 985 10.1002/hipo.20015 15390174 103. Yang Y Gozen O Watkins A Presynaptic regulation of astroglial excitatory neurotransmitter transporter GLT1 Neuron 2009 61 880 894 10.1016/j.neuron.2009.02.010 19323997 104. Bushong EA Martone ME Ellisman MH Maturation of astrocyte morphology and the establishment of astrocyte domains during postnatal hippocampal development Int J Dev Neurosci 2004 22 73 86 10.1016/j.ijdevneu.2003.12.008 15036382 105. Agius E Decker Y Soukkarieh C Role of BMPs in controlling the spatial and temporal origin of GFAP astrocytes in the embryonic spinal cord Dev Biol 2010 344 611 620 10.1016/j.ydbio.2010.05.017 20488175 106. Brahmachari S Fung YK Pahan K Induction of glial fibrillary acidic protein expression in astrocytes by nitric oxide J Neurosci 2006 26 4930 4939 10.1523/JNEUROSCI.5480-05.2006 16672668 107. Regan MR Huang YH Kim YS Variations in promoter activity reveal a differential expression and physiology of glutamate transporters by glia in the developing and mature CNS J Neurosci 2007 27 6607 6619 10.1523/JNEUROSCI.0790-07.2007 17581948 108. Cahoy JD Emery B Kaushal A A transcriptome database for astrocytes, neurons, and oligodendrocytes: a new resource for understanding brain development and function J Neurosci 2008 28 264 278 10.1523/JNEUROSCI.4178-07.2008 18171944 109. Bar-Peled O Ben-Hur H Biegon A Distribution of glutamate transporter subtypes during human brain development J Neurochem 1997 69 2571 2580 10.1046/j.1471-4159.1997.69062571.x 9375691 110. Vives V Alonso G Solal AC Visualization of S100B-positive neurons and glia in the central nervous system of EGFP transgenic mice J Comp Neurol 2003 457 404 419 10.1002/cne.10552 12561079 111. Yang Y Vidensky S Jin L Molecular comparison of GLT1+ and ALDH1L1+ astrocytes in vivo in astroglial reporter mice Glia 2011 59 200 207 10.1002/glia.21089 21046559 112. Haidet-Phillips AM Hester ME Miranda CJ Astrocytes from familial and sporadic ALS patients are toxic to motor neurons Nat Biotechnol 2011 29 824 828 10.1038/nbt.1957 21832997 113. Zamanian JL Xu L Foo LC Genomic analysis of reactive astrogliosis J Neurosci 2012 32 6391 6410 10.1523/JNEUROSCI.6221-11.2012 22553043 114. Perego C Vanoni C Bossi M The GLT-1 and GLAST glutamate transporters are expressed on morphologically distinct astrocytes and regulated by neuronal activity in primary hippocampal cocultures J Neurochem 2000 75 1076 1084 10.1046/j.1471-4159.2000.0751076.x 10936189 115. Farina C Krumbholz M Giese T Preferential expression and function of Toll-like receptor 3 in human astrocytes J Neuroimmunol 2005 159 12 19 10.1016/j.jneuroim.2004.09.009 15652398 116. Farina C Aloisi F Meinl E Astrocytes are active players in cerebral innate immunity Trends Immunol 2007 28 138 145 10.1016/j.it.2007.01.005 17276138 117. Krencik R Weick JP Liu Y Specification of transplantable astroglial subtypes from human pluripotent stem cells Nat Biotechnol 2011 29 528 534 10.1038/nbt.1877 21602806 118. Emdad L D’Souza SL Kothari HP Efficient differentiation of human embryonic and induced pluripotent stem cells into functional astrocytes Stem Cells Dev 2011 21 404 410 10.1089/scd.2010.0560 21631388 119. Serio A Bilican B Barmada SJ Astrocyte pathology and the absence of non-cell autonomy in an induced pluripotent stem cell model of TDP-43 proteinopathy Proc Natl Acad Sci USA 2013 110 4697 4702 10.1073/pnas.1300398110 23401527 120. Shaltouki A Peng J Liu Q Efficient generation of astrocytes from human pluripotent stem cells in defined conditions Stem Cells 2013 31 941 952 10.1002/stem.1334 23341249 121. Mormone E D’Sousa S Alexeeva V “Footprint-free” human induced pluripotent stem cell-derived astrocytes for in vivo cell-based therapy Stem Cells Dev 2014 23 2626 2636 10.1089/scd.2014.0151 24914471 122. Roybon L Lamas NJ Garcia-Diaz A Human stem cell-derived spinal cord astrocytes with defined mature or reactive phenotypes Cell Rep 2013 4 1035 1048 10.1016/j.celrep.2013.06.021 23994478 123. Gattazzo F Urciuolo A Bonaldo P Extracellular matrix: a dynamic microenvironment for stem cell niche Biochim Biophys Acta 2014 1840 2506 2519 10.1016/j.bbagen.2014.01.010 24418517 124. Fukusumi H Shofuda T Kanematsu D Feeder-free generation and long-term culture of human induced pluripotent stem cells using pericellular matrix of decidua derived mesenchymal cells PLoS One 2013 8 e55226 10.1371/journal.pone.0055226 23383118 125. Lam MT Longaker MT Comparison of several attachment methods for human iPS, embryonic and adipose-derived stem cells for tissue engineering J Tissue Eng Regen Med 2012 6 Suppl 3 s80 s86 10.1002/term.1499 22610948 126. Bandtlow CE Zimmermann DR Proteoglycans in the developing brain: new conceptual insights for old proteins Physiol Rev 2000 80 1267 1290 11015614 127. Kwok JCF Dick G Wang D Fawcett JW Extracellular matrix and perineuronal nets in CNS repair Dev Neurobiol 2011 71 1073 1089 10.1002/dneu.20974 21898855 128. Rauch U Brain matrix: structure, turnover and necessity Biochem Soc Trans 2007 35 656 660 10.1042/BST0350656 17635114 129. Cooke MJ Zahir T Phillips SR Neural differentiation regulated by biomimetic surfaces presenting motifs of extracellular matrix proteins J Biomed Mater Res A 2010 93 824 832 19653304 130. Flaim CJ Chien S Bhatia SN An extracellular matrix microarray for probing cellular differentiation Nat Methods 2005 2 119 125 10.1038/nmeth736 15782209 131. Marchetto MCN Muotri AR Mu Y Non-cell-autonomous effect of human SOD1 G37R astrocytes on motor neurons derived from human embryonic stem cells Cell Stem Cell 2008 3 649 657 10.1016/j.stem.2008.10.001 19041781 132. Holmqvist S Brouwer M Djelloul M Generation of human pluripotent stem cell reporter lines for the isolation of and reporting on astrocytes generated from ventral midbrain and ventral spinal cord neural progenitors Stem Cell Res 2015 15 203 220 10.1016/j.scr.2015.05.014 26100233 133. Carreau A El Hafny-Rahbi B Matejuk A Why is the partial oxygen pressure of human tissues a crucial parameter? Small molecules and hypoxia J Cell Mol Med 2011 15 1239 1253 10.1111/j.1582-4934.2011.01258.x 21251211 134. Chen G Gulbranson DR Hou Z Chemically defined conditions for human iPSC derivation and culture Nat Methods 2011 8 424 429 10.1038/nmeth.1593 21478862 135. Liu Y Mukhopadhyay P Pisano MM Repression of Zeb1 and hypoxia cause sequential mesenchymal-to-epithelial transition and induction of aid, Oct4, and Dnmt1, leading to immortalization and multipotential reprogramming of fibroblasts in spheres Stem Cells 2013 31 1350 1362 10.1002/stem.1382 23554223 136. Mathieu J Zhou W Xing Y Hypoxia-inducible factors have distinct and stage-specific roles during reprogramming of human cells to pluripotency Cell Stem Cell 2014 14 592 605 10.1016/j.stem.2014.02.012 24656769 137. Correia C Serra M Espinha N Combining hypoxia and bioreactor hydrodynamics boosts induced pluripotent stem cell differentiation towards cardiomyocytes Stem Cell Rev 2014 10 786 801 10.1007/s12015-014-9533-0 25022569 138. Ndubuizu O LaManna JC Brain tissue oxygen concentration measurements Antioxid Redox Signal 2007 9 1207 1219 10.1089/ars.2007.1634 17536959 139. Lecoq J Parpaleix A Roussakis E Simultaneous two-photon imaging of oxygen and blood flow in deep cerebral vessels Nat Med 2011 17 893 898 10.1038/nm.2394 21642977 140. Shahdadfar A Frønsdal K Haug T In vitro expansion of human mesenchymal stem cells: choice of serum is a determinant of cell proliferation, differentiation, gene expression, and transcriptome stability Stem Cells 2005 23 1357 1366 10.1634/stemcells.2005-0094 16081661 141. Halldorsson S Lucumi E Gómez-Sjöberg R Fleming RMT Advantages and challenges of microfluidic cell culture in polydimethylsiloxane devices Biosens Bioelectron 2015 63 218 231 10.1016/j.bios.2014.07.029 25105943 142. Silva MM Rodrigues AF Correia C Robust expansion of human pluripotent stem cells: integration of bioprocess design with transcriptomic and metabolomic characterization Stem Cells Transl Med 2015 4 731 742 10.5966/sctm.2014-0270 25979863 143. Ullian EM Sapperstein SK Christopherson KS Barres BA Control of synapse number by glia Science 2001 291 657 661 10.1126/science.291.5504.657 11158678 144. Faissner A Pyka M Geissler M Contributions of astrocytes to synapse formation and maturation—potential functions of the perisynaptic extracellular matrix Brain Res Rev 2010 63 26 38 10.1016/j.brainresrev.2010.01.001 20096729

PMC005xxxxxx/PMC5002044.txt

==== Front Sci Bull (Beijing)Sci Bull (Beijing)Science Bulletin2095-92732095-9281Science China Press Beijing 114610.1007/s11434-016-1146-3ArticleSolid oxide fuel cell interconnect design optimization considering the thermal stresses Xu Min 1Li Tingshuai 1Yang Ming 1Andersson Martin martin.andersson@energy.lth.se 21 School of Energy Science and Engineering, University of Electronic Science and Technology of China, Chengdu, 611731 China 2 Department of Energy Sciences, Faculty of Engineering, Lund University, P.O. Box 118, 221 00 Lund, Sweden 20 7 2016 20 7 2016 2016 61 17 1333 1344 26 4 2016 12 6 2016 6 7 2016 © Science China Press and Springer-Verlag Berlin Heidelberg 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.The mechanical failure of solid oxide fuel cell (SOFC) components may cause cracks with consequences such as gas leakage, structure instability and reduction of cell lifetime. A comprehensive 3D model of the thermal stresses of an anode-supported planar SOFC is presented in this work. The main objective of this paper is to get an interconnect optimized design by evaluating the thermal stresses of an anode-supported SOFC for different designs, which would be a new criterion for interconnect design. The model incorporates the momentum, mass, heat, ion and electron transport, as well as steady-state mechanics. Heat from methane steam reforming and water–gas shift reaction were considered in our model. The results examine the relationship between the interconnect structures and thermal stresses in SOFC at certain mechanical properties. A wider interconnect of the anode side lowers the stress obviously. The simulation results also indicate that thermal stress of coflow design is smaller than that of counterflow, corresponding to the temperature distribution. This study shows that it is possible to design interconnects for an optimum thermal stress performance of the cell. Keywords Solid oxide fuel cellThermal stressesInterconnectOptimizationFinite element methodissue-copyright-statement© Science China Press and Springer-Verlag Berlin Heidelberg 2016 ==== Body Introduction The solid oxide fuel cell (SOFC) is considered to be one of the most promising new energy technologies for high energy efficiency, environmental friendly and fuel diversity [1, 2]. The performance of an SOFC is strongly related to properties of the electrode materials, seals or interconnect materials and impurities in the fuel [3–7]. During the past decades, there have been significant research aiming to increase the performance of SOFC by exploring new materials or improving synthesis techniques and optimizing the cell structure [8–11]. However, the SOFC technology still faces many challenges, before it could be commercialized, since cost reduction and an extended lifetime are required. Numerical simulation as an economic approach to design materials, cell or stack structure was used to develop the SOFC technology further. The effect of microstructure on effective ionic and electrical electrode conductivities was investigated by numerical optimization, which determines the effective conductivities and degradation of materials [12]. The mass and heat transport, and electrochemical reaction were considered for SOFC optimization by developing numerical models [13–15]. The mixing ratio of fuel, gas flow rate and reforming reaction, utilizing various fuels were studied to archive an optimized SOFC system. Beside the material structure, heat and mass transfer processes and electrochemical problems, the thermo-mechanical phenomena attracted significant attentions in recent years [16–18]. As the most important part of thermo-mechanical phenomena, thermal stresses occurring to various components in SOFC strongly impact the cell lifetime by inducing cracks, contact loss and structural instability [19]. Thermal stress analysis of a single cell can be used to evaluate the possibility of cracks or flaws that was not able to be detected earlier and easier in experimental works [20]. For a planar anode-supported SOFC, thermal stress and thermal fluid behavior can be analyzed using a 3D integrated numerical model [21]. Maximum principal stress within the cell increases with a high current and temperature gradient, which was in good agreement with the experimental work. Since the flow distribution also had influence on the temperature distribution, thermal stresses of planar SOFC operating at coflow and counterflow were compared [22]. The results indicates that the temperature gradient near the fuel inlet for counter-flow pattern is much larger compared to that of co-flow pattern. Thermal stress in an single cell is different compared to the stress distribution in the stack, i.e., future studies are needed. Jiang et al. [23] built a thermo-electrochemical-structure model combining both the finite-volume and the finite-element approaches to investigate the thermal behavior and the thermal-stress of a SOFC with the bonded compliant seal design. It was found that, the thermal stresses depended on the location and the cell voltage. The second largest stress region within the cell was near the inlet. Except the effects of materials properties, the operating conditions such as the load of cell in a stack should be investigated for practical modeling. Effects of the applied assembly load on the thermal stress distribution in the integrated planar SOFC stack with a compressive sealing design were characterized by a 3D multi-cell model [24]. The results showed that the expansion mismatch rather than the applied compressive load dominated the thermal stress distribution of the cell components. There was also evidence that ceramic components suffered significant stress when subjected to an idealized operating duty cycle [25]. Stresses were generated due to differential thermal expansion of the layers indicating that there is a high probability of failure during these phases of the duty cycle, i.e. when cell is cooled from sintering to room temperature or is heated from room temperature to operating temperature. However, the just mentioned studies mainly focused on the thermal stresses distribution, considering the multiphysics process or operating conditions but neglecting the interconnect structural effects. As a component for connecting the electrodes and loads, the interconnect is not only essential to provide paths for electron transport in single cell (or to the neighboring cell in stack) and to protect the electrode material from damages in an ambient environment, but can also maintain the thermal mechanical properties matching to the adjacent electro-active components. Therefore, the selection criteria of interconnect materials and optimization of the structure is more important and stringent than other components of the cell. The ceramic interconnect possesses a good stability and retains a fine compatibility with other components. However, the conductivity of ceramics is not appreciable below 600 °C and the poor sinterability of ceramic interconnects is also a challenge for implementation [26]. While, the metallic alloys are low cost, ease of fabricating and high mechanical strength, made it more attractive than ceramic oxides as interconnects in intermediate temperature SOFC stacks [27]. Normally, the interconnect optimization focused on the electrical performance, degradation processes or temperature distribution, since the structure and the shape of interconnects were related to concentration polarization and electric resistance [28–30]. However, the thermal stress depends on the temperature distribution and the load conditions of a complete single cell would be affected by the interconnect structure. With contact area increasing, power density and temperature gradient would be reduced when a decreasing size of collecting pins tended to gain a better temperature homogeneity and power density [31]. Beside the contact area and the size of collecting pins, addition of interconnect ribs is another option for modifying interconnect structure. While increasing the number of interconnect ribs and reducing the gap between them at the cathode side, the lateral conduction distance can be decreased, and an enhancement of the cell performance of more than 30 % can be achieved [32]. Besides, altering the width of ribs is easier both in numerical and experimental design. 3D numerical model was built to analyze the effect of rib width on cell performance [33]. The relationship, between the contact resistance and the optimal rib widths was given as a guidance for an engineering optimization. When considering the size of ribs, the increased width gives a better conduction of the electrical current and reduced ohmic losses, while narrow ribs are needed to facilitate a more uniform distribution of reacting gases and thus promoting the electrochemical performance. This implicates that a tradeoff of rib size to the cell performance is very significant. In this work, in order to optimize the interconnect structure, based on calculations of the thermal stress field, a 3D comprehensive bi-layer model was developed. Elaborating the reliable strategies for the optimization coupling electrochemical and mechanical properties is of crucial importance for new SOFC designs. Different cell configurations with a geometry design based on the finite element method were applied in this model. The influences from the shape and size of interconnect as well as the components for SOFC involved the thermal stress distribution were investigated in terms of ion/electronic, momentum, mass and heat transport. Mathematical model Conservation and constitution laws were applied to each domain to obtain a comprehensive model containing all the above mentioned phenomena. Stress analysis of an SOFC was performed using the commercial software COMSOL Multiphysics (version 5.0). A half-cell model with bipolar channels operating with humidified hydrogen, carbon monoxide and methane as fuel. The material structure parameters and geometry were selected based on a cell developed and tested at Ningbo Institute of Material Technology & Engineering (NIMTE) [34]. The geometry was built based on the interfacial zone or layer thickness in those experiments [35, 36] and the interfacial layers in the model were assumed to be 15 and 20 μm thickness for active anode and cathode layer and 0.1 m length for the cell, as shown in Fig. 1. Note that, the coflow case was used as a demonstration for our model geometry. The parameters of the geometry are presented in Table 1. Besides, two different interconnects designs are also shown in Fig. 1. The design 1 shows symmetrical interconnects with small contact area with the electrodes, while shape 2 has large contact area with electrodes.Fig. 1 (Color online) Interconnect (lilac color)-design of three cases (not to scale) with a standard case (just coflow shown here) and b shape 1 and 2, respectively Table 1 SOFC cell geometry Geometry parameter options Thickness Direction Cell length 100 mm x Gas channel height 0.5 mm y Gas channel width 1 mm z Interconnect rib thickness 0.5 mm y Interconnect thickness 150 μm y Anode support layer thickness 400 μm y Anode active layer thickness 15 μm y Cathode support layer thickness 50 μm y Cathode active layer thickness 20 μm y Electrolyte thickness 10 μm y Electrochemical model The reactions considered are shown in Eqs. (1)–(5). Carbon monoxide is oxidized in the electrochemical reaction (Eq. (3)), but reacts faster with water in the water–gas shift reaction (WGSR, Eq. (5)). Methane reacts with steam (MSR) in Eq. (4). 1 12O2+2e-→O2-, 2 H2+O2-→H2O+2e-, 3 CO+O2-→CO2+2e-, 4 CH4+H2O→3H2+CO, 5 CO+H2O→H2+CO2. Ion and electron transport For the ion and electron transport processes, the critical voltage distribution can be described as the potential difference between the anode and cathode current collectors [37]. The operating cell voltage (E) is reduced due to the internal resistance and polarizations, as shown in Eq. (6) 6 E=EOCV-ηact+ηohm+ηconc. Here, EOCV is the open-circuit voltage, ηact, ηohm, ηconc the activation, ohmic and concentration overpotentials respectively. 7 ηa=ϕs-ϕl-Eeq,a, 8 ηc=ϕs-ϕl-Eeq,c. Here, ϕ is the potential and Eeq the equilibrium voltage. The index “a” and “c” stand for the anode and cathode, respectively. The electrode materials are noted as “s” and the electrolyte material as “l”. When a hydrogen-steam mixture is used as fuel, the reversible open-circuit voltage, EOCV in Eq. (6) corresponds to the Gibb’s free enthalpy. ΔGf,x,T is the Gibb’s free enthalpy of the oxidation reaction of hydrogen [38]: 9 EOCV=-ΔGf,H2O,T02F+RT2F·lnpH2anodepO2cathodepH2Oanode, 10 ΔGf,H2O,T0=-247.4+0.0541T. The impact from carbon monoxide and hydrocarbons should be taken into account for a complete model. However, Eq. (10) is sufficient for the study performed in this work with focus on thermal stress. The activation polarization/current density relationship is described by the Bulter–Volmer equation: 11 i=AVi0PH2O,TPBPH2O,bexpαne,aFηaRT-PH2,TPBPH2,bexp1-αne,aFηaRT. Here, p is the partial pressure at the triple phase boundary (TPB) or bulk fluid within the gas channels (b) and ne is the number of electrons transferred per reaction. In the high current region with sufficiently large ηa where 12 expαne,aFηaRT≪exp1-αne,aFηaRT. Equation (11) can be rewritten as 13 i=-AVi0PH2,TPBPH2,bexp1-αne,aFηaRT. The activation polarization is calculated by using the exchange current density. For the electron and ion transfer problem in the anode and cathode, the charge conservation equation and the governing equations for the ion and electron transport by Ohm’s law are implemented as following: 14 is=∇·-σs∇ϕs, where σ is the ion/electron conductivity and ϕ is the potential, is is the total volumetric current density. In the anode, is is the volumetric local current densities of electrochemical reactions of fuel. The electronic conductivity in the electrodes (σs, within the anode and the cathode) and ionic conductivity in the electrolyte (σl) are calculated as described in Ref. [39]. The actual length that ions and electrons are transported in the electrodes is increased when the effects of the microscopic structures in the material are considered. This is accounted for by using the effective conductivities of each material which were determined by the structure-dependent tortuosities and the volume fractions/porosities [40]. As a medium for ion and transfer from the cathode to anode, the charge conservation in electrolyte can be written as 15 ∇il=Ql, where Ql is the source term, in the electrolyte which is equal to zero because there is no production or consumption of ions in electrolyte. When the heat source Qh for the ion/charge transport is taken into account, the energy conservation equation from the Joule effect can be expressed as 16 Qh=σl∇ϕl·∇ϕl+σs∇ϕs·∇ϕs. Momentum, mass and heat transport The temperature distributions depends on the flow of gases in the channels and porosity of the electrodes that are related to the momentum, mass and heat transfer processes. The momentum transport in the porous materials as well as in the fuel and air channel are solved simultaneously [41], as 17 μk+ρ∇·u→·u→-∇-p+1εψ+2μ3∇·u→=F. Here, F is the volume force vector, p the pressure, k the permeability of the porous material, u→ the velocity vector, ε the porosity and ψ the viscous stress tensor. The viscosity (μ) and density (ρ) for the gas mixtures are dependent on local temperature and the mole fractions and are calculated as described in Ref. [42]: 18 μi=∑k=17bkT1,000k, 19 μg=∑ixiμi, 20 ρi=p∑xiMiRT, where the bk is the species dependent parameter and ‘‘k’’ stands for the number of species dependent parameters in the viscosity equation. xi mole fraction of species, Mi the molecular weight of species i. The mass balance for a gas species, i, can be expressed as 21 ρu∇ωi+∇Ji=Ri, in which Ji is the mass diffusion flux and ωi is the mass fraction, Ri is the extra source term for production and consumption of species. Note that, Ri=0 in the gas channels. 22 Ri=viniF, v is the stoichiometric coefficient for species, i is the volumetric current density for the electrochemical reaction, and ni is the number of participating electrons in the electrochemical reaction. In this work, the Maxwell–Stefan model is used because of its simplicity and good accuracy: 23 Ji=-ρωi∑jDijdk+DiT, DiT is the thermal diffusion coefficient which is neglected due to its very small effect here. A concentration gradient in the porous electrodes is considered when the effects of mass transfer under normal operating conditions for the redox reaction are calculated. In the porous medium, the molecular diffusion and Knudsen diffusion are commonly used to describe the mass diffusion mechanisms. The molecular diffusion is used to describe the situation when the pore size is significantly larger than the mean free path of the gas molecule diffusion [43]. When both Knudsen and molecular diffusion are considered, the effective diffusion coefficients can be calculated as 24 Deff,ij=ετDij·Dk,ijDij+Dk,ij. Here, ε is the porosity and τ is the effective tortuosity factor. The effective diffusivity coefficients are calculated assuming electrodes with corrected factor ετ=0.16. Dk,ij is the Knudsen diffusion coefficient of the component i with the component j in a gas mixture, which is defined as [44] 25 Dk,ij=23re8·RTπ·Mij, where re is the effective pore radius. 26 Mij=21Mi+1Mj, where Mi,Mj are the molecular weight of species i, j, respectively. Heat generation within the cell from the electrochemical reactions (electron and ion transport as well as electrode reactions) and the internal steam reforming reaction within the anode are included as source term in the model [45]: 27 rMSR=AVMSR943exp-225×103R·TPCH4PH2O-7.74×10-9exp-1937R·TPCOPH23, 28 rWGSR=kWGSRPH2OPCO-PH2PCO2kWGSR. The temperature is assumed to be locally equal for the solid material and the gases, based on the minor temperature difference between the gas- and the solid-phase. The energy conservation equation can be written as 29 ρgcp,g·u→·∇T=∇·keff∇T+Qh, Here, Qh is the heat generation or consumption, keff is the effective thermal conductivity and cp,g is the gas-phase specific heat. The heat source term Qh, due to electrochemical reactions, ion and electron transport and losses through the activation can be defined as 30 Qh=iT·ΔSrne·F+ηact+∑i2σ+∑rrefΔHref, where ΔSr is the reaction entropy change, rref is the reforming reaction rate and ΔHref is the enthalpy change of the reforming reactions. Thermal stress couplings The materials used in this model are the ones most frequently used for SOFC. Table 2 shows material mechanical properties, which are obtained from literatures [39, 43, 46]. It is noted that the mechanical properties are determined by the content of materials as well as the synthesis process, which is the reason for the difference in properties listed in Table 2. The material properties of the interfacial layers are assumed to be between those of the electrode and the electrolyte respectively. All of the materials are assumed to behave as linear elastic and isotropic. It should be noted that the boundaries of this single cell model are assumed to be free, i.e., load effects are not considered here.Table 2 Material properties used for the calculation of the thermal stress model Layer Young’s modulus (GPa) Poisson’s ratio CTE (10−6 K−1) Anode support layer (Ni–YSZ) 220 0.3 12.5 Cathode support layer (LSM) 114 0.28 12.4 Electrolyte (YSZ) 205 0.3 10.3 Active anode layer 213 0.3 11.4 Active cathode layer 160 0.3 11.4 Interconnect (stainless steel) 205 0.28 12.3 CTE thermal expansion coefficient, YSZ yttria-stabilized zirconia, LSM strontium doped lanthanum manganite SOFC components expand with temperature, causing thermal strains to develop in the material when the deformation is constrained. The overall strain results from the summation of elastic and thermal stresses, i.e., the initial contributions are neglected: 31 ε=εel+εth The form of strain as follow 32 εel=εxx,εyy,εzz,γyz,γxz,γxy, here εxx, εyy, εzz, γyz, γxz, γxy are the longitudinal and shear components for strain, respectively. The thermal strain depends on the temperature, T, the stress-free reference temperature, Tref, and the coefficient of thermal expansion (CTE), α: 33 εth=αT-Tref. Determining the stress free temperature is critical as it directly affects the magnitude of the thermal stress induced in the material. For SOFCs it is widely accepted that Tref is the sintering temperature, at which different layers are joined. The stress–strain relationship for the linear material was calculated as 34 σ=Dεel+σ0, where σ0 is the initial stress, which is treated as the residual stress in model. The elasticity matrix (D) for isotopic material is defined as, 35 D=E1+v1-2v1-vvv000v1-vv000vv1-v0000001-2v20000001-2v20000001-2v2, where E is the Young’s modulus and v the Poisson’s ratio of the material. Boundary conditions and solution methods The gas inlet velocities are defined as laminar flow profiles, and the boundary condition at the impermeable walls is a nonslip condition for the velocity. At the outlets, the pressure (1.013 × 105 Pa) is fixed. The fuel inlet fractions are defined as 30 % pre-reformed natural gas (as defined by International Energy Agency (IEA), the mole fractions of the fuel are: H2: H2O: CH4:CO: CO2 = 0.263 : 0.493 : 0.171 : 0.0294 : 0.0436) were used as fuel in this model. Note that the ion and electron transport was considered as a testing part in our model, i.e., the H2 mixture and not the 30 % pre-reformed natural mixture was considered in this section, which is sufficient for the study on thermal stress. The air inlet is defined as air, including oxygen and nitrogen. The boundary conditions for the outlets are defined as convective fluxes. The inlet gas temperature is defined by the operating temperature (1,000 K) and the outlet is defined as a convective flux. The boundaries at the top and the bottom walls of the cell are defined as the symmetry conditions, because it is assumed that the cell is surrounded by other ones with identical temperature distribution. Also the boundaries at the side walls are defined as symmetry conditions, because the channel section is assumed to be surrounded by identical channel sections. The potential at the anode current collector is set to zero and the one at the cathode current collector as the cell operating voltage (0.7 V). All other boundaries and interfaces are electrically insulated. The governing equations are segregated in five different groups:Velocity field, pressure distribution, and pressure corrections. Temperature distribution. Ion and electron distribution. Mass fraction distribution on the air side and fuel side. Thermal stress distribution. The segregated solver is applied for 7,136,654 degrees of freedom and the solution tolerance is defined to 0.001 for each segregated group. The calculation time is around 80 h on a single computer with 32 GB RAM and a CPU with 4 GHz. Note that it is hard to give an exact value for the calculation time since the model is built in several steps, where each step starts its calculation from the previous one. Results and discussion Note that the color legend and the operating parameters are kept the same for all cases. The bi-layer SOFC model geometry for a half-cell (symmetry exists in the single cell) includes interconnects with the other components of a SOFC, as illustrated in Fig. 1a. The gas flow direction of coflow case is pointed out with an arrow, while the fuel flow is in the opposite direction for the counterflow case. The experimental data from Kyushu University are used for some adjustment of the simulation conditions as well as validation for the model results, as shown in our previous work [45]. For optimization of the shape, two types were designed to investigate the support and contact effects of interconnects, as shown in Fig. 1b. The rib in shape 1 has an inverted trapezoidal, with the same width of ribs contacting to the PEN structure (support positive layer-active positive layer-electrolyte- support negative layer-active negative layer) as the standard case, but width of ribs linked to the top interconnects increased by 0.1 μm. With the shape 1, the trapezoidal ribs that contact to the PEN structure is wider than the standard case (0.1 μm), as shown in shape 2. Those two types were designed, considering the possibility that the contact area and the mechanical strength of the structure would affect the thermal stress distribution. The temperature distribution for the counterflow case of standard shape is presented in Fig. 2. The temperature increase along the main flow direction is caused by the strongly exothermic electrochemical reactions as well as from the different polarizations. Note that the temperature decrease since the endothermic reforming reactions are considered. The maximum temperature (1,124 K) is located in the inside of the cell near to the position the fuel inlet (0.04 m to the fuel inlet).Fig. 2 (Color online) Temperature (K) distribution for the counterflow case of standard shape Coupling the mass and heat transport, reforming reaction, solid mechanical and the first principal stress of SOFC with different designs are shown in Fig. 3. The positive values tell that tensile stress and negative value for compressive stress. The active electrodes and electrolyte component involved in a tensile stress while the support electrode for a compressive stress. This stress difference results from the properties mismatch and it may lead to curling of the components. It is noted that maximum tensile and compressive stresses for coflow are smaller than that for counterflow, corresponding to the temperature distribution. The greater temperature gradient in counterflow also results in an obviously tensile stress for interconnect at fuel inlet, as seen in Fig. 3a. Similar stress distribution can be seen in shape 1 and 2, shown in Fig. 3c and d. The stress decreases with the new interconnect shape design, and the shape 2 is slightly smaller than shape 1, which may be due to the subtle difference in temperature distribution of heat conduction for the two interconnect structures.Fig. 3 (Color online) First principal stresses (MPa) of a counterflow of standard shape, b coflow of standard shape, c counterflow shape 1 and d counterflow shape 2 The effects of mismatch mechanical properties can be analyzed by comparing the stress distribution at each components’ interface. Since the mismatch between the interconnect and the cathode is larger than for the anode (Table 2), the thermal stress at cathode side is larger than at anode side, as shown in Fig. 4. It should be noted that, the horizontal direction represents the z axis in the model (direction vertical to gas flow) and the vertical direction represents the x axis (direction parallel to gas flow). The palliation effects of thermal stress of shape 1 and 2 design was obvious. The interconnect shape design is more effective to reduce the tensile stress in the PEN structure. However, the thermal stress distribution would be extended in the shape 2 design case, because the interconnect has a higher heat conduction compared to the gas, leading to a higher temperature on the surfaces of support layers.Fig. 4 (Color online) First principal stresses (MPa) of counterflow a standard case for interface of interconnect and anode, b standard case for interface of interconnect and cathode, c shape 1 for interface of interconnect and anode, d shape 1 for interface of interconnect and cathode, e shape 2 for interface of interconnect and anode and f shape 2 for interface of interconnect and cathode The thermal stress on the interface close to the inlet and outlet for the different designs is shown in Fig. 5. Since the temperature gradient at air inlet is smaller, a tiny change of stress can be seen in Fig. 5a, c, e. The slight increase of stress may be due to the slightly increased temperature. However, the tensile stress of the interconnect at the anode side was obviously reduced, and this decrease can also be seen at the active layers and in the electrolyte, compared to the fuel inlet in Fig. 5b, d, f. An increased compressive stress can also be observed within the support layers.Fig. 5 (Color online) First principal stresses (MPa) of counterflow a standard case air inlet, b standard case fuel inlet, c shape 1 air inlet, d shape 1 fuel inlet, e shape 2 air inlet and f shape 2 fuel inlet Figure 6 shows that the stress has a relationship with the location, i.e. for different components. For the interconnect at the cathode side, the stress at the air inlet is bigger than that at the fuel inlet, and the minimum stress locates at the middle position of the cell. However, the maximum stress belongs to the fuel inlet for the interconnect at the anode side, and stress at air inlet is close but slightly higher than that at the middle position of cell. When considering the thermal stress distribution for the PEN structure, the middle position dominates at the electrolyte while the fuel inlet and air inlet are comparatively large at the cathode and anode support layers, as depicted in the embedded figure.Fig. 6 (Color online) First principal stress (MPa) of counterflow case along with the direction vertical to gas flow of air inlet (a), middle part of cell (b) and fuel inlet (c) It is observed in Fig. 7 that the thermal stress is almost the same when the height of rib at anode side is increased, but causes a rise at the interconnect close to the cathode side, which is illustrated in the figure embed in Fig. 7. This may be explained by the symmetry stress effect of structure.Fig. 7 (Color online) First principal stress (MPa) of counterflow case along with the direction vertical to gas flow of air inlet. Standard case (a), interconnect of anode side heighten with 10 μm (b), 15 μm (c) and 20 μm (d) Figure 8 shows that the strong influence of the rib width on the thermal stress. Wider ribs and ribs covering bigger fraction of the cell may reduce the interface resistance to current flow by increasing the electrode–interconnect contact area and reducing the current path through the possibly high resistance electrode material. However, the gas channel would be narrowed, as a consequence, the electrochemistry reaction is affected. To illustrate, considering the three designs in Fig. 8, wider interconnect at the anode side would lower the stress. Conversely, this may be useless for reducing the stress at the cathode side.Fig. 8 (Color online) Counterflow case first principal stress (MPa) for electrolyte standard case (a), interconnect rib of anode side wider 0.1 mm (b), 0.2 mm (c), and interconnect rib of cathode side wider 0.1 mm (d) The thermal stresses of the coflow case for an electrolyte with a wider rib shown in Fig. 9. One can see that the stress obviously rise with wider interconnect ribs. The stress changes increases along the flow because of the rise of temperature. This difference may occur due to the superposition of interconnect and electrodes side (ribs) changes, which affect gas concentration at the three-phase boundary (TPB) region.Fig. 9 (Color online) First principal stresses (MPa) of coflow case for electrolyte standard case (a), interconnect rib anode wider 0.1 mm (b) and interconnect rib cathode wider 0.1 mm (c) Conclusions Interconnects are critical to minimize the overall stack resistance and weight as well as to enhance the stability when thermal stresses are considered. A comprehensive model to investigate the relationship between the cell structure, temperature distribution and thermal stresses in SOFCs at certain mechanical properties was built using the finite element method. Proper design of interconnect in conjunction with single-cells were implemented based on the thermal stress optimization performed in this work. The palliation of thermal stress of different designs was obvious. The interconnect shape design was the most efficient to reduce the tensile stress in PEN structure. Besides, thicker interconnect ribs causes a rise at the interconnect close to cathode side while this effect for thermal stress at the anode side is smaller. It is also observed that, wider interconnect of anode side lowers the stress. We show, by numerical calculations, that the thermal stress can be reduced through structural correlations. This optimization can be used as a new strategy for interconnect optimization in SOFCs which is beyond single criterion for electrochemical performance. Appendix: List of symbols AVActive area to volume ratio (m2/m3) cpSpecific heat at constant pressure (J/(kg K)) DDiffusion coefficient (m2/s) DiTThermal diffusion coefficient (kg/(m s)) EActual voltage (V) EYoung modulus (Pa) EaActivation energy (J/mol) E0Ideal voltage before partial pressure consideration (V) EeqEquilibrium voltage (V) EOCVIdeal voltage after partial pressure consideration (V) FVolume force vector (N/m3) FFaraday constant (96,485 C/mol) ΔHEnthalpy change of reaction (J/mol) iCurrent density (A/m2) i0Exchange current density (A/m2) JiMass diffusion flux kThermal conductivity (W/(m K)) MjMolecular weight of species j (kg/mol, g/mol) neNumber of electrons transferred per reaction pPressure (Pa or bar) QhSource term (heat) (W/m3) rReaction rate (mol/(m3 s)) rePore radius (m) RGas constant (8.314 J/(mol K)) RiExtra source term for production and consumption of species SiSource term (mass) (kg/(m3 s)) ΔSEntropy change of reaction (J/(K mol)) TTemperature (K) TrefStress-free reference temperature (K) u→Velocity vector (m/s) uDisplacement vector VVolume fraction wiMass fraction of species i (kg/kg) x, yCoordinate system (m) xjMole fraction of species j (mol/mol) Greek symbols αCoefficient of thermal expansion (K−1) βTransfer coefficient εPorosity εelElastic stresses contribution εthThermal strain contribution ηOver potential (or polarization) (V) ΦPotential (V) kPermeability (m2) μDynamic viscosity (Pa s) ρDensity (kg/m3) σl,sIonic/electronic conductivity (Ω−1 m−1) σStress vPoisson’s ratio τTortuosity ΔNumerical difference SPECIAL TOPIC: Modeling of Solid Oxide Fuel Cells This work was financially supported by the Research Fund for International Young Scientists (51550110238) and the Fundamental Research Funds for the Central Universities (ZYGX2015J108). Conflict of interest The authors declare that they have no conflict of interest. ==== Refs References 1. Atkinson A Barnett S Gorte RJ Advanced anodes for high-temperature fuel cells Nat Mater 2004 3 17 27 10.1038/nmat1040 14704781 2. Park S Vohs JM Gorte RJ Direct oxidation of hydrocarbons in a solid-oxide fuel cell Nature 2000 404 265 267 10.1038/35005040 10749204 3. Aljaberi AD Irvine JTS Ca-substituted, A-site deficient perovskite La0.2 Sr0.7 TiO3 as a potential anode material for SOFCs J Mater Chem A 2013 1 5868 10.1039/c3ta10844a 4. Reddy AA Goel A Tulyaganov DU Thermal and mechanical stability of lanthanide-containing glass–ceramic sealants for solid oxide fuel cells J Mater Chem A 2014 2 1834 1846 10.1039/C3TA13196C 5. Jensen SH Sun X Ebbesen SD Pressurized operation of a planar solid oxide cell stack Fuel Cells 2016 16 205 218 10.1002/fuce.201500180 6. Li TS Xu M Gao C Investigation into the effects of sulfur on syngas reforming inside a solid oxide fuel cell J Power Sources 2014 258 1 4 10.1016/j.jpowsour.2014.02.041 7. Xu M Li B Wang B Mechanism of phosphorus and chlorine passivating a nickel catalyst: a density functional theory study Electrochim Acta 2015 167 147 150 10.1016/j.electacta.2015.01.222 8. Chen Y Zhou W Ding D Advances in cathode materials for solid oxide fuel cells: complex oxides without Alkaline earth metal elements Adv Energy Mater 2015 5 1500537 10.1002/aenm.201500537 9. Chen Y Lin Y Zhang Y Low temperature solid oxide fuel cells with hierarchically porous cathode nano-network Nano Energy 2014 8 25 33 10.1016/j.nanoen.2014.05.010 10. Andersson M Yuan J Sundén B SOFC cell design optimization using the finite element method based CFD approach Fuel Cells 2014 14 177 188 10.1002/fuce.201300160 11. Chen D Zhang Q Lu L Multi scale and physics models for intermediate and low temperatures H+ -solid oxide fuel cells with H+ /e− /O2− mixed conducting properties: part A, generalized percolation theory for LSCF-SDC-BZCY 3-component cathodes J Power Sources 2016 303 305 316 10.1016/j.jpowsour.2015.10.090 12. Pecho O Stenzel O Iwanschitz B 3D microstructure effects in Ni–YSZ anodes: prediction of effective transport properties and optimization of redox stability Materials 2015 8 5554 5585 10.3390/ma8095265 13. Janardhanan VM Deutschmann O Modeling diffusion limitation in solid-oxide fuel cells Electrochim Acta 2011 56 9775 9782 10.1016/j.electacta.2011.08.038 14. Ni M Modeling and parametric simulations of solid oxide fuel cells with methane carbon dioxide reforming Energy Convers Manag 2013 70 116 129 10.1016/j.enconman.2013.02.008 15. Xie Y Xue X Multi-scale electrochemical reaction anode model for solid oxide fuel cells J Power Sources 2012 209 81 89 10.1016/j.jpowsour.2012.02.077 16. Greco F Frandsen HL Nakajo A Modelling the impact of creep on the probability of failure of a solid oxide fuel cell stack J Eur Ceram Soc 2014 34 2695 2704 10.1016/j.jeurceramsoc.2013.12.055 17. Laurencin J Delette G Lefebvre-Joud F A numerical tool to estimate SOFC mechanical degradation: case of the planar cell configuration J Eur Ceram Soc 2008 28 1857 1869 10.1016/j.jeurceramsoc.2007.12.025 18. Peksen M Numerical thermomechanical modelling of solid oxide fuel cells Prog Energy Combust Sci 2015 48 1 20 10.1016/j.pecs.2014.12.001 19. Boccaccini DN Sevecek O Frandsen HL Investigation of the bonding strength and bonding mechanisms of SOFCs interconnector–electrode interfaces Mater Lett 2016 162 250 253 10.1016/j.matlet.2015.07.137 20. Fleischhauer F Terner M Bermejo R Fracture toughness and strength distribution at room temperature of zirconia tapes used for electrolyte supported solid oxide fuel cells J Power Sources 2015 275 217 226 10.1016/j.jpowsour.2014.10.083 21. Chiang LK Liu HC Shiu YH Thermal stress and thermo-electrochemical analysis of a planar anode-supported solid oxide fuel cell: effects of anode porosity J Power Sources 2010 195 1895 1904 10.1016/j.jpowsour.2009.10.011 22. Fan P Li G Zeng Y Numerical study on thermal stresses of a planar solid oxide fuel cell Int J Therm Sci 2014 77 1 10 10.1016/j.ijthermalsci.2013.10.008 23. Jiang TL Chen MH Thermal-stress analyses of an operating planar solid oxide fuel cell with the bonded compliant seal design Int J Hydrog Energy 2009 34 8223 8234 10.1016/j.ijhydene.2009.07.089 24. Lin CK Huang LH Chiang LK Thermal stress analysis of planar solid oxide fuel cell stacks: effects of sealing design J Power Sources 2009 192 515 524 10.1016/j.jpowsour.2009.03.010 25. Clague R Marquis AJ Brandon NP Finite element and analytical stress analysis of a solid oxide fuel cell J Power Sources 2012 210 224 232 10.1016/j.jpowsour.2012.03.027 26. Mahato N Banerjee A Gupta A Progress in material selection for solid oxide fuel cell technology: a review Prog Mater Sci 2015 72 141 337 10.1016/j.pmatsci.2015.01.001 27. Wu J Liu X Recent development of SOFC metallic interconnect J Mater Sci Technol 2010 26 293 305 10.1016/S1005-0302(10)60049-7 28. Lin Z Stevenson JW Khaleel MA The effect of interconnect rib size on the fuel cell concentration polarization in planar SOFCs J Power Sources 2003 117 92 97 10.1016/S0378-7753(02)00726-7 29. Huang CM Shy SS Lee CH On flow uniformity in various interconnects and its influence to cell performance of planar SOFC J Power Sources 2008 183 205 213 10.1016/j.jpowsour.2008.04.059 30. Stygar M Brylewski T Rękas M Effects of changes in MOLB-type SOFC cell geometry on temperature distribution and heat transfer rate in interconnects Int J Heat Mass Transf 2012 55 4421 4426 10.1016/j.ijheatmasstransfer.2012.04.011 31. Grondin D Deseure J Zahid M Optimization of SOFC interconnect design using multiphysic computation Comput Aided Chem Eng 2008 25 841 846 10.1016/S1570-7946(08)80146-0 32. Noh H-S Hwang J Yoon K Optimization of current collection to reduce the lateral conduction loss of thin-film-processed cathodes J Power Sources 2013 230 109 114 10.1016/j.jpowsour.2012.12.059 33. Kong W Gao X Liu S Optimization of the interconnect ribs for a cathode-supported solid oxide fuel cell Energies 2014 7 295 313 10.3390/en7010295 34. Li TS Gao C Xu M Effects of PH3 and CH3 Cl contaminants on the performance of solid oxide fuel cells Fuel Cells 2014 14 999 1005 10.1002/fuce.201300248 35. Holtappels P Bagger C Fabrication and performance of advanced multi layer SOFC cathodes J Eur Ceram Soc 2002 22 41 48 10.1016/S0955-2219(01)00238-2 36. Li TS Xu C Chen T Chlorine contaminants poisoning of solid oxide fuel cells J Solid State Electrochem 2010 15 1077 1085 10.1007/s10008-010-1166-x 37. Zhu H Kee RJ Modeling distributed charge-transfer processes in SOFC membrane electrode assemblies J Electrochem Soc 2008 155 B715 10.1149/1.2913152 38. Holtappels P Stimming U Vielstich W Gasteiger HA Lamm A Solid oxide fuel cells (SOFC) Handbook of fuel cells 2010 New York Wiley 39. Andersson M Yuan J Sundén B SOFC modeling considering hydrogen and carbon monoxide as electrochemical reactants J Power Sources 2013 232 42 54 10.1016/j.jpowsour.2012.12.122 40. Sohn S Nam JH Jeon DH A micro/macroscale model for intermediate temperature solid oxide fuel cells with prescribed fully-developed axial velocity profiles in gas channels Int J Hydrog Energy 2010 35 11890 11907 10.1016/j.ijhydene.2010.08.063 41. Le Bars M Grae Worster M Interfacial conditions between a pure fluid and a porous medium: implications for binary alloy solidification J Fluid Mech 2006 550 149 173 10.1017/S0022112005007998 42. Andersson M Yuan J Sundn B SOFC modeling considering electrochemical reactions at the active three phase boundaries Int J Heat Mass Transf 2012 55 773 788 10.1016/j.ijheatmasstransfer.2011.10.032 43. COMSOL Multiphysics Version 5.0 User Guide. Stockholm, Sweden, 2014 44. Yuan J Huang Y Sundén B CFD approach to analyse transport phenomena coupled chemical reactions relevant for methane reformers Heat Mass Transf 2009 45 471 484 10.1007/s00231-008-0449-6 45. Andersson M Nakajima H Kitahara T Comparison of humidified hydrogen and partly pre-reformed natural gas as fuel for solid oxide fuel cells applying computational fluid dynamics Int J Heat Mass Transf 2014 77 1008 1022 10.1016/j.ijheatmasstransfer.2014.06.033 46. Liu L Kim G-Y Chandra A Modeling of thermal stresses and lifetime prediction of planar solid oxide fuel cell under thermal cycling conditions J Power Sources 2010 195 2310 2318 10.1016/j.jpowsour.2009.10.064 47. Xu M Andersson M Li TS Modeling of solid oxide fuel cell with anisotropic conductivity ECS Transact 2015 68 3003 3011 10.1149/06801.3003ecst

PMC005xxxxxx/PMC5002045.txt

==== Front Cell Mol Life SciCell. Mol. Life SciCellular and Molecular Life Sciences1420-682X1420-9071Springer International Publishing Cham 221610.1007/s00018-016-2216-zOriginal ArticleRole of tau in the spatial organization of axonal microtubules: keeping parallel microtubules evenly distributed despite macromolecular crowding Méphon-Gaspard Alix 1Boca Mirela 1Pioche-Durieu Catherine 2Desforges Bénédicte 1Burgo Andrea 1Hamon Loic 1Piétrement Olivier 2Pastré David 33 1 69 47 76 22david.pastre@univ-evry.fr 11 Institut National de la Santé et de la Recherche Médicale (INSERM), UMR1204, Université Evry-Val d’Essonne, Evry, 91025 France 2 UMR 8126, CNRS, Gustave Roussy Université Paris Sud, Université Paris-Saclay, Villejuif, 94805 France 13 4 2016 13 4 2016 2016 73 19 3745 3760 27 11 2015 24 3 2016 1 4 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Opposing views have been proposed regarding the role of tau, the principal microtubule-associated protein in axons. On the one hand, tau forms cross-bridges at the interface between microtubules and induces microtubule bundling in neurons. On the other hand, tau is also considered a polymer brush which efficiently separates microtubules. In mature axons, microtubules are indeed arranged in parallel arrays and are well separated from each other. To reconcile these views, we developed a mechanistic model based on in vitro and cellular approaches combined to analytical and numerical analyses. The results indicate that tau forms long-range cross-bridges between microtubules under macromolecular crowding conditions. Tau cross-bridges prevent the redistribution of tau away from the interface between microtubules, which would have occurred in the polymer brush model. Consequently, the short-range attractive force between microtubules induced by macromolecular crowding is avoided and thus microtubules remain well separated from each other. Interestingly, in this unified model, tau diffusion on microtubules enables to keep microtubules evenly distributed in axonal sections at low tau levels. Electronic supplementary material The online version of this article (doi:10.1007/s00018-016-2216-z) contains supplementary material, which is available to authorized users. Keywords AlzheimerTauopathyCytoskeletonNeuronhttp://dx.doi.org/10.13039/501100007149GenopoleInsermissue-copyright-statement© Springer International Publishing 2016 ==== Body Introduction Tau has been the subject of extensive studies by the past owing to its central role in many neurodegenerative diseases including Alzheimer’s disease [1–6]. However, despite the in-depth study of tau functions in neurons, both its regulatory role in axonal architecture and its contribution to neurodegeneration remain open questions. In the most popular model, tau stabilizes microtubules in axons [6–15]. During neuronal degeneration, tau is phosphorylated and possibly released from microtubules, which may then destabilize axonal microtubules [12–15]. Here, we examine whether tau is specifically designed to prevent the formation of axonal microtubule bundles. Indeed, axonal microtubules are well separated in transverse sections of mature neurons [16–20]. This alternative role for tau has already been proposed by others based on the polymer brush model [21–23] but has not truly emerged in the scientific community. Tau is rather known to form microtubule bundles via complementary dimerization in between microtubules [24–26], which is hard to reconcile with its role as microtubule spacer. In addition, a continuous repulsive layer of tau on microtubules is required to keep microtubules separated in the polymer brush model [21–23, 27], which is most probably not relevant to the conditions found in axons. This point is even more critical since tau diffuses on microtubules [28] and thus has the possibility to move away from the interface between microtubules. Besides these diverging views, studying whether tau keeps microtubules well separated in axons makes sense for two reasons: (1) their cylindrical geometry and rigid structure make microtubules prone to collapse due to short-range (<5 nm) excluded volume interactions in the crowded cellular environment [29, 30]; (2) the organization of axonal microtubules into parallel arrays significantly increases the probability of microtubule bundling in axons. From a functional point of view, the surface of isolated microtubules is more accessible than in bundles. Keeping microtubules separated should then favor the accessibility of molecular motors to microtubule surface and thus the long-range transport of cargoes [20]. Here we propose and examine an alternative mechanistic model which reconciles tau-mediated microtubule bundling observed in vitro and the proposed physiological function of tau as a microtubule spacer. Indeed we describe how the formation of transient tau cross-bridges at the interface between parallel microtubules could lead to a very efficient separation of microtubules in axons. In addition, in this model, tau diffusion on microtubules enables an efficient separation of microtubules even at low tau levels, more in line with those found in axons. Results Analysis of the state of art and clarifications Four points need to be clarified from the rich literature on tau and provide a basis for the present study:In the literature, the fact that tau leads to the formation of microtubule bundles has been reported many times. Tau is then considered to be a positive regulator of microtubule bundling. This view stems mostly from the formation of microtubule bundles in neuronal and non-neuronal mammalian cells over-expressing tau [25, 26, 31–33]. As the spacing between microtubules in these bundles matches the length of the N-terminal domain of tau [31], this was considered as an evidence that tau can induce microtubule bundling. In line with this, the N-terminal domain of tau contains alternating clusters of positive and negative residues and can act as an antiparallel electrostatic zipper [24]. However, in electron micrographs of mature axon sections, microtubules appear as homogeneously distributed (for examples see [16–20]) and have a larger separation distance than in tau-mediated bundles (about 20 nm [32]). Along with this, the density of microtubules estimated from axonal sections in electron micrographs most frequently ranges from about 15 to 160 microtubules per µm2 [16, 17, 34–37]. The mean spacing between microtubules thus ranges from 79 to 260 nm. These facts indicate that tau-mediated bundling is not prevalent in mature axons. The presence of axonal microtubule bundles can nevertheless be observed at early stage during axon differentiation [34] and in the axon hillock [38] with possible neuronal functions but this is not the norm. In contrast, tau may act as a polymer brush and keep microtubules separated from each other. Based on thermal movements, the unstructured N-terminal domain of tau leads to a repulsion of entropic origin between microtubules [21, 23]. The so called “polymer brush” model works as long as tau forms a continuous repulsive layer on the microtubule surface. In addition, as recently measured [28, 39], tau is not immobile on microtubules but rather diffuses along the microtubule surface (D = 0.15 µm2/s). The consequences of tau diffusion on the putative entropic repulsion occurring in the polymer brush model need to be clarified. Indeed diffusing tau can be redistributed away from the interface between microtubules. In this case, tau could no longer exert a repulsion force to keep microtubules separated. In the absence of any additional factors, only an elevated tau:tubulin molar ratio and its associated steric hindrance can prevent tau from moving away from the interface between microtubules [40], which leaves the relevancy of this model in axons questionable. The tau:tubulin ratio in axons is actually poorly characterized. This is highly surprising since this information is critical to decipher the functions of tau in axons. In the literature and despite numerous publications on tau:microtubule interactions, only limited sources of data have been used to provide an estimation of the tau:tubulin molar ratio. In microtubules assembled from brain extracts [41], which may not be representative of axons, the reported tau:tubulin molar ratio was about 1:12 and 1:38 in gray and white matters, respectively. In another report [42], the tau:tubulin measured from undifferentiated to differentiated PC12 cell extracts ranges from about 1:34 to 1:5 and from 1:68 to 1:17. Tau concentration was also estimated using radioimmuno-slot-blot assay [43]. About 0.9 ng of tau per µg of proteins was found in postmortem human brain homogenates of grey matter. If we assume that there is about 30 ng of tubulin per µg of proteins in brain extract [44], the tau:tubulin molar ratio in neurons could be lower than 1:30 as grey matter contains more tubulin than white matter. In summary, all these data, although useful, do not provide a precise estimation of the specific tau:tubulin ratio in axons but rather a global estimation for neurons or brains. We therefore need to consider this point in the present study. Microtubules are highly prone to form bundles under macromolecular crowding conditions. In vitro, 0.3–1 % (w/v) of PEG 35K (polyethylene glycol of 35 kDa), a neutral crowding agent, is sufficient to trigger microtubule bundling while more than about 10 % of PEG 35K are required for compacting DNA under same ionic conditions [29]. Microtubules are rigid and large cylinders with diameters about 25 nm. They thus offer a large surface for excluded-volume interactions. In vitro, microtubule bundles obtained via excluded-volume interactions under macromolecular crowding conditions are tightly packed with wall-to-wall contacts, in contrast with the regularly spaced microtubules observed in tau-mediated bundles [30]. Microtubule-based transport should most probably be impaired in such compacted structures [45], which would be detrimental for most axonal functions. The role of Tau 2N on microtubule bundling is biphasic To reassess tau-mediated bundling in vitro and the impact of the tau:tubulin molar ratio in microtubule bundling, microtubule bundling was monitored by turbidimetry measurements at 37 °C [46] in the presence or absence of tau 2N, the longest tau isoform with four microtubule-binding domains (Figs. 1a and S1). The results indicate that tau 2N triggers microtubule bundling in both taxol-stabilized and non-stabilized microtubules at moderate ionic strength (Fig. 1b, c). The onset of microtubule bundling corresponds to a tau:tubulin molar ratio of about 1:30 and microtubule bundling increases steadily at higher ratios. These results were further confirmed by optical microscopy and atomic force microscopy analyzes. They revealed the appearance of aligned microtubules and loose microtubule bundles in the presence of tau 2N (Figs. 1d, e and S2). According to the literature, tau cross-bridging at the interface between two parallel microtubules is due to the presence of alternating positive and negative charges located in the N-terminal domain of tau [24]. An electrostatic zipper mechanism is therefore sensitive to the ionic strength or to the presence of perturbing zwitterions at high concentrations. In line with this, the presence of PIPES or MES at elevated concentrations (50 mM) or monovalent salts (>75 mM) impairs microtubule bundling mediated by tau 2N (figure S3), which may explain why microtubule bundling has not been detected in some reports [47, 48], in contrast to others [24, 49, 50]. We also cannot exclude that increasing the ionic strength reduces the affinity of tau for microtubules and, in turn, also reduces the probability to form tau cross-bridges. Indeed both the formation of tau cross-bridges and the binding of tau to microtubules are partly based on electrostatic interactions.Fig. 1 Tau leads to microtubule bundling in vitro via its N-terminal domain. a Schematic representation of the tau constructs used in this study. b Turbidimetry curves recorded in the presence of tubulin and tau 2N at indicated tau:tubulin molar ratios after raising the temperature from 5 to 37 °C. The increase of plateau value of the assembly curves in the presence of tau 2N could be the result of three putative and independent contributions: (1) increase of final microtubule mass, (2) tau aggregation, (3) tau-induced microtubule bundling [46]. We controlled that tau alone does not increase the absorbance so that the contribution of tau aggregation can be excluded. In addition, owing to the magnitude of the increase in the plateau value observed for tau:tubulin molar ratios of 1:5 and 1:3, microtubule stabilization cannot solely account for this phenomenon. Polymerization buffer: 10 mM HEPES–KOH pH 6.8, 30 mM KCl, 20 % glycerol, 1 mM EGTA, 4 mM MgCl2, 1 mM GTP. 20 µM tubulin were used for microtubule polymerization. c Microtubule bundling observed via turbidimetry measurements of taxol-stabilized microtubules after the addition of tau 2N, tau 0N and deltaNT. In the presence of taxol, only microtubule bundling contributes to the increase of absorbance. Tau 2N gradually induces microtubule bundling. DeltaNT does not induce microtubule bundling. Same buffer as b with 10 µM tubulin and 5 µM taxol. d Statistical analysis of microtubule bundling obtained from optical microscope images of taxol-stabilized microtubules deposited on mica in the presence or absence of tau 2N or deltaNT (see figures S2 for details). The increase of the normalized fluorescence intensity reveals that tau 2N induces a massive microtubule bundling at tau:tubulin molar ratio higher than 1:15. Results are mean ± SD (n = 60). Two-tailed t test, **p < 0.01. e High resolution imaging by atomic force microscopy shows the formation of loose microtubule bundles in the presence of tau 2N at a tau:tubulin molar ratio of 1:15. Such pattern was not observed with deltaNT. At high tau:tubulin molar ratio (1:3), tau2N leads to the formation of large bundles. Same conditions as c with taxol-stabilized microtubules. Scanned area: 5 × 5 µm2. Lower panels represent higher magnification images of the area corresponding to the dashed squares In order to examine the putative role of tau in microtubule bundling under macromolecular crowding conditions, we used PEG 35K whose size is comparable to proteins in the cell cytoplasm (about 4–5 nm). 1 % PEG 35K is sufficient to form microtubule bundles (Fig. 2a, c). This pattern was not observed with PEG 1K because its size is not sufficient to induce significant excluded volume interactions [29]. We then analyzed whether tau 2N positively or negatively regulates the packing of microtubules under macromolecular crowding conditions. The formation of microtubule bundles was monitored by turbidimetry measurements in the presence of tau 2N after the addition of PEG 35K. The results reveal that tau 2N antagonizes the collapse of microtubules into tightly packed bundles. Importantly, at low tau-tubulin molar ratios, 1:30 and to a lesser extent at 1:60, tau 2N still antagonizes microtubule bundling (Fig. 2a). Analyzes by optical, atomic force and electron microscopies further confirm that tau 2N prevents the formation of tightly packed microtubule bundles under macromolecular crowding conditions (Fig. 2c–e).Fig. 2 Tau antagonizes microtubule bundling under macromolecular crowding conditions. a Variations of absorbance recorded after the addition of PEG 35K 1 % to taxol-stabilized microtubules in the absence or presence of tau 2N at indicated concentrations. The presence of tau 2N prior to adding PEG 35K blocks microtubule bundling and so even at low tau:tubulin molar ratio (1:30). Buffer: 10 mM HEPES–KOH pH 6.8, 30 mM KCl, 20 % glycerol, 1 mM EGTA, 4 mM MgCl2, 1 mM GTP, 10 µM tubulin and 5 µM taxol. b Same experiments as a in the presence of tau 0N or deltaNT. DeltaNT fails to prevent microtubule bundling by PEG 35K. Tau 0N prevents microtubule bundling but to a lesser extent than tau 2N. c Atomic force microscopy images of taxol-stabilized microtubules after the addition of 1 % PEG 35K for 15 min. In control, PEG 35K leads to the appearance of tight microtubule bundles. In contrast with deltaNT, tau 2N inhibits the formation of tight microtubule bundles in the presence of PEG 35K 1 % for 15 min. Scanned area: 5 × 5 µm2. d Statistical analyzes of microtubule bundling from optical microcopy images of taxol-stabilized microtubules deposited on mica in the presence of tau 2N. The addition of PEG 35K 1 % leads to the formation of microtubule bundles in control. The presence of tau 2N inhibits the massive microtubule bundling induced by the addition of PEG 35K. Results are mean ± SD (n = 60). Two-tailed t test, **p < 0.01. e Electron micrographs reveal that microtubules exposed to 1 % PEG 35K for 15 min form tightly packed bundles. In the presence of tau 2N, microtubules failed to form tightly packed bundles after the addition of PEG 35K. In contrast, the presence of deltaNT does not prevent the formation of tightly packed bundles under macromolecular crowding conditions. Scale bars 100 nm. Tau:tubulin molar ratio: 1:8 for both tau 2N and deltaNT. Statistical measurements of inter-microtubule distances (center to center) inside bundles under indicated conditions in the presence of 1 % PEG 35K. The separation distance is larger in the presence of tau 2N. Interestingly, the packing of microtubule with deltaNT is tighter than in control conditions. We attributed this fact to the neutralization of the negatively charged C-terminal tail of tubulin by the positive residues of deltaNT, which reduces the electrostatic repulsion of microtubules. Results are mean ± SD, (n > 95). Two-tailed t test, **p < 0.01 The role of tau 2N in microtubule bundling could thus be misleading. On the one hand, in the absence of macromolecular crowding, tau 2N induces the formation of bundles at elevated tau:tubulin molar ratios. On the other hand, under macromolecular crowding conditions, tau 2N antagonizes microtubule bundling and so even at low tau 2N:tubulin molar ratios. Interestingly, at elevated ionic strength, tau 2N can no longer prevent microtubule bundling in the presence of PEG 35K (Figure S4A). As high ionic strengths also prevent microtubule bundling mediated by tau 2N in the absence of PEG 35K (Figure S3), there could be an unexpected correlation between the formation of tau cross-bridges between microtubules and the ability of tau to prevent microtubules from collapsing into tight bundles under macromolecular conditions. The N-terminal domain is critical for preventing microtubule bundling under macromolecular conditions at low tau:tubulin molar ratio in vitro To explore the role of the N-terminal domain of tau 2N on microtubule bundling, we considered three tau constructs (see Fig. 1a) of different N-terminal length: tau 2N, the longest tau 4R isoform used as a control, tau 0N, the smallest isoform of tau 4R, and deltaNT, tau 4R deleted from its N-terminal domain except part of the proline rich domain which is critical for the binding of tau to microtubules [51]. In the absence of macromolecular crowding, deltaNT shows no detectable microtubule bundling activity while tau 0N could form microtubule bundles but to a lesser extent than tau 2N (Fig. 1b). DeltaNT also fails to induce the formation of loose microtubule bundles as observed with tau 2N (Fig. 1e). The propensity of the tau constructs to form microtubule bundles thus correlates positively with the length of the N-terminal domain, which emphasizes the critical role played by the N-terminal domain of tau in microtubule bundling in vitro. Importantly, under macromolecular crowding conditions, deltaNT cannot prevent microtubule bundling (Fig. 2b, c, e) and tau 0N limits the formation of microtubule bundles under macromolecular crowding conditions to a lesser extent than 2N (Fig. 2b). The formation of tau cross-bridges via the N-terminal domain at the interface between microtubules is thus possibly critical to prevent microtubules from collapsing into bundles under macromolecular conditions, as observed by electron microscopy (Fig. 2e). We have also examined whether tau 2N can dissociate microtubule bundles when they are preformed under macromolecular conditions. As shown in figure S4B, microtubules cannot be released from preformed bundles in the presence of tau 2N. Microtubule bundling under macromolecular crowding can thus be considered irreversible regarding to the role of tau as microtubule spacer. In order to keep microtubules well separated from each other, tau 2N should therefore be located at the interface between microtubules before bundling takes place. The N-terminal domain of tau antagonizes microtubule bundling in non-neuronal mammalian cells at low tau:tubulin fluorescence ratio In contradiction with the experimental results obtained under macromolecular crowding conditions in vitro, tau overexpression in non-neuronal mammalian cells leads to the appearance of microtubule bundles [25, 26, 31, 32] (Fig. 3a). This fact has established tau as a bundling factor. However other facts have nuanced this view: (1) Experimental results showed that the N-terminal domain was not required for microtubule bundling ([7], Fig. 3a). This is surprising as this domain is responsible for microtubule bundling in vitro; (2) Microtubule stabilization by taxol is sufficient to induce the formation of microtubule bundles [52], which indicates that Tau, solely via its microtubule-stabilizing activity, could induce microtubule bundling [7]; (3) When overexpressed in cells, many microtubule partners lead to microtubule bundling while it is not their primary functions. EB-1 is a protein which recognizes specifically the growing plus-ends of microtubules. When artificially overexpressed in cells, EB-1 stabilizes microtubules and leads to the formation of thick bundles [53]. Spastin, a microtubule-severing enzyme, is another example of a protein [54] that forms microtubule bundles when overexpressed and prevented from cutting microtubules in cells.Fig. 3 The N-terminal domain of tau antagonizes microtubule bundling in non-neuronal cells at low tau:tubulin fluorescence ratio. a Fluorescence microscopy of HeLa cells expressing Tau 2N, Tau 0N and deltaNT, as indicated in the figure. Anti-tubulin and anti-Tau immunofluorescences are represented in red and green respectively. The anti-tau antibody is directed against the C-terminal domain of tau and recognizes all the tau constructs used in this study. Scale bar 30 µm. b Statistical analyses of the formation of microtubule bundles in HeLa cells vs tau:tubulin fluorescence ratio for the three tau constructs. For both cells displaying microtubule bundles or not (see supplementary figure S7), the mean fluorescence ratios and standard deviations were determined. For all tau constructs, the ratios of tau:tubulin fluorescence were significantly different in cells with and without microtubule bundles as indicated in the figure. **p < 0.01; two-tailed t test. Tau 2N induces microtubule bundling at a significantly higher tau:tubulin fluorescence ratio than deltaNT and to a lesser extent Tau 0N. The transition between bundling and no bundling occurs at a critical tau:tubulin fluorescence ratio indicated in the figures for each tau constructs. The critical fluorescence ratio and interval boundaries were estimated as the mean and difference between the ratios leading to 25 and 75 % of cells having microtubule bundles, respectively. The transition zone between bundling and no bundling is represented in blue in the figures. c Fluorescence microscopy of HeLa cells expressing either tau 2N or deltaNT and treated with 100 nM taxol for 8 h. Microtubule bundling appears as less marked in cells expressing tau 2N than in cells expressing deltaNT. Scale bar 30 µm. d Same as b for taxol-treated cells. Again, a higher tau:tubulin fluorescence ratio is required to induce microtubule bundling in cells expressing tau 2N than deltaNT To better understand the role of tau on the spatial organization of microtubules in a cellular context, we analyzed the formation of microtubule bundles versus the tau:tubulin fluorescence ratio on microtubule structures in HeLa cells expressing tau 2N, tau 0N and deltaNT (see figure S5). The tau:tubulin fluorescence ratio should reflect the tau:tubulin molar ratio on microtubules. This parameter is therefore more informative than the percentage of cells displaying microtubule bundles [7]. At low tau:tubulin fluorescence ratios, neither of the tau constructs induces bundling (Figs. 3a, b and S5). In contrast, at high tau:tubulin fluorescence ratios, all the tau constructs induce bundling. The difference between the three tau constructs lies in the critical tau:tubulin fluorescence ratio required to induce microtubule bundling (Fig. 3b). Microtubule bundling appears at a significantly higher expression level for the longest tau isoform, tau 2N, than for deltaNT and, to a lesser extent, for the shortest isoform of tau, tau 0N. This result indicates that the N-terminal part of tau counteracts microtubule bundling as observed in vitro under macromolecular crowding conditions. To exclude the influence of varying microtubule stability, the same experiments were repeated in HeLa cells treated with taxol to stabilize microtubules and the results again indicate a negative regulation of microtubule bundling by the N-terminal domain of tau at low tau:tubulin fluorescence ratios (Fig. 3c, d). We also considered microtubule bundling after osmotic shock in taxol-treated HeLa cells. Osmotic stress increases intracellular macromolecular crowding and leads to the formation of microtubule bundles, as previously reported [55]. In line with this, NaCl treatment indeed leads to the formation of microtubule bundles in control HeLa cells (figure S6A and B). Interestingly, the percentage of cells displaying microtubule bundles is not increasing significantly between cells expressing tau 2N under control and hypertonic conditions (figure S6A and B). Again, in cells displaying a low tau:tubulin fluorescence ratio, the presence of microtubule bundles after osmotic stress is less marked than at high tau:tubulin fluorescence ratio (figure S6C). However cells expressing deltaNT formed microtubule bundles whatever they were exposed to osmotic stress or not. Altogether these results indicate that the N-terminal domain of tau 2N, at low tau:tubulin ratio, antagonizes microtubule bundling in a cellular context. An estimation of the tau:tubulin molar ratio in axons of mouse cortical neurons reveals that tau may not trigger but rather prevent the formation of microtubule bundles Tau promotes microtubule bundling in HeLa cells at elevated tau:tubulin fluorescence ratios while the N-terminal domain responsible for tau-mediated cross-bridges clearly antagonizes their formation at lower ratios. We then wondered which of these two regimes prevails in axons of mature primary neurons. To explore this point, we co-cultured primary mouse cortical neurons and HeLa cells expressing tau 2N and measured the tau:tubulin fluorescence ratio in both axons and transfected HeLa cells (Fig. 4a). HeLa cells thus serve as internal controls to gauge the critical tau:tubulin fluorescence ratio for microtubule bundling. We noticed that the tau:tubulin fluorescence ratio was significantly lower in axons than in HeLa cells displaying microtubule bundles (Fig. 4a–c). To make a parallel between in vitro results with those obtained in cells, we recorded a calibration curve of the measured tau:tubulin fluorescence ratio on microtubules deposited on mica versus tau:tubulin molar ratio (Figs. 4d and S2). Using this calibration curve, the estimated value of the mean tau:tubulin molar ratio along axons is about 1:45, which is in line with the values previously reported. By comparison, in order to trigger microtubule bundling in HeLa cells, the tau:tubulin molar ratio should be larger than 1:8 for tau 2N. The tau:tubulin molar ratio in axons of cultured cortical neurons is thus probably not sufficient to induce microtubule bundling. These results are in line with an inhibition of microtubule bundling orchestrated by the N-terminal domain of tau. However, precautions should be taken before generalizing these results. The tau:tubulin of cortical neurons in culture for 7 days may not be representative of all neurons. In addition, the tau:tubulin ratio is not stable along axons and may increase in the apical region [56]. We thus cannot exclude that tau can initiate microtubule bundling in specific locations in axons.Fig. 4 In axons, the tau:tubulin molar ratio is not sufficient to trigger microtubule bundling and should rather prevent microtubules from bundling. a Left panel primary mouse cortical neurons were cultured for 7 days and then co-cultured with HeLa cells expressing Tau 2N for 8 h. Fluorescence microscopy reveals the expression levels of Tau in neurons and some Hela cells. The anti-tau and anti-tubulin antibody fluorescence is represented in red and green, respectively. The anti-tau antibody recognizes an epitope in the C-terminus of tau which is conserved in both mouse and human. Scale bar 30 µm. Right panel fluorescence intensities of anti-tau and anti-tubulin antibodies along the yellow line represented in the bottom image of the left panel. The positions of the cells with or without microtubule bundles and the axon are indicated in the figure. b Images of HeLa cells and a representative axon sorted by their respective tau:tubulin fluorescence ratio. In axons, the tau:tubulin fluorescence ratio appears far lower than in HeLa cells having microtubule bundling. c Statistical analyzes of the tau:tubulin fluorescence ratio measured in axons (green triangles) and HeLa cells (circles and squares) as described in Fig. 3. The axonal tau:tubulin fluorescence ratio is indeed significantly lower than in HeLa cells displaying microtubule bundles (we controlled that the presence of HeLa cells does not change the tau:tubulin fluorescence ratio measured in cultured neurons). d Calibration curve of the ratio of tau:tubulin fluorescence vs tau:tubulin molar ratio, which was measured from microtubules interacting with recombinant Tau 2N in vitro. Each dot represents the average over ten different measurements. The putative tau:tubulin molar ratio in axons was estimated by using the calibration curve and the mean tau:tubulin fluorescence ratio found in axons (0.17). The critical tau:tubulin molar ratio required to form microtubule bundles in HeLa cells was obtained by using the critical tau:tubulin fluorescence ratio for microtubule bundling (1.15 for tau 2N). The calibration curve was fitted with a second order polynome using a least square fitting Both numerical and analytical analyses predict that tau could be an efficient microtubule spacer owing to tau diffusion on microtubules In the polymer brush model [57], tau is an unstructured polymer which coats the surface of particles to prevent their aggregation. For an efficient steric hindrance in the case of microtubules, scaling laws indicate that about 1 tau molecule for 6 tubulin dimers is required [40] (see Eq. 6 in supplementary data 1). The microtubule surface should then be half-saturated with tau (saturation ratio: 1:3), which is much too high to be relevant under physiological conditions. Along with this, tau diffuses on microtubules. If there is no force to keep tau at the interface between approaching microtubules, tau has the possibility to move away from the interface of microtubules via thermal diffusion (Fig. 5a). Consequently, tau cannot efficiently block microtubule bundling at low tau:tubulin ratios in the polymer brush model.Fig. 5 Numerical simulations of the mechanism by which tau acts as a microtubule spacer. a Schematic representation of the polymer brush model and the alternative model proposed based on tau cross-bridges and tau diffusion. (1) In the polymer brush model, tau can move away from the interface between microtubules and thus cannot prevent microtubule bundling under macromolecular crowding conditions, unless sterical hindrance prevents tau movements at elevated tau:tubulin molar ratios. (2) Cross-bridging of tau at the interface between microtubule provides an energy benefit to place tau at the interface between microtubule. Tau cross-bridges then keep microtubules at distance and prevent short-range attraction between microtubules under macromolecular crowding conditions. b Numerical simulations of the spatial distribution of 10 microtubules moving on a 300 × 300 nm2 area. The number of tau molecules per microtubule, N, and the relative tau mobility, DtauDMT, are as indicated in the figure. We remarked that both increasing the number of tau proteins per microtubule and the tau diffusion decreases the occurrence of bundling. r p = 5 nm (range of excluded volume interactions); r c = 15 nm (range of tau cross-bridges). Number of iterations: 105. L the length of the microtubule, is 500 nm. See supplementary data 1 for details about the model. Here are represented the transverse views of the microtubule array. Tau molecules having different positions along microtubules are then projected on the section view. D tau,/D MT values ranging from 2 to 20 were used for the numerical simulations. This choice is based on the theoretical diffusion constant of a cylindrical molecule (Eq. 1, supplemental data 1), the estimated value of the cytoplasm viscosity and the varying length of microtubules (see supplemental data 1 for details). However, as D MT in axons has not been measured, these values may be considered as arbitrary. c Number of bundled microtubules after completion of the numerical simulations vs the number of tau proteins per microtubule and the relative tau mobility along microtubules (D tau/D MT). Same conditions as b. d Numerical simulations indicating the number of bundled microtubules among 20 microtubules moving on a 350 × 350 nm2 area vs the range of tau cross-bridges, r c. Number of iterations: 105. D tau/D MT = 3. The other parameters are the same as B) In the cross-bridge model, tau dimerization via the N-terminal domain can provide an energy benefit to keep tau at the interface between microtubules and prevent tau from bundling under macromolecular crowding conditions (Fig. 5a). In addition, tau cross-bridges will oppose the further approach of two interacting microtubules. The compression of tau cross-bridges at the interface between approaching microtubules indeed generates a repulsive force [58] and, accordingly, tau cross-bridges have been considered as strings [59]. To perform numerical simulations, we considered that tau forms antiparallel cross-bridges at the interface between microtubules and assumed that tau dimerization can be disrupted when interacting microtubules move away from each other. The latter assumption seems justified as a massive microtubule bundling was not observed at low tau:tubulin ratios (Fig. 2). Based on these hypotheses, the results of numerical simulations indicate that tau 2N is an efficient spacer for microtubules (Fig. 5b, Videos 1–4). Interestingly, tau diffusion reduces the critical tau:tubulin ratio required to prevent microtubule bundling. As the microtubule surface area scanned by tau increases with tau mobility, less tau is required to keep microtubule separated (Fig. 5b, c). To emphasize this point, an analytical approach shows that the number of diffusing tau proteins required to prevent the formation of microtubule bundles decreases linearly with the relative mobility of tau, Dtau,/DMT where Dtau, DMT are the diffusion constants of tau and microtubules, respectively (see supplementary text, Eq. 5). Interestingly, increasing the length of the N-terminal domain significantly reduces the number of tau molecules required for keeping microtubule separated (Fig. 5d and Eq. 5 in supplementary data 1). Discussion While microtubules form bundles in non-neuronal cells overexpressing tau, microtubules appear as rather homogeneously distributed in axonal sections of mature neurons [16, 17, 19]. The spatial separation of microtubules most probably favors long-range transport of vesicles, mitochondria and RNA along axons. To explain the spatial separation of axonal microtubules, tau was then considered as a microtubule spacer [21, 22]. In the polymer brush model, the separation between microtubules is due to the unstructured N-terminal domain of tau which acts as a repulsive layer. While such a mechanism deserves to be considered, its application to tau and axonal microtubules has to be carefully analyzed. The point is that near-saturating concentrations of tau are required to form a continuous repulsive layer on microtubule surface. In axons, the tau:tubulin molar ratio ranges between 1:12 and 1:68 [41–43] and, here, an estimation based on calibrated immunofluorescence in axons of cortical neurons leads to a tau:tubulin molar ratio of about 1:45. The tau:tubulin molar ratios found in axons are therefore not sufficient to form a continuous layer on microtubule surface. Importantly, tau diffuses on the microtubule lattice. Tau has thus the possibility to move away from the interface between microtubules. In the absence of additional factors, microtubules should then form bundles under macromolecule crowding conditions (Figs. 5a, 6a). Fig. 6 Spatial organization of microtubules in the presence of tau under macromolecular crowding conditions. a Schematic curves representing forces vs separation distance between microtubules. 1 Macromolecular crowding induces a strong short-ranged attraction (r < 5 nm, average size of proteins) and the formation of tightly packed microtubule bundles. 2 The formation of tau cross-bridges between microtubules generates a long-ranged attraction (about 20 nm). However the compression of tau cross-bridges generates a repulsion force at shorter distances. 3 By combining the forces due to macromolecular crowding and tau cross-bridges, we notice the appearance of an energy barrier which could prevent microtubules from collapsing into tight bundles in axons. b Spatial organization of microtubules at various tau:tubulin molar ratios. 1 In the absence of tau or at very low tau:tubulin molar ratios (<1:60 for tau 2N, in vitro), macromolecular crowding induces the collapse of microtubules into tight bundles. 2 At moderate tau:tubulin molar ratios (from about 1:60 to 1:15, in vitro), transient tau cross-bridges keep microtubules separated and prevent the formation of tightly packed microtubule bundles. No massive bundling is observed. We propose that this regime prevails in axons. 3 At elevated tau:tubulin ratios (>1:15), the formation of many tau cross-bridges induces the assembly of microtubule bundles in which the spacing between microtubules is about 15–20 nm. This regime prevails when tau is strongly over-expressed in HeLa cells To understand the mechanisms responsible for separating axonal microtubules, we considered that tau forms cross-bridges at the interface between microtubules. The energy benefit of forming cross-bridges allows the specific location of tau at the microtubule interface and thus prevents tau from moving away (Fig. 1a). At low tau:tubulin molar ratios, the formation of only few cross-bridges is not sufficient to trigger the formation of microtubule bundles by itself. However, when microtubules further approach from each other, the compression of tau cross-bridges generates an energy barrier. Consequently, the short-range attraction force due to macromolecular crowding can no longer take place (Fig. 6a), which precludes the formation of tightly packed microtubule bundles. In cells over-expressing tau, tau-cross-bridges are numerous and artificially induce microtubule bundling (Fig. 6b). In vitro, only elevated tau:tubulin molar ratios and moderate ionic strengths allow the clear detection of microtubule bundles (Figs. 1 and S3). The biphasic effect of tau on the spatial organization of microtubules is misleading and led to the biased view that tau promotes microtubule bundling in vivo. Interestingly, thermally-induced movements of tau allow the exploration of a large surface area on the microtubule lattice in search for other tau proteins on the surface of other microtubules. Tau diffusion is thus critical to prevent microtubule bundling at tau:tubulin molar ratios well below the ratio required to saturate microtubules (Figs. 5b, c, 6b). The unified view on the role of tau in axonal microtubule organization presented in this study provides new insights into the role of tau in neuronal functions. For example, tau was reported to impair the long-range transport of vesicles and mitochondria along microtubules via molecular motors, as observed repeatedly in cells overexpressing tau [60–63]. However, both tau diffusion on microtubules and its ability to form cross-bridges in between microtubules enable to keep microtubules separated at low tau levels. At such low levels, the presence of tau on microtubules should therefore not constitute an obstacle for active transport of cargoes. The consequences of alternative splicing of tau mRNA and tau phosphorylation on the spatial distribution of axonal microtubules also deserve to be considered and could enlighten the process leading to axon degeneration observed in Alzheimer’s disease. The results present here show that the tau isoform with the longest N-terminal domain (2N) is the most potent to prevent microtubules from bundling under macromolecular crowding conditions. During neurogenesis, tau isoforms with short N-terminal domain (0N and 1N) are expressed but their expression is significantly reduced in adult brain [64]. In mature axons, the presence of tau 2N could thus be necessary to keep microtubules homogeneously distributed. The role of the number of microtubule binding repeats (4R or 3R) may also matter. 4R isoforms bind to microtubules with a higher affinity than 3R isoforms but a strong binding also limits tau diffusion on microtubules. Experimental data are thus required to clarify whether tau 4R or 3R have a similar ability to prevent microtubule bundling. In addition, tau phosphorylation leads to a lower affinity of tau for microtubules [65], which is considered critical for the redistribution of tau to the somatodendritic compartment and the accumulation of tau aggregates [66]. Our unified model however provides an alternative view. Tau phosphorylation events and especially those occurring in the proline rich and N-terminal domain may alter the capacity of tau to separate microtubules. An early event in Alzheimer’s disease could thus be the collapse of axonal microtubules into tightly packed bundles due to inappropriate tau phosphorylation events. Further investigations should be carried out to explore this hypothesis. Materials and methods Preparation of sheep brain tubulin and subtilisin-treated tubulin Tubulin was purified from sheep brains and stored at −80 °C in 20 mM MES–KOH, pH 6.9, 0.5 mM DTT, 0.5 mM EGTA, 0.25 mM MgCl2, 3.4 M glycerol, and 0.1 mM GTP. Before use, an additional cycle of polymerization was performed, and tubulin was resuspended in 20 mM MES–KOH, pH 6.9, 0.25 mM EGTA, 0.25 mM MgCl2. Tubulin concentration was determined by spectrophotometry using an extinction coefficient ϵ278 nm = 1.2 × 105 M−1 cm−1 [28]. Turbidimetry measurements The kinetics of microtubule assembly and (or) bundling was monitored by turbidimetry at 370 nm using an Ultrospec 3000 spectrophotometer (GE Healthcare, Fairfield, CT) equipped with a temperature controller. Microtubule assembly was obtained after preincubating tubulin samples on ice for 5 min in polymerization buffer (10 mM HEPES–KOH pH 6.8, 30 mM KCl, 20 % glycerol, 1 mM EGTA, 4 mM MgCl2, 1 mM GTP) in the presence or absence of tau. Tubulin polymerization was then initiated by shifting the temperature to 37 °C. Microtubule bundling was monitored by turbidimetry at 37 °C after adding either 1 % PEG 35K or tau at varying concentrations to taxol-stabilized microtubules in the indicated buffer. Cloning of tau isoforms Tau 2N/4R and deltaNT cDNAs were obtained by amplifying hTau40 pET29b plasmid (catalogue no. 16316, Addgene, Cambridge, MA) either from the beginning or from P172 to L441 using Phusion Hot Start II High-Fidelity DNA Polymerase (catalogue: F-537L, Thermo Fischer Scientific). The PCR products were first cloned into pENTR/D-TOPO by TOPO cloning (primers are listed in Supplemental Table 1) and then moved into pDEST17 by the LR reaction for recombinant protein expression with a His-Tag (Life Technologies). To obtain hTau24 three PCRs was necessary) using Hot Start II High-Fidelity DNA Polymerase (catalogue: F-537L, Thermo Fischer Scientific). Tau full length plasmid was used as template to design the primers and to amplify two fragments encoding hTau24 sequence (accession number NM_016834.4). For all PCRs, cycling was a 30 s initial denaturation at 98 °C, 35 cycles with 30 s denaturation at 98 °C and a 30 s annealing at 72 °C and a final extension also at 72 °C for 5 min. PCRs product was cloned into pENTR/D-TOPO by TOPO cloning then moved into pDEST17 by the LR reaction. Production of Human Tau Protein Rosetta-gami™ 2 competent cells were transformed with hTau40 (largest isoform), hTau24 (shortest 4R isoform), and deltaNT (mutant) expressed in Gateway® pDEST™17 Vector (catalogue no. 11803-012, Life technologies). Bacteria were grown in LB medium in the presence of 100 μg/ml ampicillin and 15 µg/ml Chloramphenicol. Overexpression was induced at A600 nm = 0.5 with 0.7 mM isopropyl β-d-thiogalactopyranoside, and incubation was continued for 3.5 h at 37 °C. Bacteria were pelleted by a 10-min, 4000×g centrifugation, and the pellet was resuspended in Buffer A (25 mM Tris–HCl, 25 mM Mes–KOH, 500 mM NaCl, 0.2 mM MgCl2, 1 mM PMSF, 5 mM DTT, pH 7.5). Bacteria were then disrupted by sonication, and centrifuged for 10 min (4000×g), the supernatant was then boiled for 20 min, and ultracentrifuged at 100,000×g for 45 min at 4 °C. Clarified cell lysate was then loaded on a Ni2+ -nitrilotriacetic acid column (HisTrap HP, 1 × 1 ml, GE Healthcare Life Sciences). The proteins were eluted with 5 column volumes of buffer B (25 mM Tris–HCl, 25 mM Mes–KOH, 250 mM NaCl, 250 mM Imidazole, 0.2 mM MgCl2, 1 mM PMSF, 5 mM DTT, pH 7.5). Purified proteins were then dialyzed overnight at 4 °C to cleave the His-Tag against 20 mM Tris, 20 mM Mes, 250 mM NaCl, 0.5 mM EDTA, 1 mM DTT, pH 7.5) using TEV protease (catalogue no. T4455, SIGMA). This step was followed by a second nickel-affinity chromatography to remove the uncleaved recombinant tau and TEV protease. Fractions of interest were combined and dialyzed against 25 mM Hepes, 250 mM NaCl, 0.25 mM DTT, 0.1 mM PMSF, pH 7.4 to eliminate traces of imidazole. Proteins were concentrated by ultrafiltration (Corning® Spin-X® UF 500, 10 kDa cut-off) and the final concentration of Tau was determined by amino acid analysis. All of the purification steps were performed in the presence of complete Protease Inhibitor Cocktail tablets (catalogue no. 00000001187358000, Roche Applied Science). Atomic force microscopy Samples containing microtubules and tau under specified conditions were deposited on freshly cleaved mica and dried for atomic force microscopy (AFM) imaging, using a protocol that we developed [67]. The electrostatic adsorption of microtubules on mica is mediated by magnesium ions present in the buffer. All AFM experiments were performed in peak force mode with Nanoscope V (Bruker/Veeco, Santa Barbara, CA). The peak force tapping mode was performed using silicon tips (Scanasyst-Air-HR, Bruker). The applied force was minimized as much as possible. Transmission electron microscopy (TEM) For ultrathin sectioning, microtubules were prepared with 20 μM tubulin with or without 7 µM of tau and (or) 1 % PEG 35K in 10 mM HEPES–KOH, 50 mM KCl, pH 6.8, 1 mM EGTA, 4 mM MgCl2 and 1 mM GTP, 20 % glycerol. Microtubules were pelleted at 20,000×g for 30 min at 37 °C. The pellets were gently resuspended in 10 mM HEPES–KOH, 50 mM KCl, pH 6.8, 1 mM EGTA, 4 mM MgCl2 and 1 mM GTP, 20 % glycerol, 1 % glutaraldehyde, 0.2 % tannic acid and incubated for fixation for 1 h at room temperature. Samples were then post-fixed with 1 % osmium tetroxyde in cacodylate buffer 0.1 M, pH 7.3. After dehydration in ethanol bathes of increasing concentrations, pellets were embedded in epoxy resin (Embed-812 Embedding kit #14120, EMS). Collodion-carbon-coated copper grids were used to collect the ultrathin sections of 40 nm thickness. The sections were then stained sequentially with 2 % uranyl acetate aqueous and Reynold’s solutions and analyzed in bright field mode using a Zeiss 902 transmission electron microscope. Images were acquired using a Megaview III CCD camera with the iTEM software (Olympus Soft Imaging Solution) at a magnification of 79,000×. Immunofluorescence HeLa cells and (or) neurons were grown on poly-l-lysine-coated glass coverslips and fixed in ice-cold methanol immediately followed by fixation in 4 % paraformaldehyde at 37 °C in PBS for 30 min. After fixation, cells were then washed and incubated for 1 h with mouse monoclonal anti-tubulin antibody (E7, 1:2000 dilutions) and anti-tau antibody (SC-1995, Santa-Cruz, CA). Cells were washed extensively in PBS and incubated for 1 h with fluorochrome (Alexa Fluor® 488 and -555)-coupled secondary antibodies (Invitrogen) in blocking solution. The protocol used to measure the tau:tubulin fluorescence ratios on microtubule structures is described in figure S5. In vitro, microtubules were polymerized as described in “Turbidimetry measurements” and deposited on freshly cleaved mica. Samples were fixed with 4 % paraformaldehyde in PBS at 37 °C for 10 min. After fixation, samples were then washed three times in PBS and prepared for immunofluorescence as described above. The tau:tubulin fluorescence ratios were measured as described in figure S2. Co-culture of neurons and HeLa cells Cortical neurons from embryonic mice (E19) were prepared as described previously [68] and grown on Poly-l-ornithine-coated (Sigma-Aldrich) 14 mm coverslips at a density of 100,000 cells/coverslips in Neurobasal media, supplemented with 2 % B27, 2 mM l-glutamine. Hela cells previously transfected with 2N tau by using Lipofectamine 2000 (Invitrogen) were co-cultured on neurons at 7-day in vitro (DIV) for 8 h. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary material 1 (PDF 772 kb) Supplementary material 2 (PDF 72 kb) Supplementary material 3 (WMV 975 kb) Supplementary material 4 (WMV 981 kb) Supplementary material 5 (WMV 977 kb) Supplementary material 6 (WMV 994 kb) We gratefully acknowledge the Genopole Evry and INSERM for constant support of the laboratory. We also thank Maroun Charbel for helpful discussion about the model. ==== Refs References 1. Cleveland DW Hwo SY Kirschner MW Physical and chemical properties of purified tau factor and the role of tau in microtubule assembly J Mol Biol 1977 116 227 247 10.1016/0022-2836(77)90214-5 146092 2. Ballatore C Lee VM Trojanowski JQ Tau-mediated neurodegeneration in Alzheimer’s disease and related disorders Nat Rev Neurosci 2007 8 663 672 10.1038/nrn2194 17684513 3. Ittner LM Gotz J Amyloid-beta and tau–a toxic pas de deux in Alzheimer’s disease Nat Rev Neurosci 2011 12 65 72 10.1038/nrn2967 21193853 4. Kosik KS Joachim CL Selkoe DJ Microtubule-associated protein tau (tau) is a major antigenic component of paired helical filaments in Alzheimer disease Proc Natl Acad Sci U S A 1986 83 4044 4048 10.1073/pnas.83.11.4044 2424016 5. Gustke N Trinczek B Biernat J Mandelkow EM Mandelkow E Domains of tau protein and interactions with microtubules Biochemistry 1994 33 9511 9522 10.1021/bi00198a017 8068626 6. Morris M Maeda S Vossel K Mucke L The many faces of tau Neuron 2011 70 410 426 10.1016/j.neuron.2011.04.009 21555069 7. Lee G Rook SL Expression of tau protein in non-neuronal cells: microtubule binding and stabilization J Cell Sci 1992 102 Pt 2 227 237 1400630 8. Drechsel DN Hyman AA Cobb MH Kirschner MW Modulation of the dynamic instability of tubulin assembly by the microtubule-associated protein tau Mol Biol Cell 1992 3 1141 1154 10.1091/mbc.3.10.1141 1421571 9. Kadavath H Hofele RV Biernat J Kumar S Tepper K Urlaub H Mandelkow E Zweckstetter M Tau stabilizes microtubules by binding at the interface between tubulin heterodimers Proc Natl Acad Sci U S A 2015 112 7501 7506 10.1073/pnas.1504081112 26034266 10. Bunker JM Wilson L Jordan MA Feinstein SC Modulation of microtubule dynamics by tau in living cells: implications for development and neurodegeneration Mol Biol Cell 2004 15 2720 2728 10.1091/mbc.E04-01-0062 15020716 11. Breuzard G Hubert P Nouar R De Bessa T Devred F Barbier P Sturgis JN Peyrot V Molecular mechanisms of Tau binding to microtubules and its role in microtubule dynamics in live cells J Cell Sci 2013 126 2810 2819 10.1242/jcs.120832 23659998 12. Grundke-Iqbal I Iqbal K Tung YC Quinlan M Wisniewski HM Binder LI Abnormal phosphorylation of the microtubule-associated protein tau (tau) in Alzheimer cytoskeletal pathology Proc Natl Acad Sci U S A 1986 83 4913 4917 10.1073/pnas.83.13.4913 3088567 13. Alonso AC Grundke-Iqbal I Iqbal K Alzheimer’s disease hyperphosphorylated tau sequesters normal tau into tangles of filaments and disassembles microtubules Nat Med 1996 2 783 787 10.1038/nm0796-783 8673924 14. Lindwall G Cole RD Phosphorylation affects the ability of tau protein to promote microtubule assembly J Biol Chem 1984 259 5301 5305 6425287 15. Busciglio J Lorenzo A Yeh J Yankner BA beta-amyloid fibrils induce tau phosphorylation and loss of microtubule binding Neuron 1995 14 879 888 10.1016/0896-6273(95)90232-5 7718249 16. Peters A Vaughn JE Microtubules and filaments in the axons and astrocytes of early postnatal rat optic nerves J Cell Biol 1967 32 113 119 10.1083/jcb.32.1.113 10976204 17. Yamada KM Spooner BS Wessells NK Ultrastructure and function of growth cones and axons of cultured nerve cells J Cell Biol 1971 49 614 635 10.1083/jcb.49.3.614 4326456 18. Schulz A Baader SL Niwa-Kawakita M Jung MJ Bauer R Garcia C Zoch A Schacke S Hagel C Mautner VF Merlin isoform 2 in neurofibromatosis type 2-associated polyneuropathy Nat Neurosci 2013 16 426 433 10.1038/nn.3348 23455610 19. Quraishe S Cowan CM Mudher A NAP (davunetide) rescues neuronal dysfunction in a Drosophila model of tauopathy Mol Psychiatry 2013 18 834 842 10.1038/mp.2013.32 23587881 20. Stephan R Goellner B Moreno E Frank CA Hugenschmidt T Genoud C Aberle H Pielage J Hierarchical microtubule organization controls axon caliber and transport and determines synaptic structure and stability Dev Cell 2015 33 5 21 10.1016/j.devcel.2015.02.003 25800091 21. Mukhopadhyay R Kumar S Hoh JH Molecular mechanisms for organizing the neuronal cytoskeleton Bioessays 2004 26 1017 1025 10.1002/bies.20088 15351972 22. Marx A Pless J Mandelkow EM Mandelkow E On the rigidity of the cytoskeleton: are MAPs crosslinkers or spacers of microtubules? Cell Mol Biol (Noisy-le-grand) 2000 46 949 965 10976876 23. Mukhopadhyay R Hoh JH AFM force measurements on microtubule-associated proteins: the projection domain exerts a long-range repulsive force FEBS Lett 2001 505 374 378 10.1016/S0014-5793(01)02844-7 11576531 24. Rosenberg KJ Ross JL Feinstein HE Feinstein SC Israelachvili J Complementary dimerization of microtubule-associated tau protein: implications for microtubule bundling and tau-mediated pathogenesis Proc Natl Acad Sci U S A 2008 105 7445 7450 10.1073/pnas.0802036105 18495933 25. Kanai Y Chen J Hirokawa N Microtubule bundling by tau proteins in vivo: analysis of functional domains EMBO J 1992 11 3953 3961 1396588 26. Kanai Y Takemura R Oshima T Mori H Ihara Y Yanagisawa M Masaki T Hirokawa N Expression of multiple tau isoforms and microtubule bundle formation in fibroblasts transfected with a single tau cDNA J Cell Biol 1989 109 1173 1184 10.1083/jcb.109.3.1173 2504728 27. Chung PJ Choi MC Miller HP Feinstein HE Raviv U Li Y Wilson L Feinstein SC Safinya CR Direct force measurements reveal that protein Tau confers short-range attractions and isoform-dependent steric stabilization to microtubules Proc Natl Acad Sci U S A 2015 112 E6416 E6425 10.1073/pnas.1513172112 26542680 28. Hinrichs MH Jalal A Brenner B Mandelkow E Kumar S Scholz T Tau protein diffuses along the microtubule lattice J Biol Chem 2012 287 38559 38568 10.1074/jbc.M112.369785 23019339 29. Bounedjah O Hamon L Savarin P Desforges B Curmi PA Pastre D Macromolecular crowding regulates assembly of mRNA stress granules after osmotic stress: new role for compatible osmolytes J Biol Chem 2012 287 2446 2458 10.1074/jbc.M111.292748 22147700 30. Needleman DJ Ojeda-Lopez MA Raviv U Ewert K Miller HP Wilson L Safinya CR Radial compression of microtubules and the mechanism of action of taxol and associated proteins Biophys J 2005 89 3410 3423 10.1529/biophysj.104.057679 16100275 31. Chen J Kanai Y Cowan NJ Hirokawa N Projection domains of MAP2 and tau determine spacings between microtubules in dendrites and axons Nature 1992 360 674 677 10.1038/360674a0 1465130 32. Baas PW Pienkowski TP Kosik KS Processes induced by tau expression in Sf9 cells have an axon-like microtubule organization J Cell Biol 1991 115 1333 1344 10.1083/jcb.115.5.1333 1955477 33. Rodriguez-Martin T Cuchillo-Ibanez I Noble W Nyenya F Anderton BH Hanger DP Tau phosphorylation affects its axonal transport and degradation Neurobiol Aging 2013 34 2146 2157 10.1016/j.neurobiolaging.2013.03.015 23601672 34. Yu W Baas PW Changes in microtubule number and length during axon differentiation J Neurosci 1994 14 2818 2829 8182441 35. Brunden KR Zhang B Carroll J Yao Y Potuzak JS Hogan AM Iba M James MJ Xie SX Ballatore C Epothilone D improves microtubule density, axonal integrity, and cognition in a transgenic mouse model of tauopathy J Neurosci 2010 30 13861 13866 10.1523/JNEUROSCI.3059-10.2010 20943926 36. Liu Y Lv K Li Z Yu AC Chen J Teng J PACSIN1, a Tau-interacting protein, regulates axonal elongation and branching by facilitating microtubule instability J Biol Chem 2012 287 39911 39924 10.1074/jbc.M112.403451 23035120 37. Schaefer MK Schmalbruch H Buhler E Lopez C Martin N Guenet JL Haase G Progressive motor neuronopathy: a critical role of the tubulin chaperone TBCE in axonal tubulin routing from the Golgi apparatus J Neurosci 2007 27 8779 8789 10.1523/JNEUROSCI.1599-07.2007 17699660 38. Palay SL Sotelo C Peters A Orkand PM The axon hillock and the initial segment J Cell Biol 1968 38 193 201 10.1083/jcb.38.1.193 5691973 39. Cooper JR Wordeman L The diffusive interaction of microtubule binding proteins Curr Opin Cell Biol 2009 21 68 73 10.1016/j.ceb.2009.01.005 19185482 40. Brittain WJ Minko S A structural definition of polymer brushes J Polym Sci, Part A: Polym Chem 2007 45 3505 3512 10.1002/pola.22180 41. Binder LI Frankfurter A Rebhun LI The distribution of tau in the mammalian central nervous system J Cell Biol 1985 101 1371 1378 10.1083/jcb.101.4.1371 3930508 42. Drubin DG Feinstein SC Shooter EM Kirschner MW Nerve growth factor-induced neurite outgrowth in PC12 cells involves the coordinate induction of microtubule assembly and assembly-promoting factors J Cell Biol 1985 101 1799 1807 10.1083/jcb.101.5.1799 2997236 43. Khatoon S Grundke-Iqbal I Iqbal K Brain levels of microtubule-associated protein tau are elevated in Alzheimer’s disease: a radioimmuno-slot-blot assay for nanograms of the protein J Neurochem 1992 59 750 753 10.1111/j.1471-4159.1992.tb09432.x 1629745 44. Hiller G Weber K Radioimmunoassay for tubulin: a quantitative comparison of the tubulin content of different established tissue culture cells and tissues Cell 1978 14 795 804 10.1016/0092-8674(78)90335-5 688394 45. Conway L Gramlich MW Ali Tabei SM Ross JL Microtubule orientation and spacing within bundles is critical for long-range kinesin-1 motility Cytoskeleton (Hoboken) 2014 71 595 610 10.1002/cm.21197 25382100 46. Hamon L Savarin P Curmi PA Pastre D Rapid assembly and collective behavior of microtubule bundles in the presence of polyamines Biophys J 2011 101 205 216 10.1016/j.bpj.2011.05.022 21723831 47. Brandt R Lee G Functional organization of microtubule-associated protein tau. Identification of regions which affect microtubule growth, nucleation, and bundle formation in vitro J Biol Chem 1993 268 3414 3419 8429017 48. Choi MC Raviv U Miller HP Gaylord MR Kiris E Ventimiglia D Needleman DJ Kim MW Wilson L Feinstein SC Human microtubule-associated-protein tau regulates the number of protofilaments in microtubules: a synchrotron x-ray scattering study Biophys J 2009 97 519 527 10.1016/j.bpj.2009.04.047 19619466 49. Scott CW Klika AB Lo MM Norris TE Caputo CB Tau protein induces bundling of microtubules in vitro: comparison of different tau isoforms and a tau protein fragment J Neurosci Res 1992 33 19 29 10.1002/jnr.490330104 1360542 50. Aamodt EJ Culotti JG Microtubules and microtubule-associated proteins from the nematode Caenorhabditis elegans: periodic cross-links connect microtubules in vitro J Cell Biol 1986 103 23 31 10.1083/jcb.103.1.23 3722265 51. Goode BL Denis PE Panda D Radeke MJ Miller HP Wilson L Feinstein SC Functional interactions between the proline-rich and repeat regions of tau enhance microtubule binding and assembly Mol Biol Cell 1997 8 353 365 10.1091/mbc.8.2.353 9190213 52. Schiff PB Horwitz SB Taxol stabilizes microtubules in mouse fibroblast cells Proc Natl Acad Sci U S A 1980 77 1561 1565 10.1073/pnas.77.3.1561 6103535 53. Askham JM Vaughan KT Goodson HV Morrison EE Evidence that an interaction between EB1 and p150(Glued) is required for the formation and maintenance of a radial microtubule array anchored at the centrosome Mol Biol Cell 2002 13 3627 3645 10.1091/mbc.E02-01-0061 12388762 54. Errico A Ballabio A Rugarli EI Spastin, the protein mutated in autosomal dominant hereditary spastic paraplegia, is involved in microtubule dynamics Hum Mol Genet 2002 11 153 163 10.1093/hmg/11.2.153 11809724 55. Desforges B Savarin P Bounedjah O Delga S Hamon L Curmi PA Pastre D Gap junctions favor normal rat kidney epithelial cell adaptation to chronic hypertonicity Am J Physiol Cell Physiol 2011 301 C705 C716 10.1152/ajpcell.00128.2011 21677260 56. Black MM Slaughter T Moshiach S Obrocka M Fischer I Tau is enriched on dynamic microtubules in the distal region of growing axons J Neurosci 1996 16 3601 3619 8642405 57. Milner ST Polymer brushes Science 1991 251 905 914 10.1126/science.251.4996.905 17847384 58. Smyda MR Harvey SC The entropic cost of polymer confinement J Phys Chem B 2012 116 10928 10934 10.1021/jp302807r 22905742 59. Peter SJ Mofrad MR Computational modeling of axonal microtubule bundles under tension Biophys J 2012 102 749 757 10.1016/j.bpj.2011.11.4024 22385845 60. Stamer K Vogel R Thies E Mandelkow E Mandelkow EM Tau blocks traffic of organelles, neurofilaments, and APP vesicles in neurons and enhances oxidative stress J Cell Biol 2002 156 1051 1063 10.1083/jcb.200108057 11901170 61. Mandelkow EM Stamer K Vogel R Thies E Mandelkow E Clogging of axons by tau, inhibition of axonal traffic and starvation of synapses Neurobiol Aging 2003 24 1079 1085 10.1016/j.neurobiolaging.2003.04.007 14643379 62. Dubey M Chaudhury P Kabiru H Shea TB Tau inhibits anterograde axonal transport and perturbs stability in growing axonal neurites in part by displacing kinesin cargo: neurofilaments attenuate tau-mediated neurite instability Cell Motil Cytoskelet 2008 65 89 99 10.1002/cm.20243 63. Ebneth A Godemann R Stamer K Illenberger S Trinczek B Mandelkow E Overexpression of tau protein inhibits kinesin-dependent trafficking of vesicles, mitochondria, and endoplasmic reticulum: implications for Alzheimer’s disease J Cell Biol 1998 143 777 794 10.1083/jcb.143.3.777 9813097 64. Bullmann T Holzer M Mori H Arendt T Pattern of tau isoforms expression during development in vivo Int J Dev Neurosci 2009 27 591 597 10.1016/j.ijdevneu.2009.06.001 19540327 65. Gu GJ Lund H Wu D Blokzijl A Classon C von Euler G Landegren U Sunnemark D Kamali-Moghaddam M Role of individual MARK isoforms in phosphorylation of tau at Ser(2)(6)(2) in Alzheimer’s disease NeuroMol Med 2013 15 458 469 10.1007/s12017-013-8232-3 66. Noble W Hanger DP Miller CC Lovestone S The importance of tau phosphorylation for neurodegenerative diseases Front Neurol 2013 4 83 10.3389/fneur.2013.00083 23847585 67. Hamon L Curmi PA Pastre D High-resolution imaging of microtubules and cytoskeleton structures by atomic force microscopy Methods Cell Biol 2010 95 157 174 10.1016/S0091-679X(10)95009-7 20466134 68. Danglot L Triller A Bessis A Association of gephyrin with synaptic and extrasynaptic GABAA receptors varies during development in cultured hippocampal neurons Mol cell Neurosci 2003 23 264 278 10.1016/S1044-7431(03)00069-1 12812758

PMC005xxxxxx/PMC5002046.txt

==== Front Soc Indic ResSoc Indic ResSocial Indicators Research0303-8300Springer Netherlands Dordrecht 104410.1007/s11205-015-1044-0ArticleMultidimensional Poverty in Indonesia: Trend Over the Last Decade (2003–2013) Hanandita Wulung +44 (0) 77 7639 9580wulung.hanandita@postgrad.manchester.ac.uk Tampubolon Gindo gindo.tampubolon@manchester.ac.uk Cathie Marsh Institute for Social Research, University of Manchester, Humanities Bridgeford Street Building 2F, Oxford Road, Manchester, M13 9PL UK 1 8 2015 1 8 2015 2016 128 2 559 587 16 7 2015 © The Author(s) 2015 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.The notion of poverty as an experience of multiple deprivation has been widely acknowledged. In Indonesia, however, poverty assessment has almost exclusively been conducted within the monetary space; even when multidimensionality is admitted, it has always been computed using variants of marginal method that are indifferent to joint deprivation. Applying a novel measurement method that is sensitive to both the incidence and the intensity of multiple deprivation to data from the National Socio-economic Survey (Susenas), this paper investigates the extent and the patterns of multidimensional poverty in Indonesia from 2003 to 2013 (N=7,148,964). An Indonesian version of the multidimensional poverty index is constructed by augmenting the existing consumption poverty measure with information on health and education. Results suggest that there was an unambiguous poverty reduction over the last decade at both national and sub-national levels. The data also reveal that progress has been inclusive across population subgroups, although spatial variation remains notable. The new poverty measurement method proves to be easily adaptable to the Indonesian context and could complement the methods currently employed by the Indonesian Statistical Bureau. Keywords Poverty assessmentMultidimensional poverty index IndonesiaSusenasAlkire–Foster methodissue-copyright-statement© Springer Science+Business Media Dordrecht 2016 ==== Body Introduction Income, or consumption poverty measures such as the World Bank’s dollar-a-day headcount ratio (Ravallion et al. 2009), is still the most prevalent measure of poverty used across the globe. However, from Asia to Africa (Batana 2013; Klasen 2000; Santos 2013; Ranis and Stewart 2012; Yu 2013), and across Europe to Latin America (Battison et al. 2013; Brandolini and D’Alessio 1998; Whelan et al. 2004), scholars have consistently documented that the lack of money is not always an accurate proxy for deprivations that society cares about. It has been argued that money metrics do not tell the whole story of human suffering, because poverty is not only about one’s inability to spend on essential goods and services. More than that, it is about one’s inability to enjoy valuable beings and doings (Sen 1985). Indeed, what is now generally accepted is a notion of poverty (or well-being for that matter) as an intrinsically multidimensional construct that encompasses the whole range of ways in which an individual can participate effectively in society. Since the seminal works of Townsend (1979) and Sen (1985), different multidimensional poverty measures have been developed. Yet, as noted by Santos and Ura (2008: 1), ‘some of the proposed measures seem to have incorporated a multidimensional perspective at the cost of giving up the simplicity and intuition that characterise the unidimensional measures’. Statistical approaches to multidimensional poverty measurement (Filmer and Pritchett 2001; Sahn and Stifel 2003), for instance, rely on multivariate or latent-variable techniques to the extent that parameters are completely data-driven, leaving evaluators with limited control over the measure. Some axiomatic alternatives such as Bourguignon and Chakravarty (2003), on the other hand, satisfy a number of useful measurement properties but do strictly necessitate the availability of cardinal data; in reality, vital social indicators such as literacy and completion of primary school are usually ordinal in nature. In an attempt to address these problems, Alkire and Foster (2011a; henceforth AF) proposed a new sort of multidimensional poverty measure: one that is not only simple to construct, but also retains many of the properties of the well-known Foster–Greer–Thorbecke (FGT) measures (Foster et al. 1984) of unidimensional poverty measurement. The AF method combines the FGT with the counting approach (Atkinson 2003), which is easy to understand and has a long history in sociology. The method deals with ordinal data in a straightforward manner by dichotomising individuals’ achievement into deprived and non-deprived states. Aggregation is then performed, first across deprivations experienced by each individual, and then across individuals, yielding a measure that is intuitively interpretable as the share of deprivations that poor individuals experience out of the total deprivations that the society could possibly experience. As a generalisation of the classical FGT, the AF family of multidimensional poverty measures satisfies an array of desirable axioms (Alkire and Foster 2011a). Foremost among them are their ‘subgroup decomposition’ and ‘dimensional breakdown’ properties, which allow the overall poverty measure to be broken down into its social, geographical or dimensional constituents in a way that is both conceptually and technically defensible. A thorough characterisation of joint deprivations is further made possible by the availability of partial indices that capture the incidence as well as the intensity of poverty. The methodology is also transparent in the sense that all parameters are under the control of the evaluator, allowing normative decisions with regard to the selection of indicators, dimensional and poverty cut-offs, and weighting schemes to be easily incorporated into the analysis. In fact, acknowledging these novelties, in 2010 the United Nations Development Programme (UNDP) replaced its Human Poverty Index (HPI; first published in 1997) with the new multidimensional poverty index (MPI), based on the AF family of multidimensional measure (UNDP 2014, 2010: 95). Applying the AF method to the National Socio-economic Survey data from Indonesia, this paper seeks to estimate the extent and to investigate the regional as well as the temporal patterns of multidimensional poverty in Indonesia for 11 consecutive years spanning from 2003 to 2013. The aim of this study is not to replace the official consumption poverty estimate with a new one, but rather to augment the conventional poverty measure with additional information on health and education using the same data source that has historically been used to estimate the official consumption poverty figure in Indonesia. This version of an Indonesian multidimensional poverty index (MPI) is constructed in a way that income poor individuals are ‘automatically’ multidimensionally poor, but not the converse. The present study considers the following questions: taking into account income, health and education dimensions, how many Indonesians are poor overall? Are urban areas always better off? Which island of the archipelago is the most deprived? Did recent progress, if any, benefit the poorest of the poor? And what happened to gender and spatial inequities during the last decade? Indonesia, the world’s largest archipelagic state and the third-most populous developing country, is known for its exemplary achievement in terms of income poverty reduction and overall human development (Ranis and Stewart 2012). However, there is little research attempting to understand the nature of simultaneous deprivations experienced by its people. The majority of recent poverty evaluations have been conducted exclusively within the monetary space (Ilmma and Wai-Poi 2014; Strauss et al. 2004; Sumarto et al. 2014); even when multidimensionality is sought, it has always been computed using variants of marginal method (BPS 2015b; BPS et al. 2004) that are blind to joint deprivation (Alkire 2011: 503–504). To date, only two studies attempted to measure the extent of simultaneous deprivations in Indonesia. Alkire and Foster (2011a), in the earliest showcase of their methodology, provided a national poverty estimate for the year 2007 using the Indonesia Family Life Survey data (IFLS; Thomas et al. 2012). But it is known that the IFLS sampling frame is not entirely representative of the population (RAND 2007); it neglects individuals living in the eastern islands of the archipelago (RAND 2014), yielding a sample that favours the relatively well-developed areas in western Indonesia. Alkire and Santos (2014) carried out further study on Indonesia using the Demographic Health Survey (ICF International 2012) data as a part of a grand endeavour to construct a globally comparable MPI (UNDP 2010). While they are completely representative of the population, the DHS data do not, however, provide household consumption expenditure information, preventing a useful comparison with the official measure of consumption poverty. The contribution of the present study to the existing literature is threefold. Firstly, in estimating the extent of multidimensional poverty in Indonesia, this study uses large and nationally representative data that have been regarded as the primary source of information among Indonesian policy makers as well as international observers. While concurring with Alkire and Santos (2014: 266) who stress that data availability has been the major bottleneck in the development of an internationally comparable MPI, we would like to demonstrate that even when using an existing data source, the construction of an Indonesian MPI is not only technically feasible but also substantively meaningful. The collection of better well-being data is of course desirable, but Indonesians do not have to wait until the ‘perfect’ data becomes available to have their progress assessed. Secondly, the inclusion of consumption expenditure information makes this version of Indonesian MPI not only comparable to the official poverty measure, but also sensitive to economic fluctuations (Ravallion 2010: 11). Lastly, by providing an annual analysis of the trend of multidimensional poverty in the last 11 years, this study presents a richer picture compared to one that analyses only selected points in time over the same period. The remainder of the paper is structured as follows. The next section describes the AF method, the data and the dimensions. It then investigates the degree to which income poverty correlates with non-income deprivations. Section 3 presents the results. Initially, unidimensional deprivations are investigated using the marginal dashboard approach. Then, MPI estimates at national and sub-national levels are presented along with robustness checks. Finally, changes in the distribution of deprivations among the poor are studied. Section 4 concludes. Methods The Alkire–Foster Method This section describes the Alkire–Foster method for multidimensional poverty measurement (Alkire and Foster 2011a). For brevity, we focus only on those aspects of the methodology that are directly relevant to the present study. We also limit our attention to the general case, where the social indicators being considered might not have cardinal meaning. Further in-depth expositions are available in Alkire and Foster (2011a), Alkire and Foster (2011b), Alkire and Santos (2013) and Seth and Alkire (2014). Setup Before describing the identification and the aggregation steps of the Alkire–Foster method, it is necessary to outline some preliminary setups. First of all, let us consider an n×d dimensional achievement matrix x=xij, in which the row i=1,…,n indexes the individuals under study and the column j=1,…,j indexes indicators for every dimension that the society cares about. No restriction is placed on the cardinality of indicators entered into the matrix; ordinal variables are acceptable. In this matrix, individuals’ achievements are recorded in the row vectors (xi·), while the marginal distribution of achievements is reflected in the column vectors (x·j). We also define a deprivation cut-off vector z=(z1,…,zj) indicating the minimum level of achievement in every social indicator that should be attained by each individual in the society. The relative importance (trade-off) of each achievement indicator in the achievement matrix x is governed by a vector of weight w=(w1,…,wd) such that ∑j=1dwj=d. Of course, the choice of indicator, deprivation cut-off, and weighting scheme is largely contingent upon the specific context of study (what the society values, the aim and scope of the study, or data availability) and is open to public debate (Alkire 2011). Identification In the AF framework, identification begins with the construction of a deprivation matrix g0=gij0 whose element is defined as gij0=wj if xij<zj and gij0=0 if otherwise. This deprivation matrix contains information about ‘who is deprived in which indicator and how much weight the indicators carry’ (Alkire and Santos 2013: 242). From g0 matrix, a deprivation count vector c=(ci,…,cn) whose element is ci=∑j=1dgij0 is constructed. This column vector stores the sum of weighted deprivations experienced by each individual under study. The Alkire–Foster identification function ρk(xi;z) is such that ρk(xi;z)=1 if ci≥k and ρk(xi;z)=0 if otherwise, where k is the poverty cut-off denoting the minimum sum of weighted deprivations required to be multidimensionally poor (Alkire and Santos 2013). The plausible choice of poverty cut-off is k∈min(wj),d and like other parameters in the AF framework, its value may be subjected to sensitivity analysis. Naturally, one would expect that the larger the k, the smaller the number of individuals identified as multidimensionally poor, and vice versa. When k=min(wj), a union identification criterion is obtained, but when k=d, an intersection identification criterion is reached. In practice, however, an intermediate criterion (k = 0.33–0.50) is usually preferred (Alkire 2011). Having assessed how deprived each individual is and identified who the poor are, the next step is to construct a censored deprivation matrix g0(k)=gij0(k) whose element is defined as gij0(k)=gij0 if ci≥k and gij0(k)=0 if otherwise. Likewise, a censored deprivation count vector is constructed such that ci(k)=ci if ci≥k and ci(k)=0 if otherwise. This censoring mechanism allows analysts to focus only on those individuals who are identified as multidimensionally poor, guaranteeing that the aggregate poverty measure is insensitive to the achievement of non-poor individuals. Aggregation The main aggregation method in the AF family of multidimensional poverty measure is the adjusted headcount ratio or M0, which is ‘the proportion of weighted deprivations that the poor experience in a society out of all the total potential deprivations that the society could experience’ (Santos 2013: 261). It is obtained by taking the arithmetic mean of the censored deprivation matrix g0(k): 1 M0(x;z)=1nd∑i=1n∑j=1dgij0(k) 2 =1n∑i=1n1d∑j=1dgij0(k)⏟individual\,poverty 3 =1d∑j=1d1n∑i=1ngij0(k)⏟censored\,deprivations 4 =1nq(k)⏟incidence (H)1q(k)∑i=1q(k)ci(k)d⏟intensity(A)whereq(k)=∑i=1nρk(xi;z)⏟head\,count Intuitively, M0 can also be understood either as the weighted sum of individual poverty (Eq. 2), the weighted sum of censored deprivations by indicators (Eq. 3), or the intensity-adjusted poverty incidence (Eq. 4: M0=H×A). The measure is, as its name implies, simultaneously sensitive to both the prevalence (incidence) and the scope (average deprivation among the poor, or intensity) of poverty. By definition, it is expected that as k increases, H will get smaller and A will get larger, and vice versa. Decomposition Because the adjusted headcount ratio can be expressed as the weighted sum of individual poverty (Eq. 2), the measure is decomposable by population subgroups. It follows that overall poverty can be expressed as the weighted sum of poverty measures in l number of population subgroups: 5 M0=∑s=1lnsnM0(s) and the contribution of a population subgroup s to the overall poverty M0 is: 6 Cs=nsn×M0(s)M0fors=1,…,l where nsn and M0(s)are the population share and the adjusted headcount ratio of subgroup s, respectively. Such a decomposition enables the assessment of the extent of inequality among subgroups by comparing each subgroup’s contribution to overall poverty relative to its population share (Alkire and Santos 2013: 245). A severe deviation from Cs/nsn=1 is indicative of the fact that a particular subgroup bears a disproportionately large (or small) share of poverty. Similarly, the adjusted headcount ratio can also be broken down by its indicators, because the measure is expressible as the weighted sum of the censored deprivations by indicators (Eq. 3). The overall poverty can thus be expressed as: 7 M0=∑j=1dwjdhj(k) and the contribution of a social indicator j to overall poverty M0 is: 8 Cj=wjd×hj(k)M0forj=1,…,d where hj(k) is the censored headcount ratio of indicator j. From this, we know that whenever Cj/wjd deviates severely from unity, then there is a relatively high (or low) deprivation in an indicator (Alkire and Santos 2013: 245). Dimensional contribution is obtainable simply by adding up Cj within a particular dimension. Robustness Analysis In the AF framework, robustness is established through sensitivity analysis employing different sets of indicators, deprivation cut-off, weight, or poverty cut-off (Alkire and Santos 2014). In this study, we apply poverty cut-off dominance analysis, confidence intervals overlap testing, and rank correlation testing, which constitute the standard robustness toolbox for the AF family of multidimensional poverty measures. Rate of Change Once some degree of robustness has been established, the rate of inter-temporal change in aggregate poverty can be calculated (Alkire and Vaz 2014). The absolute (ΔM0) and relative (δM0) rates of change are defined as follows: 9 ΔM0=M0(t2)-M0(t1) 10 δM0=M0(t2)-M0(t1)M0(t1)×100 where t2 and t1 denote the later and the initial time points, respectively. When the two time points span over a number of years, it is sometimes useful to express the changes in their annualised values: 11 Δ¯M0=M0(t2)-M0(t1)t2-t1 12 δ¯M0=M0(t2)M0(t1)1t2-t1-1×100 which give us the average absolute (or relative) change during the period of observation. Inequality Among the Poor and Across Subgroups Finally, after assessing the incidence and intensity of multidimensional poverty, it is only natural to ask whether poverty reduction, if any, has been inclusive among the poor and uniform across population subgroups (the ‘triple I’ of poverty: incidence, intensity, inequality; Sen 1976). Finding the right inequality measure for such a purpose has been proven to be non-trivial in the multidimensional setting; Seth and Alkire (2014) recently proposed a decomposable inequality measure based on the positive-multiple of variance to overcome the obstacles stemming mainly from the use of non-cardinal indicator variables in the construction of M0. Following their proposal, inequality (I) among poor individuals (Iq) and across subgroups (Is) can be expressed as: 13 Iq=β~×1q(k)∑i=1q(k)ci(k)-A2 14 Is=β~×∑s=1lnsnM0(s)-M02 where β~ is a normalisation factor that must be chosen such that I=0,1, respecting the properties of any standard inequality index. Because it is known that ‘the maximum possible value that variance takes is one fourth of the range of the deprivation score vector, which is attained when half of the population have the lowest deprivation scores and the other half have the highest deprivation scores’ (Seth and Alkire 2014: 16), β~ in the between-poor equation equals the inverse of 14maxci(k)-minci(k)2. Accordingly, as M0=0,1 then it is obvious that β~=4 in the between-subgroup equation. Data We analyse data from the Survei Sosial Ekonomi Nasional (National Socio-economic Survey; henceforth Susenas), an annual cross-sectional household survey administered by the Indonesian Statistical Bureau (Badan Pusat Statistik, BPS). Initiated in 1963, Susenas is a large and nationally representative survey, which has for decades served as the main source of information not only for the government of Indonesia but also for many international bodies, including the World Bank PovcalNet (see also Surbakti 1995; van de Walle 1988). The survey consists of a yearly core module (health, education, employment, household consumption expenditure, housing, fertility, contraception, and communication) and one of three alternating modules on (1) culture and education, (2) housing and health, and (3) household consumption expenditure, each administered once every three years. Compared to other Indonesian household survey data available to date, the strength of Susenas lies in (1) its comprehensive information on consumption expenditure (more than 300 food and non-food items in 2013), education and literacy; and in (2) its large sample and periodicity which permit precise annual inferences to be made at low levels of administration. However, it should be noted that the information on health available in Susenas is neither as comprehensive as that available in the Demographic and Health Survey (DHS) nor in the Indonesia Family Life Survey (IFLS) that were analysed previously by Alkire and Santos (2014) and Alkire and Foster (2011a). While Susenas records information on the number of disabled days and morbidity for each individual, it does not provide any anthropometric measure. Of course, a household survey that was first designed nearly four decades ago is by no means ideal for the contemporary purpose of multidimensional poverty measurement, but aside from this, the consumption expenditure data available in Susenas provide us with the opportunity to address the concern about the unresponsiveness to economic fluctuations (Ravallion 2010: 11) of the living standard indicators used in the current version of UNDP’s multidimensional poverty index (Alkire and Santos 2014; UNDP 2010). Since the Susenas sample is drawn using a multi-stage stratified random sampling design (urban/rural stratification, census blocks as the primary sampling unit, households within each block as the secondary sampling unit), the survey design along with the sampling weight is always incorporated into analysis. Our exploration indicates that ignoring the unequal sampling probability underestimates the proportion of individuals living on Java island severely (30 % instead of 60 %), leading to a potentially biased estimate of the population parameter. We analyse eleven consecutive years of Susenas data, from 2003 (just before the enactment of Law 32/2004 on Regional Government that marked the decentralisation era; 346 districts) to 2013 (the latest available Susenas; 499 districts). We regroup all districts that split during the period of observation into their original 2003 districts. The unit of analysis is an individual aged 18 and older. Children are excluded from the analysis because the relevant dimensions of their well-being depend on their age (Roche 2013), and information on those dimensions are missing from Susenas. We believe that children deserve special consideration that takes into account their own specificities (see Trani et al. 2013 and the cited works therein). Only complete cases are used in the analysis: individuals that have any social indicator (presented next) missing are dropped from the sample. This yields a total complete-case sample size of 7,148,964 individuals (N≈650,000 per year) with each survey wave having a final sample size of 91–100 % of the original sample size. Dimensions, Indicators, Cut-Offs and Weights In an ideal world, the choice of dimensions, indicators, cut-offs and weights for the measurement of multidimensional poverty would be guided by the revealed preferences of the poor (what the poor think of being poor, what deprivations matter the most, and what trade-offs the poor assign between deprivations). Yet, with the notable exceptions of Mexico’s Voices of the Poor study (Székely 2003) and Bhutan’s Gross National Happiness survey (Santos and Ura 2008; Ura et al. 2012), large-scale participatory exercises are rare. In contrast to these countries, the official conceptualisation of poverty in Indonesia is still the traditional consumption (or income) poverty measurement, defined as the failure to attain the consumption level required for the fulfilment of a basket of basic food and non-food needs (BPS 2015c). The idea of poverty as an experience of simultaneous deprivations has rarely penetrated the nation’s discourse of development (for example see Hill 2014 or Strauss et al. 2004). In this light, we base our elicitation of dimensions, indicators, cut-offs and weights on the existing Human Development Index (HDI; UNDP 2010), multidimensional poverty index (MPI; UNDP 2010), Millennium Development Goals (MDGs), and the 1945 Constitution of the Republic of Indonesia (MPR RI 2011), subject to constraints imposed by data availability in Susenas survey.Table 1 Dimensions, indicators, deprivation cut-offs and relative weights Dimension Indicator variable Deprivation cut-off ‘An individual is deprived if...’ Weight w1 w2 w3 Incomea Per capita daily consumption <$1.51 PPP 13 13 13 Healthb Illness episode >4 days 16 16 13 Morbidity >3 diseases 16 16 0 Educationc Schooling Has not completed primary school 16 13 13 Literacy Cannot read and write Latin characters 16 0 0 aThe first MDGs. The fourth paragraph of preamble, article 27(2) and 28C(1) of the Constitution bThe fourth, fifth and sixth MDGs. Article 28H(1) and 34(3) of the Constitution cThe second MDGs. The fourth paragraph of preamble, article 28C(1), 31(1) and 31(2) of the Constitution Three dimensions are included in our version of Indonesian MPI: (1) income, (2) health and (3) education, mimicking the UNDP’s latest HDI and MPI. Alkire and Santos (2014: 253) note that these dimensions are not only instrumental to many other vital outcomes but also intrinsically valuable in themselves. Furthermore, they argue that having only three dimensions simplifies communication and interpretability because ‘the contribution of the chosen dimensions is widely recognized across political and ideological divides’. As shown in Table 1, these dimensions are clearly related to the values of the Constitution, not to mention the MDGs. Income is operationalised using per capita daily consumption, which is obtained by deflating total daily household expenditure by household size. The figure is measured in international dollar (an expression of purchasing power parity, or PPP; UNSD 2014) and adjusted for spatial cost-of-living differences using the provincial urban-rural adjustment factors derived from the relative differences between the national and the local poverty lines (Ilmma and Wai-Poi 2014: 132). Fixed adjustment factors (from 2008) are used for the entire 2003–2013 series because these data are not available prior to 2007 (BPS 2015a; see also Alkire et al. 2013: 2–4 for a similar approximation method). We consider an individual to be deprived in the income domain if his or her daily consumption is less than the Asia-specific poverty line of $1.51 (ADB 2014). This cut-off is more stringent than both Indonesia’s national poverty line ($1.43) and the World Bank’s extreme poverty line ($1.25). Health status is assessed using two indicators: illness episode (number of days disabled within the last month) and morbidity (number of illnesses within the last month), which, admittedly, may not be as informative as the body mass index (BMI) indicator used in Alkire and Foster (2011a) and in Alkire and Santos (2014). However, these are the best available health measures in the Susenas survey, and since the disabling burden of poor health may lead not only to potentially missed income-generating opportunities (Schultz and Tansel 1997) but also, ultimately, to an unfulfilled life, we consider that these indicators make a good representation of the health domain. The inclusion of the illness episode indicator variable into a multidimensional poverty index is not new; such a measure has been used previously in a version of Bhutanese MPI (Santos 2013). In addition, the measure is often employed to operationalise Grossman’s model of health production function (Grossman 1972) in the health economics literature. An individual is deprived in health if he or she was ill for more than 4 days or caught more than 3 diseases within the last month (Table 1). These, we believe, are reasonable cut-offs considering the high prevalence (60–70 %) of informal self- and seasonal employment in Indonesia (Nazara 2010). Like health, education is also operationalised using two indicators: the completion of primary school (schooling) and the ability to read and write Latin characters (literacy). An individual is deprived if he or she has not completed primary education or is illiterate (Table 1). Relative to other indicators described above, primary schooling and literacy are perhaps the most universally accepted social indicators. They are highly valued worldwide: their presence in the HDI, MPI, MDGs and even the 1945 Constitution of the Republic of Indonesia testifies to this. Having chosen the social indicators to be included in the multidimensional poverty index, we now define their weights, which are necessary for identification purposes (Alkire 2011: 14–16). This weight assignment, which means making the trade-offs between social indicators explicit, clearly entails value judgements (Decancq and Lugo 2013). In this study, as in many other applications of the Alkire–Foster method, we use a normative weight because of the unavailability of preferences data. For the proposed Indonesian MPI, an equal-nested weighting scheme, which assigns an equal relative weight (13) to each dimension and also an equal weight to all indicators within a dimension, is used (weight w1 in Table 1). We then set the poverty cut-off to k=13 so that an income-poor individual is ‘automatically’ multidimensionally poor, but not the converse. This choice of parameters reflects the beliefs that (1) income, health and education are equally important for human development and (2) income still holds a special position in poverty measurement ‘given its fungibility and its key role in facilitating other capabilities’ (Foster 2007: 9). Notwithstanding the preference for this setting, we still conduct sensitivity analysis employing alternative weighting schemes (weights w2 and w3 in Table 1) and/or poverty cut-offs for k∈16,1.Table 2 Spearman correlation matrix of deprivations (2003–2013 maximum value; unweighted sample) Income Illness episode Morbidity Schooling Literacy Income 1.00 Illness episode 0.02 1.00 Morbidity 0.01 0.23 1.00 Schooling 0.14 0.10 0.06 1.00 Literacy 0.13 0.10 0.05 0.34 1.00 At this point, critics may contend that an index of multidimensional poverty is unnecessary because the social indicators included in its construct are, presumably, highly correlated either to income or to each other, representing a ‘double counting’. Our Indonesian data prove that this is not the case. As shown in Table 2, the correlation between income poverty and other deprivations in health and education during the last decade is never larger than 0.14; the figure among indicators of health and education is always less than 0.35. This mismatch between income poverty and deprivations in other social indicators conforms to the general finding in the literature (see Battison et al. 2013 on Latin American countries; Batana 2013 and Klasen 2000 on Africa; Brandolini and D’Alessio 1998 on Italy; Ranis and Stewart 2012 on Bangladesh, Chile, Indonesia, Kazakhstan, Laos and Zambia; Santos 2013 on Bhutan; Whelan et al. 2004 on Europe; Yu 2013 on China), providing ‘good empirical basis to support a multidimensional approach to poverty measurement, which goes beyond income and asset ownership’ (Santos 2013: 267). Results Unidimensional deprivations We begin by describing the trend of unidimensional deprivations in Indonesia during the 2003–2013 period. The top-left panel of Fig. 1 shows a strong 83 % income poverty reduction at the national level. Nearly half (46 %) of adult Indonesians were income poor in 2003, but a decade later, the figure improved significantly to just 8 %, registering an absolute 0.38 point reduction. It is apparent from the trend-line that income poverty reduction is characterised by two quinquennial regimes. Reduction was faster in the 2003–2008 period (65 %) than in the 2008–2013 period (50 %), a phenomenon that is consistent with the fact that the global economy was relatively more buoyant in the former period (high real growth rate and high commodity prices; Bourguignon et al. 2008: 12–13) than in the latter (the 2008 financial crisis and the ensuing drop in commodity prices; Battison et al. 2013: 308). At the same time, this discontinuity could reflect the differing efficacy in governance between the first (2004–2009) and the second United Indonesia Cabinets (2009–2014).Fig. 1 Income poverty The top-middle panel of Fig. 1 shows that income poverty reduction has been accompanied by a substantial improvement in the urban/rural disparity. The rural-to-urban poverty ratio fell sharply from 1.41 in 2003 to only 1.01 in 2013. The gender gap had not been of serious concern over the 11 years of observation, as the female-to-male poverty ratio has hardly ever deviated from the 1.00–1.01 range (the top-right panel of Fig. 1). However, we should keep in mind that this figure is obtained from household expenditure data rather than from individual income data. The trend in regional disparity seems to be similar to that of the urban/rural one. When we analyse each island-group separately (the middle to bottom panels of Fig. 1), the pattern of two-regime poverty reduction pattern still holds, while the between-island variance shrank by 90 % from 0.013 in 2003 to 0.001 in 2013, indicating a converging regional poverty. Despite this tremendous progress, it should be noted that Papua and Maluku, whose headcount ratio has remained constant at 1.5 times higher than the national average since 2008, seem to be left behind. Moreover, it is noticeable that Sulawesi has failed to register any significant improvement after 2008. This suggests that while Indonesia has enjoyed substantial progress in terms of sharply reduced income poverty and gradually converging urban/rural as well as regional disparities during the last decade, the East-West divide remains (see also poverty maps in Fig. 7).Fig. 2 Non-income deprivations Having described the state of income poverty, we now investigate trends in health and education (Fig. 2). Since non-income achievements are usually represented by stock rather than flow variables, and are therefore unlikely to change in the short run (Battison et al. 2013: 296), it is expected that their trend-lines will be relatively more stable than that of income. Some clear patterns emerge from the two plots in the top panel of Fig. 2, which displays the evolution of health deprivations over the decade. First, while the reduction of income poverty was at its fastest rate (2003–2008), the nation’s illness episode deprivation increased by 7 % year-on-year (13 % for morbidity) before eventually peaking in 2007 and then gradually returning to its initial level in 2010/2011. This inverted U-shape trend-line suggests that there may have been a short-term surge of negative health-related behaviour that followed the rising income level. In addition, it might also capture the health cost of both natural and man-made disasters (the Indian ocean earthquake and tsunami, the Java earthquake, the Sumatra flood and earthquake, the Sulawesi flood and landslide, and the Sidoarjo mud flow, to name only a few) that occurred relentlessly during the 2004–2007 period. Second, the trajectory of rural-to-urban health deprivation ratio also follows this inverted-U shape. The figure was about 1.20, 1.50 and 1.20 in 2003, 2007 and 2013, respectively, suggesting little to no improvement in terms of urban/rural health equality. The plots also present evidence regarding the disturbingly weak health status in Nusa Tenggara islands. Illness episode deprivation in Nusa Tenggara was 1.53–2.13 times greater than the national average and the figure for morbidity was in the range of 2.23–3.70 times greater. Moreover, it is evident that Nusa Tenggara exhibits a very distinctive trend-line compared to the rest of Indonesia, thereby exerting undue influence over the between-island variability. In contrast to the patterns of urban/rural and regional disparities, the female-to-male health deprivation ratio has always been stable in the range of 0.80–0.90, indicating that Indonesian women seem to be slightly healthier than their male counterparts (trend-lines greyed out). The two plots at the bottom panel of Fig. 2 show the trends of schooling and literacy deprivations. With the exception of two irregularly spiking literacy deprivations in 2003 and 2005, which appear to be a data quality problem, the trends seem to be stable over the window of observation. The rural-to-urban education deprivation ratio was constant in the range of 2.00–2.25 for both indicators. The female-to-male deprivation ratio was at about 1.20 for schooling and 1.80 for literacy; and the between-island variability has barely changed over the decade. It is evident that nearly a quarter of Indonesian adults living in rural areas failed to complete primary school, despite the substantial reduction in income poverty and the constitutional mandate for the provision of universal primary schooling. Again, by studying all four plots shown in Fig. 2, one can see immidiately that Nusa Tenggara islands are doubly burdened by both poor health and education outcomes. These findings demonstrate that Indonesia’s laudatory income poverty reduction over the last decade has not been complemented by equivalently strong improvements in non-income dimensions. The findings also suggest that different population subgroups (urban/rural, men/women, island-groups) performed differently in different dimensions of well-being. We now ask some follow-up questions. Taking those social indicators altogether, how many Indonesians are poor overall? Are urban areas always better off? Which island within the archipelago is the most deprived? What happened to gender and spatial inequalities? The marginal dashboard approach (Ravallion 2010, 2011) that we have just applied throughout this section is incapable of answering these questions because it only allows us to look at the marginal distribution of deprivations, while our inquiries demand a characterisation of the joint distribution of multiple deprivations (Alkire et al. 2011). In other words, in order to be able to answer these questions, the poverty measure has to take into account the extent of simultaneous deprivation experienced by individuals in the society. The multidimensional poverty measure to be reported next does just that. Multidimensional Poverty at the National Level Figure 3 presents the trend of multidimensional poverty at the national level. The top-left panel shows that overall poverty (M0) has declined at an annual rate of 14 % over the 2003–2013 period, owing much to the sharp reduction in the proportion of individuals identified as multidimensionally poor (poverty incidence, H), but less to the improvement in the average deprivations experienced by the poor (poverty intensity, A). In 2003, 48 % of Indonesian adults were multidimensionally poor and, collectively, they experienced about one-fifth (0.19) of the total possible deprivations that all adults could experience. A decade later, only 11 % of adults were poor: the overall poverty figure was just 0.04, indicating a substantial 78 % improvement. As in the case of income poverty, reduction in multidimensional poverty was also faster in the first six years (δM0=60%) than in the second five (δM0=44%). Despite of a steady 2 % annual decline in the contribution to overall poverty, income remains the main contributor to multidimensional poverty (the top-right panel of Fig. 3). Its contribution to overall poverty in 2013 was still about two times larger than its relative weight (64 %). On the other hand, health and education contributed less than their relative weights, suggesting that there were relatively low deprivations in these dimensions. It is noteworthy, however, that as time passes, the contribution of non-income dimensions to overall poverty increases steadily.Fig. 3 Multidimensional poverty at the national level Table 3 Correlation matrix of rank orderings across different weights Ranking pair† 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 National τB(w1,w2) 1* τB(w1,w3) 1* τB(w2,w3) 1* Urban/rural τB(w1,w2) 1 1 1 1 1 1 1 1 1 1 1 τB(w1,w3) 1 1 1 1 1 1 1 1 1 1 1 τB(w2,w3) 1 1 1 1 1 1 1 1 1 1 1 Gender τB(w1,w2) 1 1 1 1 1 −1 1 1 1 1 1 τB(w1,w3) 1 1 1 1 1 −1 1 1 1 1 1 τB(w2,w3) 1 1 1 1 1 1 1 1 1 1 1 Island τB(w1,w2) 1 0.87 1 0.87 1 0.87 0.73 0.87 0.73 0.73 1 τB(w1,w3) 1 1 0.87 0.87 0.87 0.87 0.87 1 0.47 0.73 0.87 τB(w2,w3) 1 0.87 0.87 1 0.87 1 0.87 0.87 0.73 1 0.87 † Reported are Kendall’s τB correlation coefficient for pairs of ranking while holding k=13 * Correlation coefficient for pairs of year ranking at the national level (2003–2013) The middle panel of Fig. 3 displays the results of poverty cut-off dominance analysis. On the left, we plot the estimated adjusted headcount ratio (M0) for the year 2003, 2008 and 2013, along with their analytical 95 % confidence intervals, against all possible poverty cut-offs spanning from the union (k=16) to the intersection (k=1) identification criterion. It turns out that the curves never cross, nor do their confidence intervals overlap, meaning that there was an unambiguous poverty reduction over the 2003–2013 period. On the right, we present similar analysis, plotting the M0 estimates for each poverty cut-off against the year. Results suggest that the shape of the trend-line is robust to any poverty cut-off for k∈16,12. Furthermore, the data reveal that even when alternative weights are specified, the shape of the downward-sloping trend (the bottom-left panel of Fig. 3) as well as the ordering of year ranking (the first five rows of Table 3) remain largely unaltered. In the bottom-right panel of Fig. 3, we calculate the extent of mismatch when targeting the poor using the conventional measure of income poverty versus using the proposed multidimensional poverty index (k=13). We found that the mismatch is about 3 % for the baseline weight (w1), which equals approximately 4.5 million adult Indonesians in 2013. The discrepancy increases significantly as we specified alternative weights that assign more importance to the schooling indicator (w2; 7–14 %), or to both schooling and illness episode indicators (w3; 9–17 %). This finding, in combination with that of the steadily increasing contribution of non-income deprivations to overall poverty (the top-right panel of Fig. 3), underlines the growing relevance of the multidimensional conceptualisation of poverty in Indonesia. Multidimensional Poverty Across Population Subgroups Having described the trend of multidimensional poverty at the national level, we now decompose the national MPI into contextually relevant subgroups, which, in Indonesia, means measuring poverty by urban/rural, gender and island-group separately. The goals of this exercise are to understand whether the pattern of poverty reduction has been uniform across subgroups and to identify any especially disadvantaged segment of Indonesian society.Fig. 4 Multidimensional poverty by urban/rural Figure 4 displays the trend of urban/rural poverty. In both urban and rural areas, multidimensional poverty has declined significantly over the 11 years of observation (the middle-left panel of Fig. 4). Progress was faster in rural than in urban areas (δM0=79 vs. 74 %), resulting in a progressively narrowing rural-to-urban poverty ratio (1.53 in 2003 vs. 1.25 in 2013). Two distinct patterns emerge as to how poverty reduction was achieved. As shown in the top panel of Fig. 4, poverty reduction in rural areas was driven by improvement in both poverty incidence (δH=77%) and intensity (δA=7.4%) whereas in urban areas, where amelioration in intensity was minuscule (δA=1.7%), poverty reduction was chiefly attributable to the diminishing proportion of individuals experiencing multiple deprivations (δH=74%). Results of poverty cut-off dominance analysis (the bottom four plots in Fig. 4) and a rank correlation test (Table 3) suggest that multidimensional poverty was unambiguously higher in rural than in urban areas for each year in the 2003–2013 period.Fig. 5 Multidimensional poverty by gender Next, we examine multidimensional poverty by gender. The data reveal that the female-to-male poverty ratio has never oscillated outside the 0.99 (2008) to 1.10 (2013) range (the middle-left panel of Fig. 5). In fact, the proportion of Indonesian women experiencing simultaneous deprivations over the last decade did not differ much compared to that of their male counterparts (the top-left panel of Fig. 5). Poverty intensity was slightly higher (1–4 %) among women than men during the 2003–2007 period, but since 2008, there has been hardly any difference with regard to the average deprivations experienced by the poor of both genders (the top-right panel of Fig. 5). A rank correlation test in Table 3 indicates that the gender ranking is generally robust to the choice of weight, but the results of poverty cut-off dominance analysis presented in the middle and bottom panels of Fig. 5 suggest that women are not unambiguously more deprived than men. This result, however, should be interpreted with caution because we still cannot disentangle the precise income (consumption) of men and women that live in the same household using household expenditure data.Fig. 6 Multidimensional poverty by island We now turn to investigating the trend of regional poverty (Fig. 6). In general, the last decade saw a substantial poverty reduction in all six island-groups that make up the Indonesian archipelago (the middle-left panel of Fig. 6). With the exception of Nusa Tenggara islands, improvement was principally attributable to diminishing poverty incidence, with only minimal progress in terms of poverty intensity (the top panel of Fig. 6). Poverty reduction was faster in the 2003–2008 period (δM0 = 55–67 %) than in 2008–2013 (δM0 = 36–60 %). The fastest progress was observed in Kalimantan (δ¯M0=18%), whereas the slowest was in Sulawesi (δ¯M0=12%). Robustness tests for the between-island comparison fail to yield clear-cut results. On the one hand, the rank correlation test in Table 3 shows that the ranking of regional poverty is relatively robust to the selection of weight. On the other hand, poverty cut-off dominance analysis presented in the middle to the bottom panels of Fig. 6 reveals that there are only limited dominances between the islands and over the 11-year period. Yet, within the limited scope of statistically meaningful comparisons that can be drawn, we can still at least deduce that from 2010 onwards, Papua and Maluku, along with Nusa Tenggara, have always been the poorest islands of the country, whereas Kalimantan is the least poor; between these two extremes are Sumatra, Java, Bali and Sulawesi, whose level of multidimensional poverty has always been close to the national average.Fig. 7 Incidence, intensity and adjusted headcount ratio at the district level (2013) Fig. 8 Within-island variation and districts’ incidence and intensity (2013) Table 4 The top 10 poorest and the least poor districts in 2013 Incidence Intensity Adjusted Headcount Ratio Island District H Island District A Island District M0 SM Bireun 0.384 SL Boalemo 0.425 SM Bireuen 0.142 PM Jayawijaya 0.348 SL Bantaeng 0.423 SM Aceh Singkil 0.136 PM Sarmi 0.345 SL Gorontalo 0.416 PM Sarmi 0.133 SM Simeulue 0.344 SL Kepulauan Sangihe 0.413 JB Banjarnegara 0.133 SM Aceh Singkil 0.343 KA Kapuas Hulu 0.413 PM Jayawijaya 0.133 JB Banjarnegara 0.335 KA Sambas 0.413 PM Pegunungan Bintang 0.126 PM Paniai 0.325 KA Kota Banjar Baru 0.412 SM Simeulue 0.122 PM Pegunungan Bintang 0.325 JB Blora 0.409 PM Biak Numfor 0.119 PM Biak Numfor 0.323 SM Kepulauan Mentawai 0.407 JB Temanggung 0.199 SM Kota Tanjung Balai 0.311 PM Halmahera Utara 0.406 PM Paniai 0.117 SM Karo 0.018 JB Kota Yogyakarta 0.345 SM Karo 0.006 JB Badung 0.017 PM Kota Sorong 0.345 JB Sidoarjo 0.006 JB Sidoarjo 0.016 KA Kota Samarinda 0.343 JB Badung 0.006 JB Kota Denpasar 0.013 SM Kota Banda Aceh 0.342 JB Kota Denpasar 0.005 SM Kota Banda Aceh 0.013 SM Kota Sabang 0.340 SM Kota Banda Aceh 0.004 KA Kota Samarinda 0.013 SM Kota Pematang Siantar 0.340 KA Kota Samarinda 0.004 KA Kota Balikpapan 0.011 JB Kota Depok 0.337 KA Kota Balikpapan 0.004 JB Kota Depok 0.011 SM Kota Tebing Tinggi 0.337 JB Kota Depok 0.004 PM Kota Ternate 0.009 PM Kota Ternate 0.333 KA Kota Bontang 0.003 KA Kota Bontang 0.008 KA Kota Balikpapan 0.333 PM Kota Ternate 0.003 Now, what happens if multidimensional poverty is measured at the lowest level of autonomous administrative areas instead of across island-groups? Figure 7 presents poverty maps displaying the extent of incidence, intensity, and overall poverty in 346 districts in Indonesia for the year 2013. It appears that the island ranking observed above still holds generally, but such comparison tends to conceal a large amount of variation between the districts on those islands. As we dig deeper (Table 4), it turns out that while five out of the ten poorest districts in 2013 are indeed located in Papua, three of them are in Sumatra (Aceh) and two are, perhaps surprisingly, in Central Java. There is unmistakable within-island variation as well (the left panel of Fig. 8). The island of Java, for example, is paradoxically home to one of the most (Banjarnegara, M0=0.133) and the least (Kota Depok, M0=0.004) deprived districts in Indonesia. Furthermore, even within a single province, variation can be immense. It is somewhat disconcerting to note that overall poverty in Kabupaten Bangkalan (M0=0.07) can be seven times higher than in Kota Surabaya (M0=0.01), even though they belong to the same East Java province and are separated by no more than a 90-min drive. The maps in Fig. 7 further reveal that the spatial patterning of poverty incidence tends not to match that of poverty intensity (Pearson’s ρ=0.40). In contrast to a cross-national pattern reported in the Human Development Report 2010 (UNDP 2010: 98), the incidence of poverty among districts in Indonesia does not seem to be linearly correlated with its intensity (the right panel of Fig. 8). Finally, also evident from this district-level analysis is the fact that urban areas tend to dominate rural areas. It is obvious that most of the top 10 least-deprived districts listed in Table 4 are municipalities (notice the Kota prefix) that, in general, have a higher level of urbanicity than ordinary districts (the Kabupaten). This is consistent with the result obtained earlier in Fig. 4. Inequality Among the Poor and Across Subgroups Table 5 Measures of inequality Inequality across… I(2003) I(2013) ΔI∗ Individual 0.098 0.076 0.022 Urban/rural 0.006 0.000 0.006 Gender 0.000 0.000 0.000 Island 0.003 0.000 0.003 Province 0.011 0.001 0.010 District 0.023 0.003 0.020 ∗ All changes are statistically significant at 5 % level Thus far, we have analysed the trend of multidimensional poverty in Indonesia by looking at the overall (M0) and partial indices (H, A) as well as by decomposing the indices into relevant geographical or social subgroups. We found that there was an unambiguous poverty reduction between 2003 and 2013, both nationally and sub-nationally. But, with such an improvement, questions of distribution arise. Did the progress benefit the poorest of the poor? Has poverty reduction over the last decade been shared uniformly across population subgroups that make up Indonesian society? In order to evaluate these, in Table 5 we calculate an inequality index (I) using the method proposed by Seth and Alkire (2014). This index is bounded between zero and one, capturing a state of complete equality up to that of complete inequality. It turns out that the reduction of multidimensional poverty in Indonesia within the last 11 years has been accompanied by an amelioration of the distribution of deprivations among the poor. The among-the-poor inequality decreased statistically significantly from 0.098 in 2003 to 0.076 in 2013, indicating inclusive progress. Similarly, the data show that there has been a convergence in poverty, meaning that poorer subgroups improved faster than the less poor. The disparity across subgroups has gone down for all relevant groupings (urban/rural, gender, island, province and district) over the 2003–2013 period. Finally, also evident from Table 5 is the fact that, in Indonesia, spatial inequality seems to matter more than gender or urban/rural inequality. Conclusion Applying the Alkire–Foster method of multidimensional poverty measurement to the National Socio-economic Survey (Susenas) data of Indonesia, this study estimates the extent and investigates the regional as well as the temporal patterns of multidimensional poverty in Indonesia from 2003 to 2013. An Indonesian version of multidimensional poverty index (MPI) is developed through an augmentation of the existing consumption poverty measure with information on health and education that are represented by indicators of illness episode, morbidity, completion of primary school, and literacy. It is found that, irrespective of the poverty cut-offs or weights specified, there was an unambiguous multidimensional poverty reduction over the last decade at both national and sub-national levels. About half (48 %) of Indonesian adults were multidimensionally poor in 2003 and, collectively, they experienced about one-fifth (0.19) of the total possible deprivations that the society could experience. In 2013, the situation was unmistakably better: only one in ten adults (11 %) was identified as multidimensionally poor, while the overall poverty figure fell to 0.04 (78 % reduction). The data suggest that the rate of poverty reduction was faster in the 2003–2008 period (60 %) than in 2008–2013 (44 %). With the exceptions of rural areas and the Nusa Tenggara islands, there was minimal improvement with regard to the average deprivations experienced by the poor (intensity); overall poverty reduction was driven mainly by the decline in poverty incidence. It is further found that, when the overall measure is broken down into its dimensional constituents, income deprivation remains the main contributor to multidimensional poverty (60–70 %), albeit with a 2 % rate of decrease annually. Also estimated in the national-level analysis is the mismatch between income and multidimensional poverty identification. Results show that approximately 3 % of adult Indonesians (4.5 million individuals in 2013) would be classified as non-poor if poverty identification did not take into account deprivations in health and education. This figure could be as high as 7–17 % (11–26 million), depending on how much importance is assigned to schooling and/or illness episode indicators. In an attempt to gain a more complete understanding of joint deprivation, the overall poverty measure is broken down by relevant population sub-groups. The data show that for each year from 2003 to 2013, multidimensional poverty was unambiguously higher in rural than in urban areas, but the gap between them has been progressively narrowing thanks to substantial improvement in both the incidence and the intensity of poverty in rural areas (rural-to-urban poverty ratio was 1.53 in 2003 vs. 1.25 in 2013). The data further reveal that Indonesian women are not unambiguously more deprived than men, although they appeared to have slightly more deprivations on average in the 2003–2007 period. Nevertheless, we cannot ascertain whether Indonesia has fared well in terms of gender equality because we cannot disentangle fully the information of women’s income using household expenditure data available at present. In contrast to the clear trend seen in urban/rural and gender decompositions, we found only faint dominance in between-island comparisons over the 11-year period. It is only from 2010 onwards that it can be asserted with statistical confidence that poverty is unambiguously higher in Papua, Maluku and Nusa Tenggara (or lower in Kalimantan) than anywhere else in the archipelago. Even so, it is still important to note that such between-island comparisons mask a substantial amount of within-island and between-district variations, echoing both Ilmma and Wai-Poi (2014) and Sumarto et al. (2014). While five out of the ten poorest districts in 2013 are indeed located in Papua, three of them are in Sumatra and the other two are in Java, neither of which are thought of as places with extreme poverty. Analysis at the district level further reveals that, departing from the pattern observed in a cross-national study (UNDP 2010: 98), the intensity of poverty among districts in Indonesia does not seem to be related in a linear way to its incidence. When the distribution of deprivations among the poor is studied, it is found that between 2003 and 2013, there were statistically significant improvements in terms of inequality among the poor and disparity across subgroups. The data show that poorer subgroups progress faster than the less poor, irrespective of the social or geographical groupings considered (converging subgroup poverty level). This finding indicates that the progress achieved within the last 11 years is relatively inclusive, although it should be noted that the between-district inequality within the Indonesian archipelago remains striking. Overall, these trends are comparable to those obtained from recent consumption poverty evaluations conducted by Ilmma and Wai-Poi (2014) and Sumarto et al. (2014), highlighting the fact that, even a decade after a ‘big-bang’ decentralisation (Hill 2014) was initiated, spatial inequity remains a serious challenge for Indonesia. It has been argued that the immense variation in poverty levels across districts reflects heterogeneity in the ‘capacity and resources of local governments to develop and implement poverty reduction strategies, and to quickly provide good public services’ (Sumarto et al. 2014: 310). Only competent local government can formulate sound development plans, allocate budgets efficiently, and deliver public services effectively. Therefore, there is plenty of room for local administrators to learn lessons from the top-performing districts (Maharani and Tampubolon 2014). While this study has presented a thorough investigation into the state of multidimensional poverty in Indonesia over the last decade, it is inevitably bound by several limitations. Firstly, the present study is unable to include children and adolescents younger than 18 years old in the analysis because information on the relevant dimensions of their well-being (Trani et al. 2013) are not available in Susenas survey. Secondly, the health indicators used in this study (illness episode and morbidity) are weak and by no means comparable to the indicators stipulated in the Millennium Development Goals (malnutrition). Thirdly, with the absence of preference data obtained from large-scale participatory study, the trade-offs between social indicators used in this study are entirely normative. In addition, although the measurement of chronic multidimensional poverty under the Alkire–Foster methodology has recently become feasible (Alkire et al. 2014), this study was unable to make use of it due to the cross-sectional nature of Susenas survey. It is indeed indisputable that future poverty evaluations would benefit from the availability of more comprehensive micro-data. Even with these limitations, the study still contributes to the literature in at least three ways. First, using nationally representative survey data from Indonesia, the present study shows that the conventional measure of income poverty is not comprehensive. The Indonesian data reveal that income poverty only weakly correlates with deprivations in the domains of health and education, confirming the findings documented in other Asian (Ranis and Stewart 2012; Santos 2013; Yu 2013), African (Batana 2013; Klasen 2000), European (Brandolini and D’Alessio 1998; Whelan et al. 2004) and Latin American (Battison et al. 2013) countries. This may motivate future assessment of multidimensional poverty in other parts of the world. Second, in using consumption expenditure data as the indicator of income, this study allows the poverty measure to become more sensitive to economic fluctuations than the current version of the international MPI (UNDP 2010), which uses asset ownership as a proxy for deprivation in living standards. This not only addresses one of the criticisms of the MPI (Ravallion 2010: 11), but also makes the MPI more comprehensible to Indonesians, who have for decades been accustomed to the conceptualisation of poverty as a consumption shortfall in essential goods and services. Finally and most importantly, the present study demonstrates the feasibility of adapting the Alkire–Foster methodology to the Indonesian context using an existing official data source that has been in production since the 1960s (Surbakti 1995). Because the data are readily available, and the proposed multidimensional poverty measure presented here makes identification of multiply-deprived Indonesians possible, the MPI could nicely complement the existing indices that are routinely reported by the Indonesian Statistical Bureau (BPS). The new measure is suitable as a tool for monitoring the progress of national development, and could also be used as a device for prioritising investment projects or other forms of intervention that are funded by transfers from central to local governments (see Salazar et al. 2013 for a recent proposal in Colombia). With the demonstrated novelty, feasibility and utility of the Alkire–Foster method, policy makers should now more than ever want to incorporate the idea of poverty as an experience of multiple deprivations into the discourse of national development. The authors thank Suman Seth of the Oxford Poverty & Human Development Initiative (OPHI) for sharing computer code used in the computation of the inequality index. The authors also thank the anonymous referees for their helpful comments. All errors remain the authors’ responsibility. Wulung Anggara Hanandita’s research is supported by the Economic and Social Research Council, UK [ES/J500094/1] (PhD project: ‘Essays on Poverty and Health in Indonesia’). ==== Refs References ADB. (2014). Key indicators for Asia and the Pacific 2014. Asian Development Bank, Manila, Philippines, Ch. Poverty in Asia: A Deeper Look. Alkire, S. (2011). Multidimensional poverty and its discontents. Oxford Poverty & Human Development Initiative (OPHI) Working Paper No. 46. http://www.ophi.org.uk/multidimensional-poverty-and-its-discontents/ Alkire, S., Apablaza, M., Chakravarty, S. R., & Yalonetzky, G. (2014). Measuring chronic multidimensional poverty: A counting approach. Oxford Poverty & Human Development Initiative (OPHI) Working Paper No. 75. http://www.ophi.org.uk/measuring-chronic-multidimensional-poverty-a-counting-approach/ Alkire S Foster J Counting and multidimensional poverty measurement Journal of Public Economics 2011 95 7–8 476 487 10.1016/j.jpubeco.2010.11.006 Alkire S Foster J Understandings and misunderstandings of multidimensional poverty measurement Journal of Economic Inequality 2011 9 2 289 314 10.1007/s10888-011-9181-4 Alkire S Foster J Santos ME Where did identification go? Journal of Economic Inequality 2011 9 3 501 505 10.1007/s10888-011-9201-4 Alkire, S., Roché, J. M., & Summer, A. (2013). Where do the world’s multidimensionally poor people live? Oxford Poverty & Human Development Initiative (OPHI) Working Paper No. 61. http://www.ophi.org.uk/where-do-the-worlds-multidimensionally-poor-people-live/ Alkire S Santos ME A multidimensional approach: Poverty measurement & beyond Social Indicators Research 2013 112 2 239 257 10.1007/s11205-013-0257-3 Alkire S Santos ME Measuring acute poverty in the developing world: Robustness and scope of the multidimensional poverty index World Development 2014 59 251 274 10.1016/j.worlddev.2014.01.026 Alkire, S., & Vaz, A. (2014). Multidimensional poverty dynamics with time series and panel data. OPHI Summer School on Multidimensional Poverty Analysis. http://www.ophi.org.uk/wp-content/uploads/3-SS14-Analysis_Over_Time.pdf Atkinson AB Multidimensional deprivation: Contrasting social welfare and counting approaches Journal of Economic Inequality 2003 1 1 51 65 10.1023/A:1023903525276 Batana YM Multidimensional measurement of poverty among women in Sub-Saharan Africa Social Indicators Research 2013 112 2 337 362 10.1007/s11205-013-0251-9 Battison D Cruces G Lopez-Calva LF Lugo MA Santos ME Income and beyond: Multidimensional poverty in six Latin American countries Social Indicators Research 2013 112 2 291 314 10.1007/s11205-013-0249-3 Bourguignon, F., Bénassy-Quéré, A., Dercon, S., Estache, A., Gunning, J. W., Kanbur, R., et al. (2008). Millennium development goals at midpoint: Where do we stand and where do we need to go? European Commission Report. http://ec.europa.eu/development/icenter/repository/mdg_paper_final_20080916_en.pdf Bourguignon F Chakravarty SR The measurement of multidimensional poverty Journal of Economic Inequality 2003 1 1 25 49 10.1023/A:1023913831342 BPS. (2015a). Jumlah dan Persentase Penduduk Miskin, Garis Kemiskinan, Indeks Kedalaman Kemiskinan (P1), dan Indeks Keparahan Kemiskinan (P2) Menurut Provinsi, 2007–2009 (Maret), 2010–2011, 2012 (Maret dan September). http://www.bps.go.id/webbeta/frontend/linkTabelStatis/view/id/1489 BPS. (2015b). Konsep Indeks Pembangunan Manusia (online). Accessed February 27, 2015. http://www.bps.go.id/webbeta/frontend/Subjek/view/id/26#subjekViewTab1 BPS. (2015c). Konsep Kemiskinan (online). Accessed February 27, 2015. http://www.bps.go.id/webbeta/frontend/Subjek/view/id/23#subjekViewTab1 BPS, Bappenas, UNDP. (2004). Ekonomi dari Demokrasi: Membiayai Pembangunan Manusia Indonesia. Laporan Pembangunan Manusia Indonesia 2004. Brandolini, A., & D’Alessio, G. (1998). Measuring well-being in the functioning space. Research note. http://www.iberoamericana.edu.mx/humanismocristiano/seminario_capability/pdf/3.pdf Decancq K Lugo MA Weights in multidimensional indices of wellbeing: An overview Econometric Reviews 2013 32 1 7 34 10.1080/07474938.2012.690641 Filmer D Pritchett LH Estimating wealth effects without expenditure data—or tears: An application to educational enrollments in states of India Demography 2001 38 1 115 132 11227840 Foster, J. (2007). A report on Mexican multidimensional poverty measurement. Oxford Poverty & Human Development Initiative (OPHI) Working Paper No. 40. http://www.ophi.org.uk/a-report-on-mexican-multidimensional-poverty-measurement/ Foster J Greer J Thorbecke E A class of decomposable poverty measures Econometrica 1984 52 3 761 766 10.2307/1913475 Grossman M On the concept of health capital and the demand for health Journal of Political Economy 1972 80 2 223 255 10.1086/259880 Hill H Regional dynamics in a decentralized Indonesia 2014 Singapore ISEAS Publishing ICF International. (2012). Demographic and health surveys (online). Accessed February 27, 2015. http://dhsprogram.com/data/ Ilmma A Wai-Poi M Hill H Patterns of regional poverty in the new Indonesia Regional dynamics in a decentralized Indonesia 2014 Singapore ISEAS Publishing 98 132 Klasen S Measuring poverty and deprivation in South Africa Review of Income and Wealth 2000 46 1 33 58 10.1111/j.1475-4991.2000.tb00390.x Maharani A Tampubolon G Has decentralisation affected child immunisation status in Indonesia? Global Health Action 2014 7 24913 10.3402/gha.v7.24913 25160515 MPR RI. (2011). Undang–Undang Dasar Negara Republik Indonesia Tahun 1945. Majelis Permusyawaratan Rakyat Republik Indonesia. http://www.mpr.go.id/pages/produk-mpr/uud-nri-tahun-1945 Nazara, S. (2010). The informal economy in Indonesia: Size, composition and evolution. The International Labour Organization Working Paper. http://www.ilo.org/jakarta/whatwedo/publications/WCMS_145781/lang--en/index.htm RAND. (2007). IFLS-4 survey description (online). Accessed May 1, 2014. http://www.rand.org/labor/FLS/IFLS/ifls4.html RAND. (2014). Indonesia Family Life Survey East (IFLS EAST) (online). Accessed February 27, 2015. http://www.rand.org/labor/FLS/IFLS/ifls-east.html Ranis G Stewart F Success and failure in human development, 1970–2007 Journal of Human Development and Capabilities 2012 13 2 167 195 10.1080/19452829.2011.645026 Ravallion, M. (2010). Mashup indices of development. World Bank Policy Research Working Paper No. 5432. http://elibrary.worldbank.org/doi/abs/10.1596/1813-9450-5432 Ravallion M On multidimensional indices of poverty Journal of Economic Inequality 2011 9 2 235 248 10.1007/s10888-011-9173-4 Ravallion M Chen S Sangraula P Dollar a day revisited World Bank Economic Review 2009 23 2 163 184 10.1093/wber/lhp007 Roche JM Monitoring progress in child poverty reduction: Methodological insights and illustration to the case study of Bangladesh Social Indicators Research 2013 112 2 363 390 10.1007/s11205-013-0252-8 Sahn DE Stifel D Exploring alternative measures of welfare in the absence of expenditure data Review of Income and Wealth 2003 49 4 463 489 10.1111/j.0034-6586.2003.00100.x Salazar, R. C. A., Diaz, B. Y., & Pinzon, R. P. (2013). A counting multidimensional poverty index in public policy context: The case of Colombia. Oxford Poverty & Human Development Initiative (OPHI) Working Paper No. 62. http://www.ophi.org.uk/a-counting-multidimensional-poverty-index-in-public-policy-context-the-case-of-colombia/ Santos ME Tracking poverty reduction in Bhutan: Income deprivation alongside deprivation in other sources of happiness Social Indicators Research 2013 112 2 259 290 10.1007/s11205-013-0248-4 Santos, M. E., & Ura, K. (2008). Multidimensional poverty in Bhutan: Estimates and policy implications. Oxford Poverty & Human Development Initiative (OPHI) Working Paper No. 14. http://www.ophi.org.uk/working-paper-number-14/ Schultz PT Tansel A Wage and labor supply effects of illness in Côte d’Ivoire and Ghana: Instrumental variable estimates for days disabled Journal of Development Economics 1997 53 2 251 286 10.1016/S0304-3878(97)00025-4 Sen A Poverty: An ordinal approach to measurement Econometrica 1976 44 2 219 231 10.2307/1912718 Sen A Commodities and capabilities 1985 New York Elsevier Seth, S., & Alkire, S. (2014). Did poverty reduction reach the poorest of the poor? Assessment methods in the counting approach. Oxford Poverty & Human Development Initiative (OPHI) Working Paper No. 77. http://www.ophi.org.uk/did-poverty-reduction-reach-the-poorest-of-the-poor-assessment-methods-in-the-counting-approach/ Strauss J Beegle K Dwiyanto A Herawati Y Pattinasarany D Satriawan E Sikoki B Sukamdi Witoelar F Indonesian living standards before and after the financial crisis: Evidence from the Indonesia Family Life Survey 2004 Singapore ISEAS Publishing Sumarto S Vothknecht M Wijaya L Hill H Explaining regional heterogeneity of poverty: Evidence from a decentralized Indonesia Regional dynamics in a decentralized Indonesia 2014 Singapore ISEAS Publishing 285 314 Surbakti P Indonesia’s National Socio-economic Survey: A continual data source for analysis on welfare development 1995 Jakarta Badan Pusat Statistik Székely, M. (2003). Lo que dicen los pobres. Cuadernos de desarrollo humano No. 13. Secretaría de Desarrollo Social, México. Thomas D Witoelar F Frankenberg E Sikoki B Strauss J Sumantri C Suriastini W Cutting the costs of attrition: Results from the Indonesia Family Life Survey Journal of Development Economics 2012 98 1 108 123 10.1016/j.jdeveco.2010.08.015 25705070 Townsend, P. (1979). Poverty in the United Kingdom. London: Allen Lane and Penguin Books. http://www.poverty.ac.uk/free-resources-books/poverty-united-kingdom Trani J-F Biggeri M Mauro V The multidimensionality of child poverty: Evidence from Afghanistan Social Indicators Research 2013 112 2 391 416 10.1007/s11205-013-0253-7 UNDP The real wealth of nations: Pathways to human development 2010 New York Macmillan UNDP. (2014). Why is the MPI better than the Human Poverty Index (HPI) which was previously used in the Human Development Reports? Frequently Asked Questions, United Nations Development Programme (online). Accessed February 27, 2015. http://hdr.undp.org/en/faq-page/multidimensional-poverty-index-mpi#t295n138 UNSD. (2014). Purchasing power parities (PPP) conversion factor, local currency unit to international dollar. United Nations Statistics Division. http://mdgs.un.org/unsd/mdg/SeriesDetail.aspx?srid=699 Ura, K., Alkire, S., Zangmo, T., & Wangdi, K. (2012). An extensive analysis of GNH index. Centre for Bhutan Studies. http://www.grossnationalhappiness.com/wp-content/uploads/2012/10/An%20Extensive%20Analysis%20of%20GNH%20Index.pdf van de Walle, D. (1988). On the use of the susenas for modelling consumer behaviour. Bulletin of Indonesian Economic Studies, 24(2), 107–121. Whelan CT Layte R Maître B Understanding the mismatch between income poverty and deprivation: A dynamic comparative analysis European Sociological Review 2004 20 4 287 302 10.1093/esr/jch029 Yu J Multidimensional poverty in China: Findings based on the CHNS Social Indicators Research 2013 112 2 315 336 10.1007/s11205-013-0250-x

PMC005xxxxxx/PMC5002047.txt

==== Front Asian J Gambl Issues Public HealthAsian J Gambl Issues Public HealthAsian Journal of Gambling Issues and Public Health2195-3007Springer Berlin Heidelberg Berlin/Heidelberg 1810.1186/s40405-016-0018-8Research ArticleRole of smartphone addiction in gambling passion and schoolwork engagement: a Dualistic Model of Passion approach Enwereuzor Ibeawuchi K. gr8kice2@gmail.com 1Ugwu Leonard I. leonard.ugwu@unn.edu.ng 1Ugwu Dorothy I. dorothy.ugwu@unn.edu.ng 21 Department of Psychology, University of Nigeria, Nsukka, Nigeria 2 Department of Health and Physical Education, University of Nigeria, Nsukka, Nigeria 26 8 2016 26 8 2016 2016 6 1 98 5 2016 18 8 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.There are growing concerns that seem to suggest that students no longer engage in school-related activities as they ought to. Recent observation has revealed that students now spend excessive time participating in Internet gambling with their smartphone during school period. This trend could have far-reaching consequences on their schoolwork engagement and by extension, academic performance. Drawing on the Dualistic Model of Passion, this study therefore, examined the mediatory role of smartphone addiction in the gambling passion—schoolwork engagement relation. A cross-sectional design was adopted. Male undergraduates (N = 278) of a large public university in Nigeria who engage in Internet gambling participated in the study. They completed self-report measures of gambling passion, smartphone addiction, and schoolwork engagement. Results showed that harmonious gambling passion was not related to smartphone addiction whereas it was positively related to schoolwork engagement. Obsessive gambling passion had positive and negative relations with smartphone addiction and schoolwork engagement, respectively. Smartphone addiction was negatively related to schoolwork engagement and mediated only the obsessive gambling passion—schoolwork engagement relation but not that between harmonious gambling passion and schoolwork engagement. The theoretical and practical implications of the findings are discussed. Keywords Gambling passionHarmonious gambling passionObsessive gambling passionSmartphone addictionSchoolwork engagementissue-copyright-statement© The Author(s) 2016 ==== Body Background Schools represent contexts where students attend classes, work on projects and assignments, and study (Salanova et al. 2010; Salmela-Aro and Upadaya 2012). They are also places where students strive toward accomplishing specific goals such as completion of course work, academic performance, acquiring a degree and the like (Siu et al. 2014). However, there are indications that students’ interest in their schoolwork appears to be waning (Appleton et al. 2008). Insinuations suggest that the chances of students disengaging from their schoolwork increases as they progress from lower to higher levels of their education (Marion et al. 2014; Siu et al. 2014). If students are however fully engaged in their schoolwork, it could portend positive outcomes for them. Conversely, if they are not, their academic achievement could be undermined. For instance, several studies have linked increased levels of schoolwork engagement to positive outcomes such as increased performance (e.g., Bakker et al. 2015; Salanova et al. 2010; Schaufeli et al. 2002) whereas other studies linked lower levels of schoolwork engagement to a range of negative outcomes including increased school burnout and depressive symptoms (e.g., Marion et al. 2014; Salmela-Aro et al. 2009; Salmela-Aro and Upadaya 2012). Taken together, these studies are indicative of the importance of engaging in schoolwork in the lives of students. In essence, schoolwork engagement refers to energy, dedication toward, and absorption in schoolwork (Salmela-Aro and Upadaya 2012). Energy herein is characterized by positive approach to schoolwork. Dedication refers to a strong positive cognitive attitude and perceiving schoolwork as worthwhile. Absorption is characterized by being fully concentrated on studying, whereby time seems to pass quickly. In the light of the apparent importance of schoolwork engagement, it would seem important to consider the factors that could impact on schoolwork engagement as well as the psychological processes that may underpin such impact. In that sense, one possible antecedent that is purported, in the current study, to have an impact on undergraduates’ schoolwork engagement is gambling passion. This follows our recent observation that during class and off class periods, male undergraduates in particular, are seen engaging in Internet football betting/gambling. A similar observation was made by Lo et al. (2005) who decried that students now spend substantial amount of time on regular basis playing video games in cybercafés rather than engaging in school-related activities. Consistent with our observation, empirical evidence reveals that this trend is particularly common among males than females. For instance, Husky et al. (2015) recently found that women were less likely to engage in multiple gambling activities in comparison to men, while men are three times more likely to experience problems with gambling as compared to women. Tsitsika et al. (2011), reported that more males engage in gambling than females. For instance, a large cross-cultural study conducted by Peltzer and Pengpid (2014) involving 17,789 university students from 23 countries across Africa, Asia and Americas, showed that in Nigeria (n = 424), 13.8 % of the students reported gambling less than once a week whereas 6.7 % reported gambling once a week or more. In terms of gender, 10.1 % of men reported gambling once a week or more as compared to only 2.7 % of women who reported doing so. Across the countries, 4.6 % of men reported betting on sports once a week or more as compared to only 0.9 % of women who reported doing so. Furthermore, studies have also linked gambling and gambling passion to detrimental health consequences among undergraduates (e.g., Skitch and Hodgins 2005; Stuhldreher et al. 2007). Moreover, a 75 % rate of gambling participation has been reported among university students (e.g., Barnes et al. 2010). These trends appear to highlight the need to study gambling passion among male university students. Indeed, as Vallerand and colleagues posited (Vallerand et al. 2007), “… being passionate for an activity leads individuals to dedicate themselves fully to their activity, thereby allowing them to persist, even in the face of obstacles, and to eventually reach excellence” (p. 506). Even though passion may engender dedication toward the activity in question and, in the long run, performance (Vallerand et al. 2007), it is possible that it may also enhance or undermine schoolwork engagement depending on the particular type of gambling passion (i.e., harmonious or obsessive gambling passion) that is adopted. Thus, given the preceding evidence that seem to suggest that gambling passion could have far-reaching consequences on undergraduates’ schoolwork engagement, it is however surprising to note that we have not come across any published studies that have examined the relationship between gambling passion and schoolwork engagement. Moreover, our recent observation revealed that undergraduates who gamble on the Internet sometimes do so using their smartphones. Similarly, previous research has also indicated that problem gamblers are more likely to use a cell phone to gamble on the Internet than social gamblers (e.g., McBride and Derevensky 2009). In that vein, it is possible for smartphone addiction to mediate the influence of gambling passion on schoolwork engagement. In other words, having passion for Internet football betting/gambling may lead to smartphone addiction due to attempt at keeping track of ongoing/live football gambling on the Internet, which in turn may affect the schoolwork engagement of undergraduates. Again, heretofore, no published studies to our knowledge have examined the link between smartphone addiction and schoolwork engagement. Several disparate studies have been conducted on schoolwork engagement under different but related appellations including but not limited to study engagement, student engagement and the like (e.g., Bakker et al. 2015; Marion et al. 2014; Salmela-Aro et al. 2009; Salmela-Aro and Upadaya 2012; Schaufeli et al. 2002). The literature is also replete with studies on Internet gambling, problem gambling, pathological gambling, and gambling passion (e.g., Barrault and Varescon 2013; Li et al. 2012; Ratelle et al. 2004; Tsitsika et al. 2011), as well as smartphone addiction (e.g., Al-Barashdi et al. 2015; Kwon et al. 2013a, b; Mok et al. 2014). However, to our knowledge, no published research has examined gambling passion as an antecedent of schoolwork engagement and the potential mediatory role of smartphone addiction in this relation. That is, there has not been any synthesis of these constructs within the same study. This lack of research synthesis impedes our understanding of how gambling passion and smartphone addiction can affect schoolwork engagement among undergraduates. Nonetheless, it seems probable that undergraduates who are passionate about football gambling may become addicted to their smartphone in order to sustain their passion for the gambling online, which in turn, may lead to lower levels of engagement in their schoolwork as well. Gambling and smartphone addiction: the Nigerian scenario In Nigeria, the National Lottery Regulatory Commission (NLRC) is the body saddled with the responsibility of regulating lottery/gambling operations in order to promote transparency and accountability as well as protecting the interests of players, stakeholders and the public at large. It was established by the National Lottery Act (2005). Legally permitted lottery/gambling operations by the commission are: charitable lottery, lottery concierge services, online lottery, promotional lottery, short-message service (SMS) lotteries, sports lottery, USSD lotteries, and others. Thus, Internet football betting/gambling, which is the focus of the present study, by default falls under sports lottery (betting). The NLRC referred to sports betting as any activity involving prediction of sporting results and placing a bet on its outcome in anticipation of winning a set prize. All organizations currently operating the business of sports lottery (betting) and intending ones in Nigeria are required to obtain a Sports Lottery Permit (SLP) from the commission. The Sport Lottery Operator (SLO) must ensure that only those who are 18 years old and above are allowed to wager (for details, see the National Lottery Act 2005). Sports betting centres are located across the landscape of both remote and major cities in Nigeria. In most instances, such centres also serve as match viewing centres, where football fans come to watch their favourite football teams play. Bettors can bet with as low as even ₦100 in shops, over the phone and even online. All these indicate that Internet football betting/gambling is thriving in Nigeria. In addition, Nigeria is part of the global users of smartphone. Globally, it has been reported that Nigeria is ranked 17th among countries who like using smartphones (Ekpeke 2015). In a recent submission, eMarketer (2015), a digital market analytical platform, estimated that the number of smartphone users in Nigeria will increase from 18.7 % in 2015 to 27.6 % by 2019. Furthermore, a recent study involving science students from privately owned universities in Nigeria revealed that majority of the students (83.7 %) enjoyed using their smartphones as compared to other mobile devices (Fasae and Adegbilero-Iwari 2015). In another Nigerian study, 97 % of undergraduates reported that they own a mobile phone (Okafor and Malizu 2014). Our observation also reveals that most students are seen on campus with their smartphones clutched at their hands. Taken together, these indicate that undergraduates in Nigeria are active users of smartphone. The present study We aim to bridge the gaps in the literature by investigating smartphone addiction as the psychological process that could help explain the probable differential impact that the two types of gambling passion (i.e., harmonious and obsessive gambling passion) could have on schoolwork engagement. A theoretical model is proposed (see Fig. 1) based on the Dualistic Model of Passion (Vallerand et al. 2003) such that harmonious gambling passion is expected to be negatively related to smartphone addiction, whereas obsessive gambling passion is expected to be positively related to it. Smartphone addiction should in turn be associated with poorer schoolwork engagement. Thus, this study has the potential to contribute to the emerging schoolwork engagement literature by investigating gambling passion as a previously unexplored antecedent of schoolwork engagement as well as smartphone addiction as a mediating mechanism that could help account for this relationship. At the same time, it could aid in designing assessment and intervention strategies that will be useful in uncovering factors that could affect students’ schoolwork engagement as well as ways of improving such engagement.Fig. 1 Proposed theoretical model of the relationship between gambling passion, smartphone addiction and schoolwork engagement Theoretical background and hypotheses development Vallerand et al. (2003) defined passion as “a strong inclination toward an activity that people like, that they find important, and in which they invest time and energy” (p. 757). In their Dualistic Model of Passion, Vallerand et al. (2003) posit that passion represents a major motivational force behind participation in enjoyable activities by providing people with the required energy to engage in pleasurable activity. It is this pleasurable activity that is expected to lead to different outcomes. The Dualistic Model of Passion can aid in the understanding of gamblers’ motivation in participating in gambling activities. According to this model, being passionate about an activity can be so self-defining such that it embodies key characteristics on the person’s identity and thus used as criteria for defining the person. For instance, individuals who are passionate about gambling do not merely gamble; they are ‘gamblers’. In other words, passionate gambling has become internalized into their identity. The Dualistic Model of Passion (Vallerand et al. 2003) distinguishes between two types of passion that could develop based on the type of internalization process that takes place (Deci and Ryan 2000; Ryan and Deci 2000). They are: harmonious and obsessive passion. Harmonious passion stems from autonomous internalization of an activity into a person’s identity which occurs when a person has willingly accepted the activity as important and without any external rewards attached to it (Vallerand et al. 2003). There is no conflict between the enjoyable activity and other domains of the person’s life. Although the passionate activity occupies a considerable, but not overriding space in the person’s life, such that the person still remains in control while engaging in the enjoyable activity and is in harmony with other areas of the person’s life (Vallerand et al. 2003). In this respect, the Dualistic Model of Passion would suggest that undergraduates with harmonious passion for gambling, even though they are passionate about the gambling, are still in control of the gambling and thus, will not allow the gambling to override other domains of their lives such as schoolwork engagement. While the football match is going on, gamblers are usually presented with different options or odds of winning on the Internet. Thus, given our recent observation and empirical evidence (e.g., Fasae and Adegbilero-Iwari 2015) which indicate that most undergraduates now have smartphones in Nigeria, one way that those who engage in Internet football gambling among them now choose to follow-up this gambling is to use their smartphones. This affords them the opportunity to keep abreast with the best possible options that could enhance their chances of winning. Consistent with the Dualistic Model of Passion (Vallerand et al. 2003), even though undergraduates with harmonious passion toward gambling may choose to use their smartphones for Internet gambling purposes, they are unlikely to become addicted to it at the detriment of engaging in their schoolwork. This is because they are in control of the gambling situation and gamble primarily for autonomous reasons (e.g., fun) and not necessarily because of external benefits (e.g., monetary reward). Hence, the Dualistic Model of Passion would suggest that undergraduates with harmonious passion for Internet football gambling are less likely to become addicted to their smartphone and in turn should engage more in their schoolwork. In contrast to harmonious passion, obsessive passion originates from a controlled internalization of an enjoyable activity into the person’s identity. Intra- and/or interpersonal pressure is said to underlie such controlled internalization because there are certain benefits derived from the passionate activity, such as feeling socially accepted or worthy, or because the pleasure derived from the activity can no longer be controlled by the individual (Vallerand et al. 2007). Consequently, the passionate activity may become too valued, to be giving preference over all other activities in the person’s life, and to occupy so much space in the person’s identity such that it interferes with other domains in the person’s life (Vallerand et al. 2003, 2007). Such compelled engagement should prevent an undergraduate with an obsessive passion toward gambling from completely focusing on the task at hand, and may interfere with other domains of their life such as schoolwork engagement. Moreover, because of the nature of the gambling that is being done on the Internet, it is expected that undergraduates who are obsessed about Internet gambling will use their smartphones to access the gambling in order to stay abreast of the gambling. Eventually, it may get to a point when they become addicted to their smartphones given that it was internalized into their identity based on controlled or external contingencies (e.g., monetary gain). In turn, they may experience psychological dependence or addiction to their smartphones which is likely to impinge on their schoolwork engagement. Figure 1 depicts our proposed theoretical model. The dashed lines and the signs are used to represent hypothesized relations such that harmonious and obsessive gambling passions are expected to have respective negative and positive relations with smartphone addiction. Smartphone addiction in turn is expected to have a negative relation with schoolwork engagement. Harmonious and obsessive gambling passions are also expected to have respective positive and negative relations with schoolwork engagement. Gambling passion and smartphone addiction Recent advances in Information and Communication Technology (ICT) have come to bear on the way modern smartphones are designed to function. Smartphones are designed nowadays to suit our personal lifestyles and has revolutionized the way we communicate, do business, have fun, conduct research, and even work. Given the convenience, attractive features, and broad range of functions that can be performed with smartphones, many people are keen on acquiring them (Kwon et al. 2013a, b). For example, as at 2010, the number of smartphones purchased across the globe rose to 130 million and it was further projected that mobile devices will be the basic gadget for connecting to the Internet before 2020 in place of other traditional means of doing so (MacCormick et al. 2012). Hence, because smartphones are Internet-enabled, it is possible for its users to become addicted to them which may trigger physical and social problems just like those associated with Internet addiction (Shaw and Black 2008). Drawing from the definition of Internet addiction offered by Weinstein and Lejoyeux (2010), we define smartphone addiction as the excessive or irresistible compulsion regarding smartphone use that leads to impairment in psychological and social functioning. The smartphone “addicts may use … [their smartphones] … for extended periods, isolating themselves from other forms of social contact, and focus almost entirely on the[ir] … [smartphones] rather than broader life events” (Weinstein and Lejoyeux 2010, p. 277). Smartphone addiction is a form of technological addiction (Lee et al. 2014) and as such is regarded as a non-chemical behavioural addiction involving human–machine interaction (Griffiths 1996, 2000). It is characterised by increasing usage tolerances, withdrawal symptoms, salience, mood changes, conflict and relapse (Griffiths 1996, 2000). In other words, smartphone addicts can be said to be unable to do without their smartphone and may find it extremely difficult to function without it. As such, it is possible that they may become depressed and isolated when they abstain from using their smartphone. They may also be mostly preoccupied with smartphone use by devoting so much time than is required to its use rather than other life activities. What’s more, given the constant evolving of the gambling industry in terms of accessibility, affordability, availability, anonymity, convenience, endorsement, glamorization, and promotion through the Internet, many youths have taken to Internet gambling (Griffiths 2003; Messerlian et al. 2004; Monaghan et al. 2008). Such Internet gambling practices have been linked to a range of adverse health consequences for individuals who indulge in them (Griffiths et al. 2009; Jacobs 2004; Lloyd et al. 2010). This is because they are at risk of becoming addicted (Pallanti et al. 2006) even though they often see themselves as invulnerable and deny the probable negative outcomes of gambling (Derevensky et al. 2003). Hence, undergraduates constitute a potential group that could develop addictive behavioural patterns, especially smartphone addiction because they make the most use of smartphone services (Head and Ziolkowski 2012) especially in respect to surfing the Net and exploring applications which provide thrilling functions (Al-Barashdi et al. 2015). Besides, recent research (e.g., Mok et al. 2014) has reported smartphone addiction among university students which appear to corroborate our recent observation that revealed that university students use smartphone to engage in Internet football gambling. If that be the case, are all undergraduates who are passionate about Internet football gambling and use their smartphone to take part in it likely to become addicted to their smartphones? Prior studies (e.g., Mageau et al. 2005; Rousseau et al. 2002; Vallerand et al. 2003, 2007) suggest that it should be so only for those who are obsessively passionate about gambling. For instance, in two studies, Vallerand et al. (2007, Study 1 and 2) demonstrated that harmonious passion was positively related to subjective well-being (SWB) whereas obsessive passion was either not related or negatively related to SWB among dramatic arts students from different theatre schools and colleges across the Province of Quebec. In another study on background and consequences of different styles of engagement in video game play, Przybylski et al. (2009) found that low levels of basic need satisfaction were related to higher obsessive passion, higher amounts of play, more tension after play, and diminished game enjoyment, whereas higher levels of need satisfaction were not related to hours of play but were related to increased harmonious passion, game enjoyment, and energy after play. Further, Ratelle et al. (2004) found that being obsessively passionate about gambling was associated with a range of negative outcomes (e.g., poorer vitality and concentration in daily tasks, as well as increased rumination, anxiety, negative mood, guilt, and problem gambling) among participants recruited at the Montréal Casino, whereas such was not so for harmonious passion. Similar pattern of results were also obtained by Mageau et al. (2005) among those also involved in casino activities. In their investigation of the social-motivational antecedents of passion, Mageau et al. (2009) found that children and teenagers that were in autonomy-supportive environment were more likely to develop harmonious passion than obsessive passion. Meanwhile children and teenagers who attached high value to activity specialization, their activity are self-defined, and their parents exceedingly value the activity are more likely to develop obsessive than harmonious passion. Most revealing is a recent meta-analysis by Curran et al. (2015) involving 94 independent studies on the intrapersonal correlates of harmonious and obsessive passion. The results showed that harmonious passion was positively associated with positive intrapersonal outcomes (e.g., positive affect, satisfaction, flow, performance) and also had either non-significant or negative relationships with maladaptive intrapersonal outcomes (e.g., negative affect, performance avoidance goals and activity/life conflict). On the contrary, obsessive passion had a less desired and sometimes maladaptive pattern of relationships with both positive and negative intrapersonal criterion variables (e.g., negative affect, rumination, vitality). Remarkably, the effect sizes for the positive relationships between obsessive passion and adaptive intrapersonal outcomes (e.g., well-being and integrated motivation regulation) were significantly smaller in magnitude (small-to-moderate) in comparison to harmonious passion (moderate-to-large). Moreover, the previously significant positive associations between obsessive passion and these adaptive outcomes were reduced to non-significance when harmonious passion was controlled for. However, all positive relationships between obsessive passion and maladaptive outcomes remained significant while controlling for harmonious passion. Based on evidence from previous studies on passion, we contend that partaking in gambling may not inevitably lead to negative outcomes. Rather, negative consequences following the gambling involvement are likely to ensue only if obsessive passion is involved. However, gambling that stems from harmonious passion may even lead to some positive consequences. Therefore, we propose that harmonious passion toward gambling should be negatively related to smartphone addiction whereas obsessive passion toward gambling should be positively related to smartphone addiction. Hypothesis 1a Harmonious passion toward gambling will be negatively related to smartphone addiction. Hypothesis 1b Obsessive passion toward gambling will be positively related to smartphone addiction. Gambling passion and schoolwork engagement Success in school is generally desired by most individuals. Yet for students to be successful in their academic work, they need to be engaged in their schoolwork. Previous studies already indicated that factors such as maternal affection, friends (Marion et al. 2014), and achievement-related personal goal (Vasalampi et al. 2009) may influence students’ schoolwork engagement. However, we add to that list by submitting that schoolwork engagement can be influenced by gambling passion for several reasons. First, the entry into university signify a critical turning point for undergraduates because of the opportunity it affords them to leave home and socialize with their peers away from the prying eyes of their parents. This then appears to give them the leeway to get involved in not just positive behaviours but also negative ones. Supporting this stance, research has shown that university period is usually a time when students indulge in high-risk behaviours (Stuhldreher et al. 2007). Second, there are strong indications that students expend more time playing video games in cybercafés than in school-related activities (Lo et al. 2005). Third, we have observed that undergraduates constitute considerable number of fans of various football clubs across the world to the extent that they take part in Internet football wagering. Lastly, there are also strong evidences linking gambling and gambling passion to adverse health consequences among undergraduates (e.g., Skitch and Hodgins 2005; Stuhldreher et al. 2007). Collectively, these evidences suggest that gambling passion should influence schoolwork engagement. However, gambling passion may have different impacts on criterion variables depending on the particular type of passion that is at play (Curran et al. 2015; Vallerand et al. 2003). For example, undergraduates with harmonious passion toward gambling, even though they indulge in gambling, seem to understand the importance of engaging in their schoolwork because they are in control of the gambling. Hence they would not want their passion toward gambling to interfere with their schoolwork engagement. In contrast to harmonious passion, those with obsessive passion toward gambling have uncontrollable urge to indulge in the gambling. That is, gambling takes precedence over other activities in their lives and such individuals may experience negative consequences due to the gambling. For such undergraduates, the negative consequences may include poor academic performance; loss of money that was meant for school fees, feeding and other sundry expenses; insufficient sleep due to having to stay awake at night in order to follow-up the gambling (Stinchfield et al. 2006) especially if the need arises; anxiety over the amount of money invested in the gambling which could later turn into depression if huge sums of money are lost, etc. In line with Vallerand et al.’s (2003) proposition, such individuals with obsessive passion toward gambling could end up feeling guilty for expending too much time on gambling in place of engaging in their schoolwork. Indeed, previous studies seem to provide support for the notion that harmonious passion should increase schoolwork engagement whereas obsessive passion should decrease it. More importantly, studies on passion have uncovered the effects of having either a harmonious or obsessive passion toward an activity that individuals are passionate about. For instance, research (e.g., Carpentier et al. 2012) with college students between the ages of 17 and 32 years shows that harmonious passion is positively associated with psychological well-being indicators such as life satisfaction and flow in one’s studies, while being unrelated to ruminations. In turn, flow in the favourite activity was positively related to psychological well-being. On the contrary, obsessive passion was negatively related to psychological well-being but was unrelated to flow experiences during one’s favourite activity. Obsessive passion also predicted higher levels of ruminations which in turn, were negatively related to the experience of flow. Studies have also revealed that obsessive passion predicts pathological gambling whereas harmonious passion does not (Ratelle et al. 2004). Furthermore, Vallerand et al. (2003, Study 2) found that harmonious passion toward football was related to higher positive affect during the whole football season whereas engaging in the passionate activity out of obsessive passion was related to increase negative affect during the same period. These results have been replicated in ample number of studies including diary studies (Mageau and Vallerand 2007) and studies involving football fans (Verner-Filion et al. 2012). These findings suggest that harmonious passion involves positive feelings during activity engagement whereas obsessive passion entails negative feelings while participating in the same activity. Thus, gambling passion may affect undergraduates’ schoolwork engagement by engendering either positive or negative feelings depending on whether they engage in such activity out of harmonious or obsessive passion, respectively. Generally, results from previous studies robustly indicated that harmonious passion leads to positive and desirable outcomes whereas obsessive passion leads to negative and less desirable outcomes. These findings make us to hypothesize that harmonious passion toward gambling should enhance schoolwork engagement whereas obsessive passion should reduce it. Hypothesis 2a Harmonious passion toward gambling will be positively related to schoolwork engagement. Hypothesis 2b Obsessive passion toward gambling will be negatively related to schoolwork engagement. Smartphone addiction as a mediator between gambling passion and schoolwork engagement As we earlier noted, Internet gamblers sometimes use their smartphone to partake in gambling due to the nature of the gambling and the ease of access to the Internet which smartphone affords them. Thus being passionate about Internet football gambling could lead to smartphone addiction since smartphones are used for Internet gambling purposes. In turn, being addicted to smartphone could have far-reaching adverse consequences on schoolwork engagement especially for undergraduates because they constitute the highest number of individuals who consume smartphone services (Head and Ziolkowski 2012). In addition, because smartphones are Internet-enabled, there is the likelihood for its users to become addicted to them which may result to physical and social problems analogous to those associated with Internet addiction (Shaw and Black 2008). Empirical evidence from earlier studies appears to support the idea that gambling passion should impact upon smartphone addiction. In turn, smartphone addiction should be related to less schoolwork engagement. In other words, it is possible for smartphone addiction to mediate the impacts of Internet football gambling passion on schoolwork engagement. For instance, in a study involving university students, Skitch and Hodgins (2005) found that problem gamblers reported higher levels of both obsessive and harmonious passion for gambling, and that obsessive passion toward gambling was associated with severity of problem gambling behaviour whereas harmonious passion was not. Studies have also linked obsessive passion to increased rumination whereas harmonious passion was not (e.g., Ratelle et al. 2004). This could also be taken to suggest that individuals who are obsessively passionate about Internet football gambling are likely to ruminate about their smartphone because it facilitates involvement in the gambling whereas those with harmonious passion for Internet football gambling should not. Other studies have also indicated that gambling generally is associated with undesirable outcomes. For example, Lloyd et al. (2010) provided evidence that revealed that Internet gamblers of different betting patterns (i.e., non-to-minimal gamblers, sports bettors, casino and sports gamblers, lottery players, and multi-activity gamblers) were linked to varied degrees of problem gambling, mood disorders, substance misuse, history of deliberate self harm, and mental disorder. Tsitsika et al. (2011) found that Internet gambling was associated with problematic Internet use among adolescents. In a study involving online gamblers, McBride and Derevensky (2009) found that problem gamblers are more likely than social gamblers to spend more time gambling with a cell phone. All in all, given the evidence reported in previous studies, one would also expect that Internet football gambling passion will impact on smartphone addiction because most smartphones function basically through Internet-based applications. Without doubt, smartphones can be beneficial to us in many ways especially in areas of enhanced communication and interpersonal interaction. However, there are ongoing concerns on its addictive potential and the concomitant negative consequences. In this regard, Levine et al. (2007), for example, cautioned against the use of such phones basically for leisure rather than academic purposes because it can serve as a distraction to learning in an academic environment. Moreover, recent evidence suggests that excessive smartphone use can be problematic (e.g., Mok et al. 2014). Studies in academic setting have also established an association between cell phone use and academic performance. For example, a study by Lepp et al. (2014) showed that increased cell phone use/texting was associated with decreased GPA and increased anxiety in a large; Midwestern US public university. More recent studies have also provided evidence on the negative consequences of cell phone use among students. For example, in a sample of undergraduates from a large public university in the US, Lepp et al. (2015) showed that after controlling for sex, cigarette smoking, class standing, self-efficacy for self-regulated learning, self-efficacy for academic achievement, and actual high school GPA, cell phone use negatively predicted GPA. So far, since past research suggest a possible link between Internet gambling and smartphone addiction (e.g., McBride and Derevensky 2009) and smartphone use/addiction has been repeatedly related to poor academic outcomes among university students (e.g., Lepp et al. 2014, 2015), we therefore would expect that smartphone addiction may undermine schoolwork engagement and also serve as an intermediary between gambling passion and schoolwork engagement among undergraduates. Hypothesis 3 Smartphone addiction will be negatively related to schoolwork engagement. Hypothesis 4a The relationship between harmonious gambling passion and schoolwork engagement will be mediated by smartphone addiction. Hypothesis 4b The relationship between obsessive gambling passion and schoolwork engagement will be mediated by smartphone addiction. Methods Participants A total of 278 male students of a large federal university in Nigeria who used smartphones to gamble participated in this study. Our choice of only male students follows our observation that only male students engage in Internet football betting/gambling in the context where the study was conducted. Besides, similar studies have equally revealed that more males participate in Internet gambling than females (e.g., Tsitsika et al. 2011) and as such, recent studies (e.g., Wang et al. 2011) have begun to focus exclusively on male participants. Participants were between the ages of 16 and 34 years, with a mean age of 22.39 years (SD = 3.30). All participants had engaged in Internet football betting/gambling at least once over the past month preceding data collection and own a smartphone. Procedure Individuals who participate in Internet gambling were recruited as participants from three Internet football betting centres located close to a large federal university in Nigeria who came to wager on football. In total, we approached 328 males between May and August, 2015 and asked whether they are students of the federal university. The 319 individuals who identified themselves as students of the university were then asked if they own a smartphone and use it for gambling purposes. A total of 310 students used smartphones for gambling purposes. These students (310) were then asked to participate in a study aimed at gaining a better understanding of their gambling, smartphone, and school-related activities. Of the 310 students, 290 consented to participate and were thus handed a questionnaire that included the Gambling Passion Scale (GPS; Rousseau et al. 2002), Smartphone Addiction Scale-Short Version (SAS-SV; Kwon et al. 2013a), and Schoolwork Engagement Inventory (EDA; Salmela-Aro and Upadaya 2012) as well as demographic items on gender and age. All the two hundred and ninety (290) copies of the questionnaire distributed were completed and collected on the spot. However, only properly completed questionnaires (i.e., 278) were subjected to statistical analysis. The remaining 12 were discarded due to improper completion. All in all, the response rate was 95. 86 %. Measures Participants completed self-report questionnaire assessing gambling passion, smartphone addiction, schoolwork engagement, and demographic information (gender and age). All response formats were based on Likert scales except demographic items. Schoolwork engagement was assessed with the Salmela-Aro and Upadaya’s (2012) Schoolwork Engagement Inventory (EDA), which comprises nine items that measure energy (e.g., “I feel strong and vigorous when I am studying”), dedication (e.g., “My schoolwork inspires me”), and absorption (e.g., “I feel happy when I am working intensively at school”) in relation to schoolwork. Respondents rated the degree to which each item applied to them based on estimation from the previous month, using a 0 (never) to 6 (daily) response scale. Items were averaged to form an overall scale score. Higher scores indicate greater engagement with schoolwork. The inventory has satisfactory level of Cronbach’s alpha (α) coefficient of .77 in the present study. Gambling passion was measured with the Rousseau et al.’s (2002) Gambling Passion Scale (GPS). Respondents were asked to think about their favourite Internet football betting/gambling games and then complete the items on the GPS with respect to this activity. The GPS consists of 10 items, 5 for each subscale: obsessive gambling passion (e.g., “I have almost an obsessive feeling for this gambling game”) and harmonious gambling passion (e.g., “This gambling game reflects the qualities I like about myself”). The responses are rated on a 7-point scale ranging from 1 (not agree at all) to 7 (very strongly agree). Possible score on the GPS ranged from 5 to 35 for each of the subscales. Each respondent’s scores were averaged to form a subscale score. Higher scores represent higher obsessive or harmonious passion toward gambling. Estimated Cronbach’s α coefficient of both obsessive and harmonious gambling passion subscales respectively, are .84 and .85 in the present study. Smartphone addiction was measured with the Smartphone Addiction Scale-Short Version (SAS-SV) developed by Kwon et al. (2013a). The SAS-SV consists of ten items. Participants’ responses to each item were scored using a 6-point scale ranging from 1 (strongly disagree) to 6 (strongly agree). Sample items include: “Having my smartphone in my mind even when I am not using it”, and “The people around me tell me that I use my smartphone too much.” Possible score on the scale ranged from 10 to 60. Scores on each item were averaged to form an overall score for the SAS-SV. Higher scores represent greater addiction to smartphone. The SAS-SV is internally consistent with Cronbach’s α coefficient of .76 obtained in the present sample. Ethical considerations We tried to address some ethical issues during the study period. For all prospective respondents, we explained the nature and purpose of the study orally to them. We also clearly stated, both orally and through a letter that accompanied the questionnaire, that participation was voluntary, anonymous, and confidential. Finally, we informed them that they were free to withdraw at any time in the study without any penalty. Strategy of data analysis We conducted some preliminary analyses on the means, standard deviations, reliabilities, and correlations among all key variables using the Statistical Package for the Social Sciences (SPSS) version 21.0. The internal consistency reliability was estimated based on Cronbach’s alpha. Correlations are presented here to test whether the independent variables were too related (i.e., an indication of the risk of multicollinearity) that may warrant combining the scores of the variables to form a composite score. To test our main effects hypotheses, we conducted two hierarchical multiple regression with smartphone addiction and schoolwork engagement as outcome variables, respectively. To test our mediation effects hypotheses, we conducted bootstrapping. Because previous research has demonstrated that age can influence schoolwork and study engagement (e.g., Bakker et al. 2015; Marion et al. 2014), mobile phone use (e.g., Walsh et al. 2011), and gambling severity (e.g., McBride and Derevensky 2009), we controlled for this variable in the analyses. Further, due to conceptual similarities between harmonious and obsessive gambling passion, we controlled for each of them to ensure that our results were not spurious and misleading as well as to be able to detect the unique influence that each type of gambling passion exerts on smartphone addiction and schoolwork engagement. Similar approach was also adopted in a recent study (e.g., Curran et al. 2015). Results Preliminary analysis Means, standard deviations, reliabilities (Cronbach’s alpha), and correlations matrix for all measured variables are presented in Table 1. Inspection of the Cronbach’s alpha coefficients showed satisfactory levels of internal consistency reliabilities of the scales that exceeded the cutoff rules-of-the thumb of .70 commonly recommended for research purpose (Kaplan and Saccuzzo 2013). Besides, the correlations among the study variables, ranging from −.06 to .27, indicated no threat of multicollinearity.Table 1 Means, standard deviations, alpha coefficients, and correlations among the study variables Variable M SD 1 2 3 4 5 1. Age 22.39 3.30 – 2. Harmonious passion 28.14 2.87 −.23*** (.84) 3. Obsessive passion 8.27 2.55 −.10 .11 (.84) 4. Smartphone addiction 18.55 4.02 −.13* .06 .27*** (.76) 5. Schoolwork engagement 46.09 3.30 −.06 .18** −.11 −.19*** (.77) Cronbach’s alpha coefficients for the scales are on the diagonal of the matrix in parentheses N = 278, * p < .05 (two-tailed); ** p < .01 (two-tailed); *** p < .001 (two-tailed) The results in Table 1 indicated that age was significantly and negatively correlated with smartphone addiction (r = −.13, p = .033) but not significantly correlated with schoolwork engagement (r = −.06, n.s.). Harmonious passion was not significantly correlated with smartphone addiction (r = .06, n.s.) whereas it was significantly correlated with schoolwork engagement (r = .18, p = .003). Obsessive passion was significantly and positively correlated with smartphone addiction (r = .27, p < .001) but was not significantly correlated with schoolwork engagement (r = −.11, n.s.). Smartphone addiction was significantly and negatively correlated with schoolwork engagement (r = −.19, p = .001). Hypotheses testing The results of the hierarchical multiple regression and bias corrected (BC) bootstrapping appear in Fig. 2 and Table 2, respectively. Bootstrapping is a re-sampling method computed in order to derive more precise confidence intervals (CI) that has advantage over the causal steps method of Baron and Kenny and the Sobel test that is based on the product of coefficients and the assumption that the sample must be normally distributed before mediation can occur (Hayes 2009; Preacher and Hayes 2004). These assumptions are not necessary in bootstrapping in that it can well be used on small sample size. Thus, given the inherent limitations in the Baron and Kenny’s approach as well as the Sobel test, researchers have begun to discourage the use of such procedures in mediation analysis (Hayes 2009; Preacher and Hayes 2004). We used the PROCESS for SPSS macro (version 2.13) to produce the bootstrapped indirect effects. We drew new samples (with replacements) from our original sample 1000 times and calculated the indirect estimates of the model based on 95 % CI. Mediation is demonstrated if the entire CI lies above or below zero. That is, a predictor X has an indirect effect on Y through the mediating variable M (Hayes 2009; Hayes and Preacher 2010; Preacher and Hayes 2004).Fig. 2 Results of hierarchical multiple regression of predictors of schoolwork engagement. Note *p < .05; **p < .01; ***p < .001. The numerical values on the arrows pointing to smartphone addiction are beta weights taken from the first regression while those pointing to schoolwork engagement are taken from the second regression Table 2 Mediating effect of smartphone addiction in the gambling passion—schoolwork engagement relation Indirect effect pathways Estimate BC 95 % CI Lower Upper HGP → SA → Schoolwork engagement −.004 −.020 .010 OGP → SA → Schoolwork engagement −.034* −.073 −.010 BC bootstrapping results were based on 1000 bootstrapped samples BC bias corrected, CI confidence interval, HGP harmonious gambling passion, OGP obsessive gambling passion, SA smartphone addiction * p < .05 The results of the hierarchical multiple regression in Fig. 2 indicated that after controlling for the influence of age (β = −.13, p < .05), harmonious gambling passion was not significantly related to smartphone addiction (β = .03, n.s.) whereas obsessive gambling passion was significantly and positively related to smartphone addiction (β = .26, p < .001). Therefore H1a was not confirmed whereas H1b was confirmed. Both harmonious (β = .17, p = .005) and obsessive (β = −.14, p = .024) gambling passion were significantly related to schoolwork engagement in positive and negative directions, respectively. Furthermore, smartphone addiction had a significant negative relation with schoolwork engagement (β = −.19, p = .002). Thus, H2a, H2b, and H3 were confirmed. We also tested if smartphone addiction mediated the relationship between the two components of gambling passion (i.e., harmonious and obsessive gambling passion) and schoolwork engagement. The results showed that the BC bootstrap CI for the indirect effect of harmonious gambling passion on schoolwork engagement via smartphone addiction was not statistically different from zero (H4a; bootstrap estimate = −.004, lower CI = −.020, upper CI = .010, n.s.) whereas that of obsessive gambling passion on schoolwork engagement via smartphone addiction was statistically different from zero (H4b; bootstrap estimate = −.034, lower CI = −.073, upper CI = −.010, p < .05). Thus the mediating role of smartphone addiction was confirmed for obsessive gambling passion but not for harmonious gambling passion. That is, obsessive gambling passion was related to decreased schoolwork engagement through smartphone addiction. Discussion The current study focused on gaining insight into how passion toward gambling relates to schoolwork engagement as well as in uncovering the psychological processes that might be responsible for the hypothesized opposing consequences that the two types of gambling passion (i.e., harmonious and obsessive gambling passion) may have on schoolwork engagement. Our expectations were partially supported as results revealed that harmonious passion toward gambling was unrelated to smartphone addiction (i.e., H1a was not confirmed) whereas obsessive passion toward gambling was related to increased levels of smartphone addiction (i.e., H1b was confirmed). These results suggest that having harmonious passion toward gambling may not lead to smartphone addiction, whereas the more individuals are obsessed about gambling, the more they seem to be addicted to their smartphone. Our results appear to be in line with the Dualistic Model of Passion (Vallerand et al. 2003), considering the possibility that individuals driven by harmonious passion toward gambling are in control of their gambling situation and expend great autonomy in participating in the gambling. As such, even if they use their smartphone for Internet gambling purposes, they may not become addicted to it. However, for those who are obsessively passionate about gambling, the reverse seems to be the case given that they are being driven by irresistible urge to indulge in the gambling which incidentally, using their smartphone makes easier for them. These findings are consistent with past research that also revealed that harmonious passion is either unrelated or negatively related to negative outcomes whereas obsessive passion is linked to increased negative outcomes (e.g., Carpentier et al. 2012; Curran et al. 2015; Mageau et al. 2005; Ratelle et al. 2004; Skitch and Hodgins 2005). These would explain why harmonious gambling passion was unrelated to smartphone addiction and the positive relationship found between obsessive gambling passion and smartphone addiction in the present study. As expected, the results showed that both harmonious (H2a) and obsessive (H2b) gambling passion were related to schoolwork engagement and in the hypothesized directions (i.e., positive and negative, respectively). These findings suggest that having a harmonious passion toward gambling may also increase schoolwork engagement among undergraduates, whereas participating in the gambling activity out of obsessive passion seems to undermine their schoolwork engagement. These findings are consistent with the Dualistic Model of Passion (Vallerand et al. 2003) which posits that for those driven by harmonious passion; the passionate activity does not conflict with other aspects of the person’s life since the person is able to juggle successively between the passionate activity and other life domains. However, for those who are obsessed about gambling, they seem to be so preoccupied with the gambling, thereby leading to less attention being paid to their schoolwork. Indeed, consistent with past researches that have shown that having a harmonious passion toward an activity is linked to increased desirable outcomes such as psychological well-being indicators, mastery goal pursuit, performance and the like, whereas obsessive passion toward an activity is unrelated or negatively related to such desirable outcomes (e.g., Carpentier et al. 2012; Vallerand et al. 2007, 2008, Study 1 and 2), the current study shows that having a harmonious passion toward gambling is linked to increased schoolwork engagement whereas obsessive passion toward gambling is linked to decreased schoolwork engagement. The results also provide evidence for H3 which predicted a negative relationship between smartphone addiction and schoolwork engagement. This finding suggests that increased smartphone addiction appears to lead to reduced levels of schoolwork among undergraduates. This finding is not surprising given that recent research has found that excessive smartphone/cell phone use has been linked to negative psychological health outcomes (e.g., Mok et al. 2014), and poor academic performance, especially among university students (e.g., Lepp et al. 2014, 2015). What this means in essence is that using smartphone to the extent of becoming addicted to it may lead to poorer schoolwork engagement among undergraduates. This would explain the negative relationship found between smartphone addiction and schoolwork engagement. We expected that smartphone addiction would mediate the relationship between the two types of gambling passion and schoolwork engagement. However, our data only provided evidence for the mediatory role of smartphone addiction in the obsessive gambling passion—schoolwork engagement relation (H4b) but not for harmonious gambling passion—schoolwork engagement relation (H4a). These results suggest that smartphone addiction is an important avenue that helps in conveying the influence of obsessive gambling passion on schoolwork engagement. If undergraduates take part in Internet gambling out of obsessive gambling passion, they seem to become addicted to their smartphone because it serves as a medium of partaking in the gambling with ease, which in turn decreases their schoolwork engagement. Our results are consistent with previous studies that showed that obsessive passion was positively related to maladaptive outcome such as ruminations, which in turn was associated with decreased experience of flow (e.g., Carpentier et al. 2012). However, a plausible explanation as to why smartphone addiction did not mediate the relationship between harmonious gambling passion and schoolwork engagement could be because gambling based on harmonious passion may not necessarily lead to maladaptive outcomes as revealed in the present as well as past studies (e.g., Carpentier et al. 2012; Curran et al. 2015; Ratelle et al. 2004). Thus, maladaptive outcome such as smartphone addiction in turn may be unable to transmit the influence of harmonious gambling passion on schoolwork engagement in the present study. Supporting this stance, our earlier results (H1a) showed that harmonious gambling passion was unrelated to smartphone addiction in the first instance, and as such smartphone addiction could not convey/mediate the impact of harmonious gambling passion on schoolwork engagement. Limitations, strengths, and future directions Any study of this nature has inherent limitations and needed to be mentioned. Therefore the findings of the present study must be interpreted in the light of these limitations. First, the study adopted a cross-sectional design and is correlational in nature such that the relationship between Internet gambling passion and smartphone addiction were tested in only one direction. We cannot completely rule out the possibility of a reciprocal relationship between Internet gambling passion and smartphone addiction such that smartphone addiction provides easy access to Internet gambling or whether Internet gambling passion leads to smartphone addiction. An important step for future research is to utilize longitudinal designs to determine whether changes in Internet gambling passion and smartphone addiction can account for significant variance in schoolwork engagement over a period of time. However, the method adopted in the current study is consistent with current approach (e.g., Carpentier et al. 2012). Second, this study collected data based on only self-report measures. This may raise the issue of common method variance. Future studies should endeavour to utilize data from other sources such as the students’ peers and lecturers. However, we tried to limit the problem of common method variance by utilizing self-report measures with different response formats, ensuring anonymity and confidentiality of responses, and encouraging respondents to be honest in their responses (Podsakoff et al. 2003). This study therefore contributes to the literature by extending the influence of gambling passion on an emerging construct (i.e., schoolwork engagement). The study also adds to the area of positive psychology by providing further support to the idea that having passion for an activity can also lead to positive outcomes in school setting depending on the type of passion that is adopted. Moreover, to our knowledge, this is the first study that has attempted to synthesize the relationships between gambling passion, smartphone addiction, and schoolwork engagement in a sample of male undergraduates within the same study. This study is also informative given that the data were gathered at an actual Internet football betting centres while undergraduates’ bettors were involved in actual Internet wagering activities. Theoretical and practical implications of findings The findings of our study have some theoretical and practical implications. Indeed, we found evidence that the two types of gambling passion (i.e., harmonious and obsessive gambling passion) relate in different ways to common criteria—smartphone addiction and schoolwork engagement. This provides further support for the Vallerand et al.’s (2003) Dualistic Model of Passion that contends that engaging in passionate activity may not always lead to detrimental consequences, and can even be beneficial depending on the type of passion that is at play. In essence, even in academic setting, empirical evidence seems to favour the theory. Thus, the theory can well provide explanations as to why we found that harmonious and obsessive gambling passion related in different ways to smartphone addiction and schoolwork engagement. From a practical view point, gambling has traditionally been viewed from a negative lens. For instance, Stuhldreher et al. (2007) have already cautioned that “Gambling is an emerging high-risk behavior that has sounded the alarm bell on campuses nationwide” (p. 75). However, as our findings suggest, such negative connotation usually comes into play especially when obsessive gambling passion is involved. Thus, rather than condemning or promoting gambling in its entirety, vocational counsellors and academic advisers should inform their students that it is possible for them to be passionate about gambling, while at the same time being highly engaged in their schoolwork activities, to the degree to which their passion for gambling is in harmony with other life pursuits, especially those that have to do with schoolwork. In other words, they can be passionate gamblers and still be able to strike a balance between their gambling activities and schoolwork engagements. They should also sensitize them on the dangers of being obsessively passionate about gambling as this may adversely interfere with their schoolwork engagement and could also impact negatively on other important school-related outcomes. School management should also endeavour to conduct regular screening exercises for gambling among their students using indices such as the revised South Oaks Gambling Screen (SOGS; Lesieur and Blume 1993) in order to identify those who are already or likely to become pathological gamblers. After the screening, they can then commence treatment for those who require such. Moreover, gambling regulatory agencies in Nigeria such as the National Lottery Regulatory Commission (NLRC) should focus on promoting responsible gambling activities. They should emphasize the need for gambling advertisements to clearly point out that gamblers should engage in gambling activities primarily for the fun of it rather than as a means of enriching themselves. They should also encourage the inclusion of pop-up messages in adverts during gambling on the Internet to occasionally remind Internet gamblers of the potential detrimental consequences of excessive gambling activities. Following our observation, the NLRC should also ensure that operators of gambling centres strictly comply with the gambling age of 18 years and above by penalising defaulters. We also discovered that smartphone addiction seems to serve as an impediment to undergraduates’ schoolwork engagement. Although we do not advocate for the ban of the use of smartphone on campus given that it could also facilitate learning if used for learning purposes, we however concur with Levine et al. (2007) by urging smartphone users not to use such phones exclusively for leisure purposes on campus because it can be distractive to learning. Thus, school management could prevent or minimize the negative impact of smartphone addiction on their students by taking proactive steps at preventing or minimizing the occurrence of smartphone addiction through sensitization programmes on the dangers of being addicted to their smartphone. In a similar vein, conducting regular screening exercises for smartphone addiction using indicators such as the Smartphone Addiction Scale-Short Version (SAS-SV; Kwon et al. 2013a) not just for newly admitted students, but also for those admitted earlier could be beneficial. Such screening exercises have the potential in diagnosing students who are likely to become addicted or those already addicted to their smartphone and thus preventive measures or prompt commencement of treatment as the case may be, can ensue. In addition, given the apparent similarities between smartphone addiction and Internet addiction, some of the interventions designed for Internet gambling addiction purposes such as those highlighted in Griffiths (2003) could also be effective in preventing and alleviating smartphone addiction among university students. Conclusion This study aimed at studying how the two types of gambling passion— harmonious and obsessive gambling passion relate to schoolwork engagement and whether smartphone addiction mediates these relations among a sample of undergraduates in Nigeria. The study was guided by the Dualistic Model of Passion (Vallerand et al. 2003) which contends that passion can lead to different consequences depending on the specific type of passion that is at play. Indeed, the results support the central proposition of the Dualistic Model of Passion, and thus broaden our understanding of the two types of gambling passion and their relationship with schoolwork engagement for undergraduates involved with Internet football gambling. It also offers additional and interesting insight into the mediating role of smartphone addiction. Thus, besides being associated with poorer schoolwork engagement, we also found strong evidence for the mediating role of smartphone addiction in the relationship between obsessive passion toward gambling and schoolwork engagement. Nonetheless, the findings of the current study call for prospective research to investigate other potential intervening variables, including but not limited to educational level, academic self-efficacy, gamblers’ beliefs, gambling motives, and gambling expectancy, in the association between gambling passion and schoolwork engagement in a more representative sample of not just university students but also secondary school students. Authors’ contributions IKE conception and design, draft of the manuscript. LIU literature search, manuscript revision. DIU acquisition of data and analysis. All authors read and approved the final manuscript. Competing interests The authors declare that they have no competing interests. Ethical approval All procedures performed in this study involving human performance were in accordance with the ethical standards of the institution and/or natural research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standard. Funding We did not receive any grant from any governmental or non-governmental agencies. Informed consent Participants were informed of the purpose of the study and were told that participation was voluntary. They were also informed that they were free to withdraw from the study at any stage if they so wished. ==== Refs References Al-Barashdi HS Bouazza A Jabur NH Smartphone addiction among university undergraduates: A literature review Journal of Scientific Research & Reports 2015 4 3 210 225 10.9734/JSRR/2015/12245 Appleton JJ Christenson SL Furlong MJ Student engagement with school: Critical conceptual and methodological issues of the construct Psychology in the Schools 2008 45 5 369 386 10.1002/pits.20303 Bakker AB Vergel AIS Kuntze J Student engagement and performance: A weekly diary study on the role of openness Motivation and Emotion 2015 39 1 49 62 10.1007/s11031-014-9422-5 Barnes GM Welte JW Hoffman JH Tidwell MC Comparisons of gambling and alcohol use among college students and noncollege young people in the United States Journal of American College Health 2010 58 443 452 10.1080/07448480903540499 20304756 Barrault S Varescon I Cognitive distortions, anxiety, and depression among regular and pathological gambling online poker players CyberPsychology, Behavior, and Social Networking 2013 16 3 183 188 10.1089/cyber.2012.0150 Carpentier J Mageau GA Vallerand RJ Ruminations and flow: Why do people with a more harmonious passion experience higher well-being? Journal of Happiness Studies 2012 13 3 501 518 10.1007/s10902-011-9276-4 Curran T Hill AP Appleton PR Vallerand RJ Standage M The psychology of passion: A meta-analytical review of a decade of research on intrapersonal outcomes Motivation and Emotion 2015 Deci EL Ryan RM The “what” and “why” of goal pursuits: Human needs and the self-determination of behavior Psychological Inquiry 2000 11 4 227 268 10.1207/S15327965PLI1104_01 Derevensky JL Gupta R Winters K Prevalence rates of youth gambling problems: Are the current rates inflated? Journal of Gambling Studies 2003 19 4 405 425 10.1023/A:1026379910094 14634300 Ekpeke, M. (2015). Nigeria ranked 17th in global smartphone usage. Retrieved from http://itpulse.com.ng/nigeria-ranked-17th-in-global-smartphone-usage/. eMarketer. (2015). Worldwide Internet and mobile users: emarketer’s updated estimates for 2015. Retrieved from https://insights.ap.org/uploads/images/eMarketer_Estimates_2015.pdf. Fasae JK Adegbilero-Iwari I Mobile devices for academic practices by students of college of sciences in selected Nigerian private universities The Electronic Library 2015 33 4 749 759 10.1108/EL-03-2014-0045 Griffiths M Gambling on the internet: A brief note Journal of Gambling Studies 1996 12 4 471 473 10.1007/BF01539190 24234164 Griffiths M Does Internet and computer “addiction” exists? Some case study evidence CyberPsychology & Behavior 2000 3 2 211 218 10.1089/109493100316067 Griffiths M Internet gambling: Issues, concerns, and recommendations CyberPsychology & Behavior 2003 6 6 557 568 10.1089/109493103322725333 14756922 Griffiths M Wardle H Orford J Sproston K Erens B Sociodemographic correlates of internet gambling: Findings from the 2007 British Gambling Prevalence Survey CyberPsychology & Behavior 2009 12 2 199 202 10.1089/cpb.2008.0196 19072080 Hayes AF Beyond Baron and Kenny: Statistical mediation analysis in the new millennium Communication Monographs 2009 76 4 408 420 10.1080/03637750903310360 Hayes AF Preacher KJ Quantifying and testing indirect effects in simple mediation models when the constituent paths are nonlinear Multivariate Behavioral Research 2010 45 4 627 660 10.1080/00273171.2010.498290 26735713 Head M Ziolkowski N Understanding student attitudes of mobile phone features: Rethinking adoption through conjoint, cluster and SEM analyses Computers in Human Behavior 2012 28 6 2331 2339 10.1016/j.chb.2012.07.003 Husky MM Michel G Richard J-B Guignard R Beck F Gender differences in the associations of gambling activities and suicidal behaviors with problem gambling in a nationally representative French sample Addictive Behaviors 2015 45 45 50 10.1016/j.addbeh.2015.01.011 25644586 Jacobs DF Derevensky JL Gupta R Youth gambling in North America: Long-term trends and future prospects Gambling problems in youth: Theoretical and applied perspectives 2004 New York, NY Kluwer Academic/Plenum 1 25 Kaplan RM Saccuzzo DP Psychological assessment and theory: Creating and using psychological tests 2013 8 Belmont, CA Wadsworth Cengage Learning Kwon M Kim D-J Cho H Yang S The Smartphone Addiction Scale: Development and validation of a short version for adolescents PLoS ONE 2013 8 12 e83558 10.1371/journal.pone.0083558 24391787 Kwon M Lee J-Y Won W-Y Park J-W Min J-A Hahn C Kim D-J Development and validation of a Smartphone Addiction Scale (SAS) PLoS One 2013 8 2 e56936 10.1371/journal.pone.0056936 23468893 Lee H Ahn H Choi S Choi W The SAMS: Smartphone addiction management system and verification Journal of Medical Systems 2014 Lepp A Barkley JE Karpinski AC The relationship between cell phone use, academic performance, anxiety, and satisfaction with life in college students Computers in Human Behavior 2014 31 343 350 10.1016/j.chb.2013.10.049 Lepp A Barkley JE Karpinski AC The relationship between cell phone use and academic performance in a sample of U.S. college students SAGE 2015 Lesieur HR Blume SB Revising the South Oaks Gambling Screen in different settings Journal of Gambling Studies 1993 9 3 213 223 10.1007/BF01015919 Levine LE Waite BM Bowman LL Electronic media use, reading, and academic distractibility in college youth Cyber Psychology & Behavior 2007 10 4 560 566 10.1089/cpb.2007.9990 Li H Mao LL Zhang JJ Wu Y Li A Chen J Dimensions of problem gambling behavior associated with purchasing sports lottery Journal of Gambling Studies 2012 28 1 47 68 10.1007/s10899-011-9243-3 21365440 Lloyd J Doll H Hawton K Dutton WH Geddes JR Goodwin GM Rogers RD Internet gamblers: A latent class analysis of their behaviours and health experiences Journal of Gambling Studies 2010 26 3 387 399 10.1007/s10899-010-9188-y 20422442 Lo SK Wang CC Fang W Physical interpersonal relationships and social anxiety among online game players CyberPsychology & Behavior 2005 8 1 15 20 10.1089/cpb.2005.8.15 15738689 MacCormick JS Dery K Kolb DG Engaged or just connected? Smartphones and employee engagement Organizational Dynamics 2012 41 3 194 201 10.1016/j.orgdyn.2012.03.007 Mageau GA Vallerand RJ The moderating effect of passion on the relation between activity engagement and positive affect Motivation and Emotion 2007 31 4 312 321 10.1007/s11031-007-9071-z Mageau GA Vallerand RJ Charest J Salvy S-J Lacaille N Bouffard T Koestner R On the development of harmonious and obsessive passion: The role of autonomy support, activity specialization, and identification with the activity Journal of Personality 2009 77 3 601 646 10.1111/j.1467-6494.2009.00559.x 20078732 Mageau GA Vallerand RJ Rousseau FL Ratelle CF Provencher PJ Passion and gambling: Investigating the divergent affective and cognitive consequences of gambling Journal of Applied Social Psychology 2005 35 1 100 118 10.1111/j.1559-1816.2005.tb02095.x Marion D Laursen B Kiuru N Nurmi J-E Salmela-Aro K Maternal affection moderates friend influence on schoolwork engagement Developmental Psychology 2014 50 3 766 771 10.1037/a0034295 24015690 McBride J Derevensky J Internet gambling behavior in a sample of online gamblers International Journal of Mental Health and Addiction 2009 7 1 149 167 10.1007/s11469-008-9169-x Messerlian C Byrne AM Derevensky JL Gambling, youth and the internet: Should we be concerned? Canadian Child and Adolescent Psychiatry Review 2004 13 1 3 6 19030146 Mok J-Y Choi S-W Kim D-J Choi J-S Lee J Ahn H Song W-Y Latent class analysis on internet and smartphone addiction in college students Neuropsychiatric Disease and Treatment 2014 10 817 828 24899806 Monaghan S Derevensky J Sklar A Impact of gambling advertisements and marketing on children and adolescents: Policy recommendations to minimise harm Journal of Gambling Issues 2008 22 252 274 10.4309/jgi.2008.22.7 National Lottery Act. (2005). Law regulating lottery business in Nigeria. Retrieved from the National Lottery Regulatory Commission website: http://www.natlotregcom.gov.ng/downloads/NLA_2005.pdf. Okafor GO Malizu CF Mobile phone use patterns among university undergraduates in Nigeria International Journal of Advance Research 2014 2 10 1 16 Pallanti S Bernardi S Quercioli L The shorter PROMIS questionnaire and the internet addiction scale in the assessment of multiple addictions in a high-school population: Prevalence and related disability CNS Spectrums 2006 11 12 966 974 10.1017/S1092852900015157 17146410 Peltzer K Pengpid S Gambling behaviour and psychosocial correlates among university students in 23 low, middle and emerging economy countries Mediterranean Journal of Social Sciences 2014 5 27 1086 1092 Podsakoff PM MacKenzie SB Lee JY Podsakoff NP Common method biases in behavioral research: A critical review of the literature and recommended remedies Journal of Applied Psychology 2003 88 5 879 903 10.1037/0021-9010.88.5.879 14516251 Preacher KJ Hayes AF SPSS and SAS procedures for estimating indirect effects in simple mediation models Behavior Research Methods, Instruments, & Computers 2004 36 4 717 731 10.3758/BF03206553 Przybylski AK Weinstein N Ryan RM Rigby CS Having to versus wanting to play: Background and consequences of harmonious versus obsessive engagement in video games CyberPsychology & Behavior 2009 12 5 485 492 10.1089/cpb.2009.0083 19772442 Ratelle CF Vallerand RJ Mageau GA Rousseau FL Provencher P When passion leads to problematic outcomes: A look at gambling Journal of Gambling Studies 2004 20 2 105 119 10.1023/B:JOGS.0000022304.96042.e6 15060328 Rousseau FL Vallerand RJ Ratelle CF Mageau GA Provencher PJ Passion and gambling: On the validation of the Gambling Passion Scale (GPS) Journal of Gambling Studies 2002 18 1 45 66 10.1023/A:1014532229487 12050847 Ryan RM Deci EL Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being American Psychologist 2000 55 1 68 78 10.1037/0003-066X.55.1.68 11392867 Salanova M Schaufeli WB Martinez IM Breso E How obstacles and facilitators predict academic performance: The mediating role of study burnout and engagement Anxiety, Stress, and Coping 2010 23 1 53 70 10.1080/10615800802609965 Salmela-Aro K Kiuru N Leskinen E Nurmi J-E School-Burnout Inventory (SBI): Reliability and validity European Journal of Psychological Assessment 2009 25 1 48 57 10.1027/1015-5759.25.1.48 Salmela-Aro K Upadaya K The Schoolwork Engagement Inventory: Energy, dedication, and absorption (EDA) European Journal of Psychological Assessment 2012 28 1 60 67 10.1027/1015-5759/a000091 Schaufeli WB Martinez IM Pinto AM Salanova M Bakker AB Burnout and engagement in university students: A cross-national study Journal of Cross-Cultural Psychology 2002 33 5 464 481 10.1177/0022022102033005003 Shaw M Black DW Internet addiction: Definition, assessment, epidemiology and clinical management CNS Drugs 2008 22 5 353 365 10.2165/00023210-200822050-00001 18399706 Siu OL Bakker AB Jiang X Psychological capital among university students: Relationships with study engagement and intrinsic motivation Journal of Happiness Studies 2014 15 4 979 994 10.1007/s10902-013-9459-2 Skitch SA Hodgins DC A passion for the game: Problem gambling and passion among university students Canadian Journal of Behavioural Science 2005 37 3 193 197 10.1037/h0087256 Stinchfield R Hanson WE Olson DH Problem and pathological gambling among college students New Directions for Student Services 2006 113 63 72 10.1002/ss.196 Stuhldreher WL Stuhldreher TJ Forrest KY-Z Gambling as an emerging health problem on campus Journal of American College Health 2007 56 1 75 83 10.3200/JACH.56.1.75-88 17711830 Tsitsika A Critselis E Janikian M Kormas G Kafetzis DA Association between Internet gambling and problematic Internet use among adolescents Journal of Gambling Studies 2011 27 3 389 400 10.1007/s10899-010-9223-z 20953681 Vallerand RJ Blanchard C Mageau GA Koestner R Ratelle C Léonard M Marsolais J Les passions de l’âme: On obsessive and harmonious passion Journal of Personality and Social Psychology 2003 85 4 756 767 10.1037/0022-3514.85.4.756 14561128 Vallerand RJ Mageau GA Elliot AJ Dumais A Demers M-A Rousseau F Passion and performance attainment in sport Psychology of Sport and Exercise 2008 9 3 373 392 10.1016/j.psychsport.2007.05.003 Vallerand RJ Salvy S-J Mageau GA Elliot AJ Denis PL Grouzet FME Blanchard C On the role of passion in performance Journal of Personality 2007 75 3 505 534 10.1111/j.1467-6494.2007.00447.x 17489890 Vasalampi K Salmela-Aro K Nurmi J-E Adolescents’ self-concordance, school engagement, and burnout predict their educational trajectories European Psychologist 2009 14 4 332 341 10.1027/1016-9040.14.4.332 Verner-Filion J Lafrenière M-AK Vallerand RJ On the accuracy of affective forecasting: The moderating role of passion Personality and Individual Differences 2012 52 7 849 854 10.1016/j.paid.2012.01.014 Walsh SP White KM Cox S Young RM Keeping in constant: The predictors of young Australians’ mobile phone involvement Computers in Human Behavior 2011 27 1 333 342 10.1016/j.chb.2010.08.011 Wang CKJ Liu WC Chye S Chatzisarantis NLD Understanding motivation in internet gaming among Singaporean youth: The role of passion Computers in Human Behavior 2011 27 3 1179 1184 10.1016/j.chb.2010.12.012 Weinstein A Lejoyeux M Internet addiction or excessive Internet use The American Journal of Drug and Alcohol Abuse 2010 36 5 277 283 10.3109/00952990.2010.491880 20545603

PMC005xxxxxx/PMC5002048.txt

==== Front Evol IntellEvol IntellEvolutionary Intelligence1864-59091864-5917Springer Berlin Heidelberg Berlin/Heidelberg 14010.1007/s12065-016-0140-7Special IssueReduced projection angles for binary tomography with particle aggregation http://orcid.org/0000-0002-1798-9615al-Rifaie Mohammad Majid m.majid@gold.ac.uk Blackwell Tim t.blackwell@gold.ac.uk Department of Computing, Goldsmiths, University of London, London, SE14 6NW UK 8 8 2016 8 8 2016 2016 9 3 67 79 1 5 2016 15 7 2016 22 7 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.This paper extends particle aggregate reconstruction technique (PART), a reconstruction algorithm for binary tomography based on the movement of particles. PART supposes that pixel values are particles, and that particles diffuse through the image, staying together in regions of uniform pixel value known as aggregates. In this work, a variation of this algorithm is proposed and a focus is placed on reducing the number of projections and whether this impacts the reconstruction of images. The algorithm is tested on three phantoms of varying sizes and numbers of forward projections and compared to filtered back projection, a random search algorithm and to SART, a standard algebraic reconstruction method. It is shown that the proposed algorithm outperforms the aforementioned algorithms on small numbers of projections. This potentially makes the algorithm attractive in scenarios where collecting less projection data are inevitable. Keywords Binary tomographyDiscrete tomographyParticle aggregationUnderdetermined linear systemsReduced projectionsissue-copyright-statement© Springer-Verlag Berlin Heidelberg 2016 ==== Body Introduction Tomographic reconstruction is the process of inferring the internal structure of an object from a set of projected images. The projected images are records of the quantity of penetrating radiation that has passed through, or has been emitted from the interior of, the object in question. There are many applications, ranging from medical imaging (CT, SPECT, PET and MRI) [4, 5, 18] to oceanography (seismic tomography) [16] and quantum tomography (quantum state tomography) [6]. Although an exact reconstruction is possible by use of the inverse Radon transform, in practice the discrete nature of the imaging, and the finite number of available projections, mean that approximate and discrete techniques must be employed. The continuous density distribution of the object is modelled as a grid of pixels and the projections are acquired in bins because cameras consist of arrays of detectors of finite size [5]. Even after discrete modelling, the remaining mathematical problem may be ill-defined due to underdetermination: the number of independent relationships amongst the unknown quantities is fewer than their number. As a result, the solution of the inverse problem is not unique, and indeed very many solutions might exist. This incompleteness of data arises from cost, time and geometrical concerns. For instance, the importance of cost reduction in industrial applications results in shortened scan duration and fewer projected images; similarly, in electron tomography, the damage caused to the sample by the electron beam reduces the number of collectable projections [15]. The classical filtered back projection [8, 13] technique is a relatively quick and effective reconstruction procedure. However, increasing computation power means that algebraic reconstruction techniques (algebraic-RT or ART) are gaining prominence. This is due to ART’s potential for greater accuracy, albeit at increased time of execution. The first ART algorithm was a rediscovery [7] of the Kaczmarz method for solving linear equations [12]. An improved Kaczmarz method for image reconstruction, SART, (simultaneous-ART) was proposed by Andersen and Kac [2]. SART remains popular to this day and has been the subject of mathematical analysis (for example, [10]). Prior knowledge can inform algorithms and speed up computation. For example, if it is known that the object is composed of just a few regions of homogeneous density, discrete tomography can be employed. The aim is to reconstruct an image that is composed of just a few greyscale values. And, as an extreme instance of discrete tomography, if just two greyscale values are assumed, corresponding to the interior and exterior of the object, the problem is to find a binary reconstruction [9]. The aim of this paper is to further investigate a binary reconstruction technique [1] based on the aggregation of particles. The idea is to suppose that pixel values 0 and 1 represent particles that may be absent or present in a particular cell (a pixel), and for particles to move freely until they meet, and thereupon “stick” to, clusters of other particles, subject to a concomitant reduction in error. The underlying assumption is that the preferred solutions to the inverse problem will be those solutions that are more homogeneous. Particles will, therefore, tend to move to unoccupied pixels with a greater neighbourhood count. The selection of a particle for movement was random in the previous version of PART [1]; this meant that many moves had to be rejected. In the updated version of reported here, isolated particles are preferentially selected for movement. The paper continues with an overview of tomography and of reconstruction. Then, the aggregation algorithm, Particle aggregate-RT (PART) is specified along with its updated version; after highlighting the importance of smaller number of projections, a section detailing a sequence of experiments compares the performance of the updated version (referred to as PART 2), to the original PART algorithm, PART 1, SART, random search (RS) and filtered back projection (FBP) on a number of phantoms (i.e. pre-prepared exact images). Additionally, in a second set of experiments, the newly proposed algorithm, is analysed under several number of projections and is compared against the other algorithms. The paper ends with a summary of the main findings and suggestions for future research. Tomography and algebraic reconstruction There are two important imaging modalities, parallel beam and fan beam tomography. In either modality, an array of detectors is rotated to lie at a number of (usually) equally spaced angles in [0,π). Figure 1 shows the two modalities and the pixellated representation of the object. Ideally, if the detectors have perfect collimators, each detector will record the amount of radiation received in a finite width beam.Fig. 1 Tomography geometry. Left parallel beam geometry; right fan beam geometry However, an approximate model of the physical measurement must be built in order to formalise the mathematical reconstruction problem. This approximation is called the forward model. Beams are typically modelled by parallel rays (Fig. 1-left). Each ray is incident on the centre of each detector or projection bin. The imaging process is approximated by a projection matrix A∈R≥0m×n where m is the total number of rays collected (equal to the number of rays at each projection angle multiplied by the number of projection angles) and n is the number of pixels in the reconstructed image. If b∈Rm is a vector of detector values, the continuous/discrete reconstruction problem can be stated as: 1 findx∈Rn∈{0,1,…,k-1}n,k>1such thatAx=b. The binary problem is k=2 i.e. with x∈{0,1}n. The methods used to computed the intersection between the ray and the pixels vary. One such common method is the line model, where the entries in the projection matrix, aij, are computed by measuring the length of the intersection between the line of the ray and the pixel (see Fig. 2-left). In this model, when the projection rays are parallel to the horizontal or vertical axes, the weight function exhibits two discontinuities; these, caused by floating point error, could lead to wrongly setting the weight entries to 0 instead of 1, or vice versa. In order to overcome this issue, Joseph’s weighting scheme [11] could be used instead. In this model, the interpolation coefficients are calculated when following the line column by column or row by row (based on the projection angle chosen). Thus, linear interpolation between the centres of the two adjacent pixels are applied. See Fig. 2-middle. In another model, the strip mode, strips are used with width larger than a unit instead of lines. Therefore, the intersection area between strip i and pixel j determines the weight aij as displayed in Fig. 2-right. While in the the strip model the column sums of the projection matrix is constant, this is does not hold for the line and Joseph models. In this work, in order to compute the entries Aij of the projection matrix, a more refined line model which uses the length of the intersection between the ray and the pixel is used.Fig. 2 Three projection models. From left to right: line, Joseph and strip models Since the equation Ax=b is, in general, underdetermined, it cannot be inverted. Instead an approximate solution y must be obtained (for example, by FBP, or SART). This trial solution is forward projected according to the measurement model: Ay=c with an associated lp projection error ϵ(y)=||b-c||p where the lp,p≥1, norm is defined ||v||p≡∑|x|p1p. An iterative scheme will produce a sequence of candidate solutions, y(k),k=1,2,…, of decreasing error. A zero projection error might yield a reconstructed solution y that is not identical to the original object x. This is due to underdetermination. However, in cases where the reference image is known, the proximity of y to x offers a second and more stringent measure of algorithm performance. Consider the following measures: 2 e1=||b-c||1 3 e2=||y-x||1 A zero value of e1 solves the problem Ay=b but does not guarantee reconstruction proximity. e2 provides a check: a value of zero corresponds to a reconstructed image that is the exact replica of the original. Reconstruction by particle aggregation In many applications, the reconstructed image is expected to consist of patches of various sizes of uniform pixel value, since many physical objects of interest consist of uniform structures. Non-uniform regions with randomly varying pixel values would be construed as noisy and unphysical. Relevant reconstructed images are therefore those with low entropy. This observation suggests the following assumption: given a number of distinct candidate reconstructions, {y:Ay=c}, with identical error ϵ(y), the preferred reconstruction is the one with the lowest entropy (or one of the reconstructions of lowest entropy, in the case of non-uniqueness). It would clearly be beneficial to equip a reconstruction algorithm with this assumption, under those conditions where the assumption might be expected to hold. The principal idea underlying the aggregation technique proposed in this paper and motivated by the low-entropy assumption, is to suppose that pixel values are mobile particles, moving from pixel to pixel. The low-entropy assumption is implemented by requiring that particles stick together in clusters to form aggregates of uniform pixel value. A model of aggregation for any random deposition process that is dominated by diffusive transport, for example electodeposition and mineral growth, was proposed by Witten and Sander [20]. Their model, known as Diffusion Limited Aggregation or DLA, is remarkably simple: a particle is released from a random point on a boundary and subsequently follows a random walk until it strikes a stationary particle at some location within the enclosing boundary. The walking particle sticks to the stationary particle and another particle is released. Surprisingly complex dendrite-like clusters with fractal structure are formed by repeated application of this simple rule. The reconstruction problem is converted into particle aggregation with the following correspondence:image x→ configuration of particles, pixels → cells pixel values 1/0 → presence/absence of a single particle, image → a grid of cells. Furthermore, an objective functionerror → objective function converts the growth model into an optimisation problem: only those aggregates that lower the objective function are permitted to form. A direct implementation of DLA as a reconstructive process would be very expensive since a randomly walking particle might pass by many isolated cells before arriving at a boundary cell; diffusion can be accelerated by causing a particle to jump from cell a to a vacant cell b, picked uniformly at random from all vacant cells. Although a jump has been made, the particle might not necessarily ‘stick’. Suppose a particle has jumped from a to b and that b is a boundary cell of a particle cluster1. We might suppose that whether the particle sticks or not to the cluster is conditional on the number of occupied neighbours of the boundary cell b relative to the neighbour count for cell a - with higher neighbourhood counts being preferred, and on the fitness of the new configuration. There are a number of ways to deal with a particle that has jumped to a vacant cell but does not stick. For example, it could simply return to a. With these considerations in mind, the particle aggregate reconstruction technique (PART) can be specified.2 Algorithm 1 specifies an application of PART to a single particle. Here, y is the reconstructed image, Select (see Algorithm 2) returns pixels a,b∈y,a≠b, such that a is occupied and b is empty. n is the number of occupied cells in the neighbourhood (Moore or von Neumann) of a particular cell and ϵ(a→b) is the error of the new image with the pixel a set to zero and pixel b set to 1. u is a sample drawn from U(0, 1) (the uniform distribution on [0, 1]). The algorithm has two parameters p1 and p2. p1 governs the influence of the local neighbourhood constraint: the requirement to move to a neighbourhood of higher local particle density. p1=1 corresponds to a random search and the neighbourhood constraint is ignored. A move a→b will always be attempted even if the neighbourhood function n is lowered. In contrast, p2 governs the influence of the global constraint on the particle configuration as a whole. If p2=0, a move a→b will always be rejected if it does not lower or equal the current error. The algorithm is greedy. If p2>0, the algorithm is not greedy and a configuration with higher error will be accepted with probability p2. Movement away from a local minima of ϵ can occur. In principle, p2 might depend on the change in error (and on a steadily reducing temperature parameter as in simulated annealing). While finding the optimal value for p2 is not explored in this paper (and p2 is set to zero here), optimising this parameter is a subject of an ongoing research. Algorithm 1 specifies a trial update of a single particle. Each application incurs a cost of a single function evaluation (ϵ(y)). The algorithm is iterated until zero error or until a set number of function evaluations (FEs) has been achieved. As stated by Reynolds [17], the three simple rules of interaction in flocks are collision avoidance, velocity matching and flock centring. Swarms differ from flocks in the sense that there is no velocity matching. The aggregating particles in PART can be considered as individuals in a swarm. The dynamic rules of particles swarms are of the form:If too close or colliding to neighbouring particles, move away Else if too far from neighbours, move closer. where rule 1 opposes crowding and rule 2 brings the particles together in a swarm. The single occupancy condition implements the anti-crowding rule, and the (conditional) move to a neighbourhood of higher particle density, as measured by the neighbourhood function n implements rule 2. The error function ϵ(y) imposes a global constraint on the swarm as a whole. In an altered version of PART, further emphasis is placed on the aggregation of particles by more systematically choosing isolated particles as more suitable a pixels to be placed in b pixels. This is arranged by creating an ordered list of particles’ neighbourhood counts (see Algorithms 1 and 3). Experiments and results In [1], three experiments were conducted in order to investigate the performance of PART 1 in the context of binary image reconstruction: the first and preliminary, experiment, aimed at finding a suitable value for the local constraint parameter p1 for a single phantom of one size only. This value is set to p1=0.1; the second experiment investigated the convergence properties of PART 1 and random search, which can be seen as a limiting case of PART 1. The results demonstrate the outperformance of PART 1 in all cases except when reference images (or phantoms) are only noise, in which case, as expected, random search performs better. The final experiment provided a comparison between random search, the commonly used reconstruction algorithm, simultaneous algebraic reconstruction technique (SART), and PART 1 with p1 set to the empirical value determined in the first experiment. The result of this set of experiments demonstrated that PART 1 converges rapidly when compared to random search for phantoms with all nonzero pixel values occurring in connected regions. And in the case that there are isolated nonzero values pixels, PART 1 will find better reconstructions at fewer iterations. Additionally, PART 1 performs (statistically) significantly well when compared to random search and a standard algebraic reconstruction technique for 32×32 and 64×64 phantoms, except for the case of isolated nonzero pixel values; it is also shown that for a larger 128×128 phantom with proportionally fewer angles of projection, PART wins out over random search and SART. In this work further experiments are conducted with the focus on the important issue of reconstruction with fewer number of projections, as in practice, merely a small number of projections can be collected, thus giving rise to what is known as limited data problems. There are several reasons behind this, including cost, time, and geometrical constraints. For instance, the importance of cost reduction in the industry applications results in shortened scan duration, which in turn leads to less projections; similarly, in electron tomography, the damage caused to the sample by the electron beam reduces the number of collectable projections [15]; and in nuclear imaging, reducing the number of projection means reducing the duration in which patients should be exposed to radioactive materials as well as the inconvenience of long scanning time. This stresses that algorithms need to return sufficiently suitable approximations of the original phantoms even with smaller number of projections, which is what some of the experiments in this section are allocated to. In this section, PART 2 is contrasted against PART 1, RS, SART and FBP over all the phantoms used in this work. Methodology Forward model The acquisition geometry used for the experiments is parallel beam topology and the experiments use simulated objects (i.e. virtual phantoms). In all cases, the elements of the projection matrix were calculated from the line model. Phantoms Phantoms 1 and 2 (see Fig. 3) are commonly used in binary tomography [19] and the third phantom resembles the Jaszczak phantom used to calibrate the SPECT and PET scanning machines. The size of all the phantoms is 512 × 512. To carry out the experiments in images with different sizes, the phantoms or reference images have been scaled to create images of varying sizes (namely, 64×64 and 128×128). PART 1 & 2 PART is used with the Moore neighbourhood. There are a number of alternatives for line 1 of Algorithm 1, the selection step in PART. The purpose of this step is to find an occupied cell, a, and a vacant cell, b. The following experiments use random selection: a and b are selected uniformly at random from the sets of all occupied/unoccupied cells. A list implementation would have been efficient, but since the numbers of occupied/unoccupied cells is roughly similar, uniform sampling over the entire grid y was used due to the ease of implementation and small time overhead. Algorithms 2 and 3 specify Select for PART 1 & 2; U(y) is a uniform random selection of a single cell from the grid y. The value of the global constraint parameter p2 was fixed, in all experiments, to zero. Fig. 3 Phantom images used in the experiments. a Phantom 1, b phantom 2, c phantom 3 Random search (RS) For the purposes of these experiments, random search is defined as the PART algorithm with the neighbourhood parameter p1 set to 1 with the consequence that a particle will always attempt a move to an unoccupied cell b even if the neighbour count of b, n(b), is less than n(a). Simultaneous algebraic reconstruction technique (SART) The implementation of SART used here was based on Andersen and Kac’s algorithm, [2]. The projection angles were selected uniformly at random [3]. The value of the relaxation parameter λ was set to 1.9 in accordance with the recommendation of [14]. SART needs to be modified for binary reconstruction since in the unaltered form SART produces a continuum of pixel values. The following modifications were made: any negative pixel values occurring after updating at any angle were set to zero; the final image y after updating all projection angles was normalised so that the total pixel value count of the phantom image and the reconstructed image were equal; y was thresholded at the average pixel value so that values below the average were set to zero, values above or equal tot he average were set to 1. Filtered back projection (FBP) In order to provide a more comprehensive account to the experiments conducted in this work, FBP algorithm is also used. FBP algorithm is capable of fast and adequate reconstruction, but requires a large number of projections. FBP generates an image in a single iteration. Measure The principle performance measure is the image proximity e2=||y-x||1 (defined in Eq. 3) where, for the phantom image x,Ax=b and for the reconstructed image, Ay=c. However, while the algorithm uses the projection error e1=||b-c||1 as the objective function. This is because, in practice, x is unknown. Experiments and results In this section, phantoms of 64×64 and 128×128 are used with 8 and 5 projections respectively. In these experiments p1=0.10, p2=0.00 and 30 runs were conducted for test in order to acquire adequate statistics. The termination condition for each run is 20, 000 function evaluations (FEs). For the purposes of this study, the number of FEs does not vary with the size of the phantoms and the number of projections. In this section the five algorithms are used (e.g. PART 1 & 2, RS, SART and FBP) on Phantoms I, II, and III (see Fig. 3). The results of running the five algorithms on the three phantoms in 64×64×8 are shown in Fig. 4. The results show a clear and almost homogeneous picture on the performance of the algorithms. The algorithms’ performance ranking appear in the following order: PART 2, PART 1, RS, FBP and SART. The only exception appears in phantom 3 where SART outperforms FBP. Given the large error margin of FBP and SART, Fig. 4 does not clearly show the difference between PART 1 & 2, where Fig. 5 zooms into the graph to show the difference between these two variations of PART.Fig. 4 e2 in 64×64×8 Fig. 5 e2 in 64×64×8 for phantom 3. In the bottom plot, PART 2 is shown in green and PART 1 is highlighted in blue (color figure online) The results of running the five algorithms on the three phantoms in 128×128×5 are shown in Fig. 6. The results match the previous observations on the smaller phantoms, with the difference that SART outperforms FBP in phantoms 2 and 3. The performance ranking of the other algorithms is maintained (i.e. PART 2, PART 1 and then RS). Again, in order to visually compare the results of PART algorithms in phantom 3, Fig. 7 illustrates the difference on a few of the experiments, zooming into the graph.Fig. 6 e2 in 128×128×5 Fig. 7 e2 in 128×128×5 for Phantom 3. In the bottom plot, PART 2 is shown in green and PART 1 is highlighted in blue (color figure online) The summary of figures from which the plots are derived are reported in Table 1. This table shows the performance of the five algorithms used in this work when reconstructing three phantoms in 64×64×8 and 128×128×5 configurations.Table 1 Comparing PART2, PART1, RS, SART and FBP in 64×64×8 and 128×128×5 experiments Min Max Median Mean StDev (a) 64×64×8  Phantom 1   PART2 510 3060 1530 1445 600.21   PART1 18,360 28,560 23,460 23,749 2966.79   RS 81,090 97,410 87,975 88,281 4263.5   SART 71,145 368,475 169,575 192,508 85,173.71   FBP 156,190 156,190 156,190 156,190 0  Phantom 2   PART2 5100 31,620 16,320 17,493 6556.06   PART1 56,610 77,010 66,555 66,759 4932.56   RS 124,950 143,820 139,230 138,380 4216.93   SART 151,725 560,490 297,585 305,209.5 109,065.36   FBP 238,960 238,960 238,960 238,960 0  Phantom 3   PART2 510 3060 1020 1037 560.01   PART1 2550 10,200 4845 5389 2073.51   RS 30,090 40,800 35,190 35,275 2384.91   SART 52,020 200,430 95,880 95,557 31,399.4   FBP 337,340 337,340 337,340 337,340 0 (b) 128×128×5  Phantom 1   PART2 107,610 191,760 135,405 138,669 20,283.75   PART1 313,650 357,510 339,150 339,745 11,298.13   RS 786,930 834,360 809,370 810,662 11,662.70   SART 659,175 2,512,260 1,219,410 1,389,206 588,415.16   FBP 1,070,800 1,070,800 1,070,800 1,070,800 0.00  Phantom 2   PART2 477,870 616,590 548,760 544,561 32,452.80   PART1 609,450 736,950 676,515 678,504 29,424.83   RS 1,049,580 1,095,480 1,069,725 1,070,286 12,500.01   SART 522,750 1,840,845 785,017.5 901,340 330,794.76   FBP 1,654,200 1,654,200 1,654,200 1,654,200 0.00  Phantom 3   PART2 116,280 185,640 151,725 152,507 17,061.81   PART1 170,340 199,920 182,070 182,121 7035.41   RS 302,430 329,460 314,415 315,112 6456.23   SART 294,270 742,305 528,615 533,103 120,685.03   FBP 1,971,500 1,971,500 1,971,500 1,971,500 0.00 Table 2 Statistical analysis of the performance of the algorithms PART2–PART1 PART2–RS PART2–SART PART2–FBP (a) 64×64×8  Phantom 1 X–o X–o X–o X–o  Phantom 2 X–o X–o X–o X–o  Phantom 3 X–o X–o X–o X–o  ∑ 3–0 3–0 3–0 3–0 (b) 128×128×5  Phantom 1 X–o X–o X–o X–o  Phantom 2 X–o X–o X–o X–o  Phantom 3 X–o X–o X–o X–o  ∑ 3–0 3–0 3–0 3–0 Based on Wilcoxon 1×1 Non-Parametric Statistical Test, if the error difference between each pair of algorithms is significant at the 5 % level, the pairs are marked. X–o shows that the left algorithm is significantly outperforming its counterpart algorithm; and o–X shows that the right algorithm is significantly better than the one on the left. The figures, n – m, in the last row present a count of the number of X’s and o’s in the respective columns For the purpose of providing a more meaningful analysis and comparison of results, a statistical analysis would help identify the presence of any significant difference in the behaviour of the algorithms (i.e. finding e2) across the phantoms. For this reason, a Wilcoxon 1×1 non-parametric statistical test is deployed. Investigating Table 2 validates the previous finding and confirms that PART 2 exhibit a statistically significant difference when compared to the rest of the algorithms used in this study. This result holds both for the experiments conducted on phantoms of size 64×64 with 8 projection angles and phantoms of size 128×128 with 5 projection angles. Figures 8 and 9 present the reconstructed phantoms by FBP, SART, RS, PART1 and PART2 using two configurations 64×64×8 and 128×128×5.Fig. 8 Reconstructed phantoms in 64×64×8. From left to right original phantoms, FBP, SART, RS, PART1, PART2 Fig. 9 Reconstructed phantoms in 128×128×5. From left to right original phantoms, FBP, SART, RS, PART1, PART2 These results indicate that the proposed algorithm performs better than the rest of the algorithms when a small number of projection angles are deployed. In contrast to the previous experiments reported in [1], where a larger number of projections where used (i.e. α=n/2, where α is the number of projection angles, and n is the number of pixels in the phantom), in these experiments a small number of projections where deployed, therefore adding to the undeterministic nature and thus complexity of the problem. Despite this, the proposed algorithm is exhibiting a competitive performance. To verify the strength of PART and explore the reduction in the number of projections, a set of experiments are designed and the results are reported in the next section. Impact of the number of projection angles In this section one phantom is used with the image size of 64×64. The aim of this experiment is to investigate the role of the number of projection angles on the performance of the reconstructing algorithms. The three algorithms picked are PART 2 along with the classically used SART and FBP algorithms. Both SART and FBP are not able to improve their reconstructed images after the end of the iteration. However, PART in principle, can iteratively reconstruct the phantoms. In this experiment each algorithm is run 10 times and the termination criteria for PART 2 is reaching 50,000 FEs. The plots in Fig. 10 clearly illustrate the impact of the number of projections on the quality of the reconstructed images (measured using e2). In this set of experiments, FBP shows a clear impact of α on its performance (i.e. the higher the number of projections, the smaller the error); this picture changes slightly in SART where for example, in several runs, the error in α=2 is better than α=4. Consistent with the previous experiments where PART 2 is outperforming the other algorithms, the algorithm still maintains its superiority in terms of the resultant e2. Table 3 presents the summary of the numerical values of these experiments.Table 3 Varying values of α for PART2, SART and FBP in 64×64 Min Max Median Mean StDev PART 2  α=2 54,570 78,030 64,005 65,178 6189.49  α=4 1020 23,460 7905 9027 6495.91  α=8 510 1020 510 561 161.28  α=16 510 1020 510 612 215.03  α=32 510 1020 510 612 215.03 SART  α=2 171,105 171,105 171,105 171,105 0.00  α=4 118,575 236,640 158,100 176,817 51,089.77  α=8 55,845 166,515 79,943 88,077 31,420.86  α=16 6630 18,360 10,965 11,118 3492.66  α=32 1275 4845 1658 2015 1084.87 FBP  α=2 1,462,900 1,462,900 1,462,900 1,462,900 0.00  α=4 717,020 717,020 717,020 717,020 0.00  α=8 337,340 337,340 337,340 337,340 0.00  α=16 144,530 144,530 144,530 144,530 0.00  α=32 59,328 59,328 59,328 59,328 0.00 One of the interesting observations in the PART 2 plot is the presence of some instances where α=4 finds equally good error values in comparison with α=8,16,32. This suggests that PART 2 is less dependant on the value of α and can perform well even in cases where only smaller number of projections can be obtained in the real-world experiments and clinical setups. Figure 11 shows a closer view on PART 2 with α=4,8,16,32. The error of 510 shared by most trials with α=8,16,32 means that only two pixels are misplaced (i.e. two white pixels, 510=2×255).Fig. 10 Varying number of projection angles (α) for PART2, SART and FBP Fig. 11 PART2: Varying number of projection angles (α) Conclusions This paper extends the previously introduced particle aggregate reconstruction technique (PART) with emphasised focus on the aggregation of isolated particles. PART is based on the idea that an image can be interpreted as a grid of cells populated by particles. Pixel values represent cell occupancy; particles are mobile and diffuse throughout the grid by random jumps, preferably landing adjacent to regions of increased particle density. The algorithm is intuitive, and easily implemented. This work also puts a particular emphasis on the reduction in the projection angles and therefore aiming to use less data to reconstruct the phantom images. A number of experiments were designed based on three phantoms in two sizes of 64×64 and 128×128 and the proposed variation of PART algorithm is contrasted against FBP, simultaneous algebraic reconstruction technique (SART) as well as random search (RS). Based on the results and in terms of the error, the dominance of PART over the aforementioned techniques is suggestive. Furthermore, the results demonstrate the statistically significant outperformance of PART in all instances. An experiment was designed to show the impact of the number of projections on the reconstruction quality of the phantom images. It is shown that PART is less sensitive to the number of projection angles, making the algorithm attractive when less data are available, or in situations where collecting less projection data are inevitable (i.e. in medical scenarios where patients cannot be kept for long duration for the scanning purposes, or where long exposure to radiation is lethal). One of the main research questions is whether aggregation by particle diffusion can be extended to the general discrete case, which is the topic of ongoing research. 1 Note that the boundary might lie within the cluster i.e. bounding a hole. 2 PART source code can be downloaded from http://doc.gold.ac.uk/~map01mm/PART/. ==== Refs References 1. al-Rifaie MM, Blackwell T (2016) Binary tomography reconstruction by particle aggregation. In: Applications of evolutionary computation: 19th European conference, EvoApplications 2016, Porto, Portugal, March 30–April 1, 2016, proceedings, part I. Springer, Cham, pp 754–769. doi:10.1007/978-3-319-31204-0_48 2. Andersen AH Kak AC Simultaneous algebraic reconstruction technique (SART): a superior implementation of the ART algorithm Ultrason Imaging 1984 6 81 94 10.1177/016173468400600107 6548059 3. Batenburg K Sijbers J DART: a practical reconstruction algorithm for discrete tomography IEEE Trans Image Process 2011 20 9 2542 2553 10.1109/TIP.2011.2131661 21435983 4. Block KT Uecker M Frahm J Undersampled radial MRI with multiple coils. Iterative image reconstruction using a total variation constraint Magn Reson Med 2007 57 6 1086 1098 10.1002/mrm.21236 17534903 5. Bruyant PP Analytic and iterative reconstruction algorithms in spect J Nucl Med 2002 43 10 1343 1358 12368373 6. D’Ariano G Presti PL Quantum tomography for measuring experimentally the matrix elements of an arbitrary quantum operation Phys Rev Lett 2001 86 19 4195 10.1103/PhysRevLett.86.4195 11328133 7. Gordon R Bender R Herman GT Algebraic reconstruction techniques (ART) for three-dimensional electron microscopy and X-ray photography J Theor Biol 1970 29 3 471 481 10.1016/0022-5193(70)90109-8 5492997 8. Herman GT Fundamentals of computerized tomography: image reconstruction from projections 2009 Germany Springer 9. Herman GT Kuba A Advances in discrete tomography and its applications 2008 Germany Springer 10. Jiang M, Wang G (2001) Convergence of the simultaneous algebraic reconstruction technique (SART). In: Conference record of the thirty-fifth Asilomar conference on signals, systems and computers, 2001. vol 1, pp 360–364. doi:10.1109/ACSSC.2001.986951 11. Joseph PM An improved algorithm for reprojecting rays through pixel images IEEE Trans Med Imaging 1982 1 3 192 196 10.1109/TMI.1982.4307572 18238275 12. Kaczmarz S Angenaherte auflosung von systemen linearer gleichungen Bull Int Acad Polon Sci Lett A 1937 35 335 357 13. Kak AC Slaney M Principles of computerized tomographic imaging 2001 USA Society for Industrial and Applied Mathematics 14. Kazemini E Nedialkov N Taveres JMRS Natal JRM An empirical study of algebraic reconstruction techniques Computational vision and medical image processing IV 2014 UK Taylor and Francis 93 98 15. Midgley PA Dunin-Borkowski RE Electron tomography and holography in materials science Nat Mater 2009 8 4 271 280 10.1038/nmat2406 19308086 16. Nolet G (1987) Seismic wave propagation and seismic tomography. In: Nolet G(ed) Seismic tomography. Springer, Germany 17. Reynolds CW (1987) Flocks, herds and schools: a distributed behavioral model. In: ACM siggraph computer graphics, vol 21. New York, USA, pp 25–34 18. Ter-Pogossian MM (1983) Positron emission tomography (PET). In: Nolet G(ed) Diagnostic imaging in medicine. Springer, Germany 19. Van Dalen B Stability results for uniquely determined sets from two directions in discrete tomography Discrete Math 2009 309 12 3905 3916 10.1016/j.disc.2008.11.018 20. Witten TA Sander LM Diffusion-limited aggregation, a kinetic critical phenomenon Phys Rev Lett 1981 47 1400 1403 10.1103/PhysRevLett.47.1400

PMC005xxxxxx/PMC5002049.txt

==== Front Carbon Balance ManagCarbon Balance ManagCarbon Balance and Management1750-0680Springer International Publishing Cham 5910.1186/s13021-016-0059-4ResearchModelling forest carbon stock changes as affected by harvest and natural disturbances. II. EU-level analysis http://orcid.org/0000-0003-4541-7056Pilli Roberto +39 0332 785655roberto.pilli@jrc.ec.europa.eu 1Grassi Giacomo giacomo.grassi@jrc.ec.europa.eu 1http://orcid.org/0000-0003-4576-7849Kurz Werner A. werner.kurz@canada.ca 2Moris Jose V. joseantonio.vazquezmoris@unito.it 3Viñas Raúl Abad raul.abad-vinas@jrc.ec.europa.eu 11 European Commission, Joint Research Centre, Directorate D – Sustainable Resources - Bio-Economy Unit, Via E. Fermi 2749, 21027 Ispra, VA Italy 2 Natural Resources Canada, Canadian Forest Service, Victoria, BC V8Z 1M5 Canada 3 Department of Agriculture, Forest and Food Sciences, University of Torino, Via Leonardo Da Vinci 44, 10095 Grugliasco, TO Italy 26 8 2016 26 8 2016 12 2016 11 2016 2 2016 12 8 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Background Forests and the forest sector may play an important role in mitigating climate change. The Paris Agreement and the recent legislative proposal to include the land use sector in the EU 2030 climate targets reflect this expectation. However, greater confidence on estimates from national greenhouse gas inventories (GHGI) and more comprehensive analyses of mitigation options are needed to seize this mitigation potential. The aim of this paper is to provide a tool at EU level for verifying the EU GHGI and for simulating specific policy and forest management scenarios. Therefore, the Carbon Budget Model (CBM) was applied for an integrated assessment of the EU forest carbon (C) balance from 2000 to 2012, including: (i) estimates of the C stock and net CO2 emissions for forest management (FM), afforestation/reforestation (AR) and deforestation (D), covering carbon in both the forest and the harvest wood product (HWP) pools; (ii) an overall analysis of the C dynamics associated with harvest and natural disturbances (mainly storms and fires); (iii) a comparison of our estimates with the data reported in the EU GHGI. Results Overall, the average annual FM sink (−365 Mt CO2 year−1) estimated by the CBM in the period 2000–2012 corresponds to about 7 % of total GHG emissions at the EU level for the same period (excluding land use, land-use change and forestry). The HWP pool sink (−44 Mt CO2 year−1) contributes an additional 1 %. Emissions from D (about 33 Mt CO2 year−1) are more than compensated by the sink in AR (about 43 Mt CO2 year−1 over the period). For FM, the estimates from the CBM were about 8 % lower than the EU GHGI, a value well within the typical uncertainty range of the EU forest sink estimates. For AR and D the match with the EU GHGI was nearly perfect (difference <±2 % in the period 2008–2012). Our analysis on harvest and natural disturbances shows that: (i) the impact of harvest is much greater than natural disturbances but, because of salvage logging (often very relevant), the impact of natural disturbances is often not easily distinguishable from the impact of harvest, and (ii) the impact of storms on the biomass C stock is 5–10 times greater than fires, but while storms cause only indirect emissions (i.e., a transfer of C from living biomass to dead organic matter), fires cause both direct and indirect emissions. Conclusions This study presents the application of a consistent methodological approach, based on an inventory-based model, adapted to the forest management conditions of EU countries. The approach captures, with satisfactory detail, the C sink reported in the EU GHGI and the country-specific variability due to harvest, natural disturbances and land-use changes. To our knowledge, this is the most comprehensive study of its kind at EU level, i.e., including all the forest pools, HWP and natural disturbances, and a comparison with the EU GHGI. The results provide the basis for possible future policy-relevant applications of this model, e.g., as a tool to support GHGIs (e.g., on accounting for natural disturbances) and to verify the EU GHGI, and for the simulation of specific scenarios at EU level. Keywords EUCarbon Budget ModelForest managementAfforestationDeforestationHarvestNatural disturbancesGHGIissue-copyright-statement© The Author(s) 2016 ==== Body Background An effective role of forests in climate change mitigation requires a comprehensive assessment, from scientific and policy perspectives. From a scientific point of view, recent studies demonstrate the relevance of biophysical aspects of the forest-climate interactions, that may be important locally or in specific time frames [1]. Even if the recent paper by Naudts et al. [2], casting doubts on the role of European forests in mitigating climate change over the last centuries, there is increasing and largely consistent scientific evidence that forests in Europe are currently making a relevant and positive contribution to climate change mitigation (see [3, 4]). From a policy perspective, it is relevant to understand how this contribution may be efficiency translated into different mitigation options, including the sink in the forest, the sink outside the forest (in harvested wood products, HWP) and the use of wood for energy and material substitution [5]. Given the heterogeneity of the European forest system, assessing the specific regional circumstances, opportunities and challenges is fundamental [6]. At the same time, maximizing the sum of these mitigation options requires an integrated, dynamic modeling framework to quantify in a robust way the unavoidable trade-offs (e.g. between the forest sink and the bioenergy), which are often not appropriately considered [7]. Furthermore, for such a modeling framework to be directly policy-relevant, the policy context such as the rules for reporting and accounting emissions and removals from forests need to be taken into account. In particular, the current rules under the Kyoto Protocol (KP) significantly changed for the second commitment period (CP2, 2013–2020) (see, [8–10]). According to these new rules, as reflected in the latest IPCC guidance [11]: (i) the reporting and accounting of forest management (FM, i.e., land in the forest land use category in 1989) are now mandatory (through a ‘forest reference level’), in addition to the already mandatory accounting of afforestation/reforestation and deforestation (AR and D, i.e., forest land-use changes since 1990); (ii) the accounting of FM shall include the carbon (C) stock changes in the HWP pool; and (iii) emissions and subsequent removals from natural disturbances may be excluded from the accounting under certain conditions. While some further change is foreseen under the proposed post-2020 EU regulation on land use and forestry [12], most of these rules are expected to continue (e.g. forest reference level, HWP, natural disturbances). The greenhouse gas inventories (GHGIs) represent the basis to assess the compliance of any climate mitigation target. The GHGI of the EU, submitted annually to the United Nations Framework Convention on Climate Change (UNFCCC) and its KP, is the sum of the inventories of 28 Members States (MS), which include about 158 Mha of forests [13]. The species composition, the current and past management practices, the amount of natural disturbances, and the quality and type of information available on the forest resources, differ among countries. Moreover, conceptual and methodological differences in countries’ GHGIs produce discrepancies in the resulting estimates that are currently not entirely addressed and require further work to achieve reliable and consistent estimates throughout Europe [14, 15]. The complex EU forest sector can be represented using a process-based approach (e.g. [1, 16]) or using, for each country, empirical forest-inventory based models (e.g. [17]). Traditionally, process-based models have mainly been used to simulate the long-term evolution of forest C dynamics at large scales, including the potential effects of climate change [18], but they generally do not include a detailed analysis of forest management practices. Therefore, empirical models still remain the primary tool to simulate the detailed effects of different management options on short-term forest C dynamics [19, 20] at small to medium spatial scales (e.g. from forest stands to countries). When compiled from regional or country level scales to a continental scale, the empirical model results can be compared with the data produced through process-based approaches [21, 22], and can provide additional information on the main drivers of forest carbon dynamics at the EU scale. Different forest-inventory based models were used in the European context, to estimate the future forest C sink under different policies and management scenarios (i.e., [20]) or the impact of natural disturbances on the forest C stock [23] or the realizable potential supply of woody biomass [24]. None of these studies, however, considered in a comprehensive way the overall EU forest C sink consistently with the current international reporting and accounting regulations, i.e., including FM, AR and D, HWP and natural disturbances. In a recent study, the Carbon Budget Model (CBM), developed by the Canadian Forest Service [25] was applied to 26 EU MSs to model the forest C dynamics from FM at the country level for the period 2000–2012, including the impact of the major natural disturbances [26]. In that study, after having validated the CBM results for a representative country, the country-specific results were evaluated against the individual 2014 GHGIs submitted to the UNFCCC by each EU MS. This evaluation is an essential pre-requisite to analyze the overall EU forest C balance and the level of confidence on the EU GHGI. Achieving this confidence is key to allow the forest sink to be included in the EU climate target [27]. The goal of this paper is to provide a tool for verifying the whole EU GHGI and for simulating specific policy and forest management scenarios at EU level. In particular, with the present paper, focused at the EU level and largely based on the same methodological assumptions used by Pilli et al. [26], we aim to: (i) estimate the C stock and the CO2 emissions and removals for FM, covering the carbon both in the forest pools (total living biomass, dead organic matter, mineral soil) and in HWP pool; (ii) estimate the CO2 emissions and removals for forest land-use changes (i.e., AR and D); (iii) provide an overall EU-level analysis of the impacts of harvest and natural disturbances (mainly storms and fires); and (iv) compare our estimates with the estimates reported in the EU GHGI and other continental-scale studies. Results and discussion The aggregated results at the EU level for the forest-related activities defined by the KP are reported for FM (i.e., forest existing in 1989) AR and D (i.e., forest and land-use changes since 1990) in “Forest C stock (2008–2012)” section. Here, the results obtained using the CBM are compared with the data reported in the 2014 EU GHGI1 [13]. According to the EU GHGI, the two MSs not considered in this study, Cyprus and Malta, provide a negligible contribution to the EU forest sink (0.02 %). The resulting forest C dynamics are described in “Forest C dynamics (2000–2012)” section. Here, the data are reported from an atmospheric perspective, where negative values represent a sink (CO2 removals) and positive values a source (CO2 emissions). Results cover only CO2 and exclude organic soils. Even if emissions from drained organic soils and non-CO2 emissions from forest fires may be relevant in some countries (e.g., [28, 29]), at the EU level they account for 5 and 2 % (in terms of CO2-eq.), respectively, of net annual forest sink [13]. In “Main drivers determining forest C dynamics” section, we discuss the main drivers determining the forest C sink dynamic, further distinguished between harvest (“Harvest” section) and natural disturbances (“Natural disturbances” section). Forest C stock (2008–2012) The average C stock per hectare estimated for FM by the model for the period 2008–2012 (i.e., the First Commitment Period, CP1, of the KP), is equal to 142.3 Mg C ha−1 at the EU level, including 68.4, 19.3 and 54.5 Mg C ha−1, for living biomass, DOM and mineral soil, respectively (Table 1). According to Pilli et al. [30], the total C stock of HWP is about 1921 Tg C (average for the CP1), which compares the total C stock in living biomass of 9437 Tg C in this study.Table 1 Average C stock estimated for FM by the CBM model for the period 2008–2012 (KP-CP1) at EU level Average historical (2008–2012) C stock DOM Soil Living biomass Total ecosystem Deadw. Litter Abovegr. Belowgr. Tot Liv. Biom. C stock Av. (Mg C ha−1) 10.8 8.5 54.5 56.0 12.4 68.4 142.3 Total (Tg C) 1493 1177 7524 7720 1717 9437 19,632 Overall, our estimates on living biomass are in good agreement with the data reported by most of other reports or studies. The Global Forest Resource Assessment 2010 [31] reports for 2010 in Europe, excluding the Russian Federation but including some non-EU countries (i.e., for a total forest area that is about 8 Mha larger than the forest area considered by our study), an average C stock equal to 63.9, 18.6 and 96.6 Mg C ha−1, for the living biomass, dead wood plus litter and soil (including peat), respectively. The State of Europe’s Forests 2015 [32] reports for EU-28 and assuming a forest area slightly larger than our study (+9 %) a biomass C stock in 2010 equal to 58.8 Mg C ha−1. Providing estimates for dead wood and litter C stock is difficult [33]. Verkerk et al. [34] applied the EFISCEN model to 24 EU MS (i.e., the same considered by our study, except Croatia and Greece) and estimated an average amount of deadwood of 12.3 t ha−1 in 2005. Assuming a C content of 0.5, this equals an average C stock of 6.1 t C ha−1, about 40 % lower than our estimate. An accurate assessment of the soil C stock is also difficult due to the range of available model and inventory results [33, 35]. Our estimate of the total C stock in forest mineral soils (7524 Tg C) is intermediate between the values reported by other studies: about 5000 Tg C reported by Liski et al. [36] and 13,700 Tg C (including the O-layer) estimated by Goodale et al. [37]. Since the soil C stock is affected by both natural and anthropogenic factors [38, 39], comparing the average C stock estimated by our model, equal to 54.5 Mg C ha−1, with other studies is even more difficult. Shulp et al. [40] report, for the Netherlands, a mean C stock in mineral soil between 53 and 97 Mg C ha−1, with significant statistical differences mainly due to the species composition. The average C stock in mineral soil estimated by the last Italian National Forest Inventory (NFI) (concluded in 2009) varies between 68 and 96 Mg C ha−1, depending by the species composition [41], while the Swedish NFI estimated, for 2000, an average C stock for coniferous forest soil, equal to 73 ± 10 Mg C ha−1 [42]. Forest C dynamics (2000–2012) The sum of the net CO2 removals from all land-use activities and carbon pools considered (FM + AR + D + HWP) is, on average, equal to −409 Mt CO2 year−1 between 2000 and 2012 (Fig. 1, panel c). This corresponds to about 8 % of the total GHG emissions in the EU for the same period (without LULUCF). This amount may be further distinguished between different land-use activities (FM, AR and D) and pools. About 90 % of the total C sink is due to FM (Fig. 1, panel a, including HWP), while AR (Fig. 1, panel b) contributes to the remaining 10 %, with an increasing fraction due to the ageing of the new forest area. Deforestation is a source by about 33 Mt CO2 year−1 between 2000 and 2012. About 80 % of the total C sink (after subtracting emissions from D) is due to the living biomass pool (70 % accounted as FM and 10 % as AR), 10 % is due to DOM (mainly accounted as FM) and the remaining 10 % is related to the HWP pool.Fig. 1 EU-level net CO2 emissions (in Mt CO2 year−1) for: a Forest management (FM), as estimated by the CBM for the living biomass, DOM and soil pools, by Pilli et al. [30] for the HWP pool and as reported for the living biomass pool in the EU GHGI (see [13] and the “Methods” for further details); b afforestation/reforestation (AR) and deforestation (D) since 1990, as estimated by the CBM for living biomass and all the pools (total) and as reported (all pools) in the KP CRF tables of the EU GHGI; c sum of FM, AR and D, as estimated for all the pools by the CBM (with and without HWP) and as reported in the EU GHGI (without HWP). Organic soils are always excluded from GHGIs, to allow a more consistent comparison with CBM. All the values are reported from an atmospheric perspective, i.e., negative values represent a sink and positive a source Within FM, the highest inter-annual variations (due to harvest and natural disturbances) were estimated for the living biomass pool, varying from −237 Mt CO2 year−1 in 2000 to −311 Mt CO2 year−1 in 2012 (Fig. 1, panel a). As expected, the DOM pool in the CBM has the opposite trend, because natural disturbances such as storms, fires and insect attacks transfer carbon from biomass to DOM pools (see for example 2000, 2005 and 2007) from where the carbon will be released to the atmosphere through subsequent decomposition. For this pool, we estimated an average C sink, for the entire period, equal to −43 Mt CO2 year−1. For the mineral soil we estimated a modest and rather stable C sink over the entire period, equal on average to −3 Mt CO2 year−1. This trend is consistent with increasing biomass of the EU forests, which means increasing inputs from litter and dead wood to the soil pool, and with the short time horizon considered, i.e., the process of soil C accumulation is typically a slow process. This process is simulated by the CBM through a series of biomass annual turnover rates and transfer rates [25]. Similarly to other soil models [43] the results provided by CBM may be influenced by uncertainty in the model initialization that may directly affect the estimate of the C stock change of this pool [44]. In terms of C stock change, we estimate average values, for the entire period, equal to 0.01, 0.08 and 0.60 Mg C ha−1 year−1 respectively for living biomass, DOM and soil. The CO2 sink of the FM living biomass pool estimated by the CBM is about 12 % lower than the data reported in the EU GHGI2 (see Fig. 1, plot a) and is in line with most of other studies with similar area and time frames (e.g. [45, 46]). The largest differences with the GHGI, in 2000, 2005 and 2007, are related to the different assumptions about the impact of natural disturbances (mainly storms that occurred in central and northern European countries). Indeed, the effect of natural disturbances on the redistribution of C among pools (from living biomass to DOM), well represented in the CBM, is often not evident in the GHGIs [22, 26]. The difference between the CBM and the EU GHGI is reduced to 8 % when DOM and mineral soil are also considered. Since the uncertainty of CO2 estimates for “forest land remaining forest land” at the EU level is around 18 % [13],3 and given the fact that methods to estimate emissions/removals by the CBM are largely independent from those of the EU countries [26], we consider the match between the CBM results and the EU GHGI to be satisfactory. The increasing discrepancy in more recent years is mainly due to few countries (mainly Poland and France), which will deserve further investigation, e.g. inappropriate data or assumption used by the CBM model or problems with the GHGIs. Indeed, at country level, where updated NFI data are available and the model’s assumptions on harvest and natural disturbances are consistent with the countries’ input data, the estimates provided by the CBM are generally consistent (both in the trend and in the amount) with the GHGI data [26]. For AR (Fig. 1, panel b), the living biomass C sink gradually increases until 2003, due to the constant annual rate of AR prior to 2008 (see Fig. 7 in the “Methods” section). We estimated a very small source for the atmosphere from DOM and soil, due to the effect of afforestation on the soil pool during the first years [47]. As explained in the methods, the biomass C sink is directly related to the values reported by the yield tables applied by model. In most cases, due to the young age of the forests that were afforested since 1990, we assumed that no silvicultural treatment was applied to broadleaved species stands younger than 15 years and to coniferous stands younger than 20 years. The only exception was for Portugal’s Eucalyptus plantations, where we assumed a minimum rotation length of 12 years. Due to the effect of these treatments, the biomass sink has a first step in 2003–2004 (due to the harvest on Eucalyptus plantations), and a second step in 2009–2010, due to the first harvest applied to coniferous plantations. Overall, the total amount of harvest provided by AR is negligible, equal to about 6.3 Mm3 in 2012, i.e., about 1.2 % of the total amount of harvest obtained from FM at the EU level. The total annual C sink (−54 Mt CO2 year−1) provided by AR for the CP1 (2008–2012) is equal to about 1.2 % of the total GHGs in the EU for the same period (without LULUCF). For deforestation (Fig. 1, panel b), the CBM estimates the loss of C from the living biomass, DOM and soil pools based on the areas subject to deforestation (taken from countries’ GHGIs). Overall, for the CP1 (2008–2012), these emissions equal about 0.7 % of the total GHGs in the EU for the same period (without LULUCF). Our estimates for AR and D compare very well with the EU GHGI for CP1 (2008–2012) [48]. While emissions from drained organic soils may be important in some MS (e.g. Finland, Sweden, Ireland) at the EU level the impact of organic soils for AR and D is small (about 2–3 Mt CO2 year−1). Overall, this good match with the GHGIs is expected, because the CBM uses the same rates of AR and D areas as reported in the GHGIs and because of the good agreement for the estimates of biomass carbon densities. For AR, however, a certain degree of independence between CBM and GHGIs arises from the choice of the yield tables and the harvest assumptions. Main drivers determining forest C dynamics The forest C sink is essentially the difference between the net increment and the losses, i.e., harvest and natural disturbances. Forest growth, and the evolution of net annual increment4 over time is estimated during the model run, by combining, for each country and time step (i.e., year), a yield table library based on the NFI annual increment with the forest inventory and its age class distribution (see [25, 44]). Given that the net increment typically changes relatively slowly and in recent years appears rather stable at EU level [49] here we focus on harvest (Fig. 2, panel a) and natural disturbances, such as storms and ice (panel b) and fire (panel c).Fig. 2 a Harvest-related C dynamics (in Mt CO2 year−1, reported on the left axis) and volume removed according to our model (in million m3, on the right axis), in terms of fuelwood (FW) immediately released into the atmosphere and industrial roundwood (IRW) moved from forest to HWP; b indirect total CO2 emissions (in Mt CO2 year−1, reported on the left axis) due to storm and insect attacks (reported as volume, in million m3, on the left axis), further distinguished between the amount of biomass moved from living biomass to DOM pool (black dots) due to the disturbances and directly recovered as salvage logging (white dots); c CO2 emissions (in Mt CO2 year−1) due to fires, distinguished between direct—i.e. immediately released into the atmosphere (according to our estimates and compared with the data reported in the EU GHGI [13])—and indirect emissions, i.e. moved from biomass to DOM pool (black dots), from where it will decompose in subsequent years Harvest The CBM represents the amounts used for fuelwood (FW) and for industrial roundwood (IRW) (Fig. 2, panel a). According to the IPCC [11] and the UNFCCC [50], for the Second Commitment Period of the KP (KP-CP2) the C in the FW has to be accounted as a direct CO2 emission into the atmosphere, while the C in the IRW products has to be further quantified to estimate the C stock changes in the HWP pool, including product categories. End-of life disposal of HWP in landfills is considered an instantaneous oxidation. According to our estimates using the CBM model, the direct emissions related to the FW continuously increased, from about 82 Mt CO2 year−1 in 2000 to about 100 Mt CO2 year−1 in 2012 (i.e., +2 % year−1). Based on the country-specific assumptions applied by CBM, the FW may be provided by: (i) direct harvest removals, i.e., specific silvicultural treatments applied to forest stands (e.g., clearcuts on coppices or commercial thinnings on high forests); (ii) indirect harvest of branches, other wood components and snags during other silvicultural treatments (i.e., thinnings and clearcut where the merchantable biomass is used as IRW); and (iii) the salvage logging after disturbance events (mainly fires). The amount of IRW that was moved from living biomass to the HWP pool was equal, on average, to 309 Mt CO2 year−1 between 2000 and 2012. The two peaks reported in 2005 and 2007 are due to salvage logging after the storms in 2005 and 2007 (see Fig. 2, panel b). Interestingly, no major peak is reported by the statistics after the big storm that occurred in 1999/2000 and after the storms occurred in some countries (i.e., Austria, Estonia and, above all, France) in 2009, suggesting that salvage logging either was not very relevant or it was spread over more years. According to Rüter [51], the analysis of IRW data to obtain C stock changes in the HWP pools involves the service life of products (i.e., the annual decay rate), the estimate of the domestic production, the balance between C inflow and outflow from the HWP pool, the exclusion of harvest from deforestation, and many other factors. Taking into account all these factors and applying the same harvest rate used in the present study, Pilli et al. [30] have applied the IPCC Tier 2 production approach [11] to estimate the HWP mitigation potential at the EU level. The resulting net CO2 sink in the HWP pool is equal on average to −44 Mt CO2 year−1 between 2000 and 2012. The ratio between the IRW C sink and the FW direct emissions to the atmosphere (on average 93 Mt CO2 year−1 between 2000 and 2012) is equal, on average, to 0.48 at EU level but it varies between countries, as highlighted by the labels in Fig. 3.Fig. 3 HWP mitigation contribution to the forest management (FM) C sink (colors as per legend). The numbers report the ratio between the industrial roundwood (IRW) C sink and the fuelwood (FW) direct emissions to the atmosphere The sum of the net CO2 emissions of the forest pools plus the HWP (as estimated by [30]) is, on average, equal to −409 Mt CO2 year−1 between 2000 and 2012 (Fig. 1, panel c). Our estimates on HWP at the EU level are very similar to the data submitted by the countries to the KP [52], and indicate that the HWP mitigation contribution is currently equal to about 10 % of the total forest net CO2 emissions at the EU level. Pan et al. [53] estimated that at the global level, the C sequestration in HWP accounted for 8 % of the total C sink in established forests. As expected, at the EU level this percentage is higher and in five out of 26 countries the contribution of the HWP pool to the total FM C sink is >20 %, for the historical period 2000–2012 (see Fig. 3). However, because these estimates are based on the IPCC production approach, the C sink in HWP for countries with large exports is attributed to the country of harvest, i.e. where the wood originated and which may not be the country where the wood is in use. As highlighted by Pilli et al. [30], in the future, the current HWP sink can be maintained either by further increasing (on average by 1 % per year) the current harvest, or by shifting more of the harvest to long-lived products [54]. In some countries, this contribution is negligible compared to the total forest C sink. In four countries where the IRW pool is a C source, the HWP pool has negative impacts on the overall C sink. This highlights the need to consider the specific national circumstances, when analyzing the possible contribution of the HWP C pool as potential mitigation tool. Natural disturbances Storms and ice The overall C dynamics related to storms and ice are shown in Fig. 2, panel B. These disturbances do not produce any direct emission of CO2 to the atmosphere, but they cause a transfer of C from the living biomass to both the DOM pool and the HWP pool (due to direct salvage logging). This process, reported in detail at the country level in Fig. 4, is simulated by the CBM through disturbance matrices for each disturbance type applied to each country (according to available information from the literature). Disturbance matrices quantify the proportion of C that is moved: (i) from the living biomass to DOM and (ii) from the living biomass to the HWP pool (see Fig. 5). Indeed, a consistent fraction of the living biomass damaged by these events is removed as salvage logging immediately after the disturbance (i.e., in the same year) or few years later (see [26, 55, 56]). The storm that occurred in 1999/2000 (reported as 2000 in Figs. 2, 4 and 5), with a total C stock transfer from the living biomass to DOM of about 150 Mt CO2, caused opposing peaks in the living biomass and DOM pools (Fig. 1, panel a). The same effect is clearly reported for the other major disturbances (e.g., 2005, 2007 and 2009). On average, we estimated that, between 2000 and 2012, about 34 Mt CO2 year−1 were moved from the living biomass to DOM and to HWP due to the effects of storms; excluding the amount of biomass directly removed as salvage logging, this amount decreases to 21 Mt CO2 year−1.Fig. 4 The map reports, for each year and country, the amount of living biomass C stock (expressed as Mt CO2 year−1, even if these are not direct CO2 emissions to the atmosphere) damaged by storm, ice and insect disturbances, as estimated by our model Fig. 5 Salvage logging after storms, ice and insect disturbances, as estimated by the CBM between 2000 and 2012, reported as relative amount of C removed from forest, with the maximum amount reported for Sweden (2005) equal to 42.2 Mt CO2 The main storms at the EU level (the area affected by ice is negligible compared with storms) occurred in 1999/2000 and 2005 (see Figs. 2, 8, panel b in the Material). In the first case (which we count as 2000), the so called storms “Lothar” and “Martin”, occurred on 27th–28th December 1999, and affected mainly France and Germany (Fig. 4). According to Gardiner et al. ([55], Appendix 3) between 184 and 204 Mm3 were directly damaged by these events. Based on our model run, about 170 Mm3 were damaged. In 2005, different storms affected many European countries, including “Gudrun” and “Erwin”, which damaged about 75 million m3 in northern Europe (see [55], Appendix 3). Overall, between 2000 and 2012, about 272,000 ha year−1 (based on the data collected by our study) were affected by storms in EU countries and, according to our estimates, on average 36.5 Mm3 year−1 were damaged. The direct salvage of storm residues simulated by CBM was equal, on average, to 13 Mm3 year−1 but with high interannual variability (see Fig. 5). Indeed, when large disturbances occur, we cannot expect that all the biomass affected by storms will be removed during the same year (this is the case for the two main disturbances that occurred in 1999–2000 and 2005). A fraction of this biomass will be removed during the following years (see for example the case of Germany in Fig. 5 between 2000 and 2003). A further amount of salvage logging may be recovered through the normal silvicultural practices (i.e., thinnings, clear-cuts, etc.) applied by the model at the country level but this amount cannot be directly estimated. On average, excluding the largest disturbance events (i.e., 2000 and 2005), we estimated that about 40 % of the total biomass damaged by storms (including branches and other wood components) was directly removed as salvage logging between 2000 and 2012. Fire The second major disturbance type considered by our study is wildfire and slash burning (if relevant). Average annual direct emissions to the atmosphere due to the burning of biomass and dead organic matter are equal to 5.27 Mt CO2 year−1 between 2000 and 2012 (Fig. 2, panel c). As for the storms, strong inter-annual variations are related to the amount of area burned with peaks in 2003, 2005 and 2007, mainly concentrated on the Mediterranean countries (see Figs. 6, 8, panel c). In some cases (Slovak Republic and Lithuania), the emissions reported in Fig. 6 are mainly due to burning of harvesting residues.Fig. 6 The map reports, for each year and country, the direct CO2 emissions to the atmosphere (in Mt CO2 year−1) due to fire disturbances, as estimated by our model. Where no direct fire emissions is reported, they were assumed as negligible, according to the information reported by each country on the National Inventory Reports submitted to UNFCCC Based on our assumptions, the total emissions due to fires are equal, on average, to 15 % of the total indirect losses due to the effect of storms. Indeed, while storms have a clear effect on the forest living biomass pool (Fig. 1, panel a), the effect of fires on this pool appears, at the EU level, much less evident. However, while storms do not contribute to direct CO2 emissions and a considerable amount of wood is recovered through salvage logging, fires do cause direct emissions to the atmosphere and the amount of salvage logging varies considerably among countries. For example, in Portugal a substantial amount of wood may be recovered as salvage logging after fire disturbances, while this practice is negligible in Greece and Italy [26]. This explains the peak on transfers of living biomass to DOM reported in 2007 (see Fig. 2, panel c), when fires mainly occurred in Italy and Greece. Overall, our estimates are consistent with the total CO2 emissions reported in the EU GHGI, except for few a years (e.g. 2000, 2003 and 2007). A full comparison between the CBM and the EU GHGI however is difficult because of the different assumptions and methodologies used by some MS (see [13, 14]). Conclusions This paper analyzed the CBM results at the EU level, in terms of C emissions and removals for all forest activities foreseen by the Kyoto Protocol: forest management (i.e. land in the forest land-use category in 1989), afforestation/reforestation and deforestation (, i.e. forest related land-use changes since 1990). We considered all forest carbon pools (including HWP for FM) and we analyzed the main drivers of forest C stock and sink dynamics over the period 2000–2012 (i.e., harvest, natural disturbances and land-use changes). Overall, the sink estimated by the CBM from FM, AR and D during 2000–2012 corresponds to about 7 % of total GHGs at the EU level for the same period. The sink from the HWP pool contributes an additional 1 %. The CBM results for FM, AR and D are very close to the reported values from the EU GHGI, i.e. 8 % lower for FM (in the CBM vs. GHGI) and almost identical for AR and D. In absolute terms, the impact of harvest is much greater than natural disturbances but, because of salvage logging, the impact of natural disturbances is often not easily distinguishable from the impact of harvest. Over the period analyzed in our study, the sum of these two drivers often caused year-by-year variations of about 10–15 % of the FM sink at the EU level. Between 2000 and 2012, land-use changes are also important: the annual sink from AR reached 15 % of the FM sink, while annual emissions from D were on average equal to nearly 10 % of the FM sink. Our analysis on natural disturbances showed that at the EU level the impact of storms on the C stock balance is quantitatively far more important than the impact of fires, i.e. 5–10 times greater when both direct and indirect emissions are considered. While other studies quantified the impact of these events on the European forest carbon balance (e.g. [23, 57]), the detailed analysis of the C dynamics of natural disturbances presented in this study is useful both to identify climate change mitigation options and to allow better process understanding for the purpose of accounting emissions and removals under the KP and the post-2020 EU regulations. In particular, while storms cause a transfer of C from biomass to DOM pools (from where it will decompose unless salvage logged), fires cause an approximately equal amount of direct emission (of CO2 to the atmosphere) and indirect emission (i.e. C moved from living biomass to DOM from where it will decompose). A consistent fraction of the total amount of harvest may be provided by salvage logging after disturbance events, and in several cases the impact of natural disturbances is not visible in the GHGIs either because the dynamics described above are not taken into account (e.g. Italy, [44]) or because the methods used to estimate GHGs do not detect interannual variations (e.g. Germany, [26]). In the light of the rules under the KP [11], our results suggest that: (i) different management strategies (e.g., salvage logging) applied after natural disturbances may significantly affect the total C balance, and (ii) modelling the detailed C dynamics associated to natural disturbances is an essential prerequisite to apply the provision of natural disturbances under the KP (i.e. excluding the corresponding emissions and subsequent removals from the accounting under certain circumstances). In conclusion, this study presents the application of a consistent methodological approach, based on an inventory-based model, adapted to the forest management conditions of 26 different EU countries. The approach captures, with satisfactory detail, the C sink reported in the EU GHGI and the country-specific variability due to harvest, natural disturbances and land-use changes. These results provide the basis for possible policy-relevant future applications of the CBM, e.g., as a tool to support GHGIs (e.g. on accounting for natural disturbances) and to verify GHGIs, and for the simulation of specific scenarios at EU level. Applying the same model from a regional [58], to country [26, 44] and at the EU level (this study), may help a consistent assessment of different forest sector mitigation strategies appropriate to the specific regional circumstances, and in evaluating the overall contribution of the forest sector towards EU emissions reduction targets. To our knowledge, this is the most comprehensive study of its kind at EU level. Indeed, even if the EU forest C sink was previously analyzed by many other studies, none of them provided an overall assessment disaggregated between FM, AR and D, including all the forest pools, HWP and natural disturbances, and a comparison with the EU GHGI. Such comprehensive assessments and comparisons are increasingly needed to help improving the quality of GHGIs and ultimately increasing the credibility of the forest sink as a potential mitigation option within the EU [27] and the international frameworks [10]. As part of an integrated modeling framework, further possible developments include linking the CBM to models predicting land-use changes, the impact of climate change on primary productivity, the harvest demand and the material substitution effects of the industrial roundwood products. Methods This study used the Carbon Budget Model (CBM, [25]) to estimate forest C stock and the net CO2 emissions in 26 EU countries, covering a total forest area of about 146 Mha, disaggregated in 189 forest types (FTs) distributed over 178 administrative regions and 35 climatic units. The area included about 138 Mha of FM (at time step zero of the model runs) and, in 2012, 8 Mha of AR, and 2.8 Mha of D, since 1990. Due to the D area, the area of FM slightly decreased during the model runs, but this area decrease was compensated by an increase in AR area. Unproductive forests (according to countries’ GHGIs) and overseas territories were not included in this study [26]. The spatial framework applied by the CBM conceptually follows IPCC Reporting Method 1 [11] in which the spatial units are defined by their geographic boundaries and all forest stands are geographically referenced to a spatial unit (SPU). For FM, the Carbon Budget Model and the general methodological assumptions applied to each country were described previously [26]. Further details can be found elsewhere for the model itself (e.g. [25]), for its applications to European countries [26, 44] and for the representation of natural disturbances [26]. A summary of the main NFI input data used for each country is reported in Table 2. We considered 26 administrative units (i.e., European countries, as reported by Table 2) and 35 climatic units (CLUs, as defined by [59]), with mean annual temperatures (mainly affecting the DOM turnover rate), ranging from −7.5 to +17.5 °C.Table 2 Summary of the main parameters applied by CBM model for each country Country Original NFI year Time step 0 (years) CBM FM area (Mha)b Harvest rate (av. 2000–2012, Mm3) County specific biomass equations Austria 2008 1998 3.2 22.9 X Belgium 1999 1999 0.7 4.3 Bulgaria 2000 2000 3.2 5.3 Croatia 2006a 1996 2.0 4.6 Czech Republic 2000 2000 2.6 17.0 X Denmark 2004 1994 0.5 2.3 Estonia 2000 2000 2.1 7.9 Finland 1999 1999 21.7 55.0 France 2008 1998 14.6 54.9 Germany 2002 1992 10.6 74.7 X Greece 1992a 1992 1.2 1.6 Hungary 2008 1998 1.6 6.2 X Ireland 2005 1995 0.5 2.8 Italy 2005 1995 7.4 10.2 X Latvia 2009 1999 3.2 15.8 X Lithuania 2006 1996 2.0 7.7 Luxembourg 1999 1999 0.1 0.3 Netherlands 1997 1997 0.3 1.2 Poland 1993 1993 8.9 37.8 Portugal 2005 1995 3.6 12.2 X Romania 1985 1985 6.6 17.2 X Slovakia 2000 2000 1.9 9.0 Slovenia 2000 2000 1.1 3.3 Spain 2002 1992 12.6 16.8 Sweden 2006 1996 22.6 79.5 United Kingdom 1997 1997 2.5 9.8 EU 137.9 480.7 8 countries The table reports the NFI original reference year; the year since the model was applied; the FM area used by CBM at time step 0; the average harvest rate used; the countries where specific equations to convert the merchantable volume into aboveground biomass were selected. Two countries were not modeled: Cyprus (no NFI data available) and Malta (very small forest area, mainly covered by shrub lands) aAnalysis based on data from forest management plans bFM area used by CBM at time step 0. According to KP rules, FM is the area of forest in 1990, decreased with any subsequent deforestation. The FM area is taken from the official submissions made by countries to UNFCCC/Kyoto Protocol, giving priority to data from KP-CRF tables when available (i.e., if FM had been elected during the first KP commitment period), or alternatively taking data from the convention CRF tables (using ‘forest land remaining forest land’ in 1990 as a proxy for FM). To obtain FM area at time step 0, the D area reported by all countries under the Kyoto Protocol was used. Please note that CBM runs did not include forests reported as “not productive” (e.g., 0.4 Mha in Austria, 0.02 Mha in Bulgaria, 5 Mha in Sweden) and overseas territories (8.2 Mha in France) Within a SPU, each forest stand is characterized by age, area and seven classifiers that provide administrative and ecological information, the link to the appropriate yield curves, and parameters defining the silvicultural system such as the forest composition (defined according to different FTs), the management type (MT), and the main use of the harvest provided by each SPU, as fuelwood or industrial roundwood. For each country, these parameters were mainly derived by national forest inventories. According to country-specific information, MTs may include even-aged high forests, uneven-aged high forests, coppices and specific sylvicutural systems such as clear-cuts (with different rotation lengths for each FT), thinnings, shelterwood systems, partial cuttings, etc. (detailed information on the main model’s assumptions for five representative countries are reported in the Supplementary Information of [26]). To assess the FM area, data from KP reporting were used when available (for 18 countries in the period 2008–2012); alternatively “forest land remaining forest land” data (from UNFCCC reporting) were used, i.e. for all countries before 2008 and for those counties that did not elect FM during 2008–2012. Countries’ data for AR and D for 2008–2012 always came from KP reporting. The following carbon pools were considered: living biomass (aboveground and belowground), dead organic matter (DOM), mineral soil and harvested wood products (HWP). Even if the CBM estimates CH4 and N2O emissions, this study includes only CO2 and excludes organic soils. In the CBM, species-specific, stand-level equations [60] convert merchantable volume production into aboveground biomass, partitioned into merchantable stemwood, other (tops, branches, sub-merchantable size trees) and foliage components. Where additional information provided by NFIs or by literature was available (see last column in Table 2), country-specific equations were selected to convert the merchantable volume into aboveground biomass. The CBM starts the initialization process with all DOM pools containing zero C stocks and then simulates multiple iterations of growth and stand-replacing disturbances, gradually increasing the size of the DOM pools [25]. The rotations continue until the slowly-decaying C pools at the end of two successive rotations meet a difference tolerance of 0.1 %. Once this criterion has been met, the CBM applies a user-selected last disturbance event which affects the amount of C in the DOM pools, and then links DOM dynamics to biomass dynamics. Inputs from biomass to DOM pools, during the model run, result from biomass litterfall and turnover as well as natural and human-caused disturbances. The DOM parameters were first calibrated and validated on some specific study at country and regional level [44, 58] and, if necessary, further modified for specific countries [26]. We use two sets of yield tables (YT) in these analyses [26, 44]. Historical YTs derived from the standing volumes per age class reported by the NFI represent the impacts of growth and partial disturbances during stand development. Current YTs derived from the current annual increment reported in country NFIs represent the stand-level volume accumulation in the absence of natural disturbances and management practices. For 21 countries, we also evaluated the impact of natural disturbance events (a summary at EU level is reported in Fig. 8, below), including storms and ice (15 countries), fires (11 countries) and insect attacks (i.e., bark beetles attacks, for 2 countries). Specific information on the assumptions on natural disturbances are reported by [26], The following sections provide specific information on the application of CBM to areas affected by land-use changes (AR, D lands) and a summary, at the EU level, of the main assumptions on harvest applied by our study. AR and D assumptions We used for the analyses of AR and D areas the same 26 administrative units (i.e., European countries) and 35 CLUs applied for FM. For both AR and D, we used the area reported up to 2012 by each country in the GHGIs submitted the KP (Fig. 7). Before 2008, only the cumulative values since 1990 are available [13] and therefore we used an average annual rate of AR and D for the period 1990–2007.Fig. 7 Total annual rate (kha year−1) of afforestation (AR) and deforestation (D), used by CBM at the EU level, based on countries’ GHGIs under the Kyoto Protocol. Data from 1990 to 2007 are the cumulative area of AR and D in 2008 converted to annual average rates of land-use change AR was modelled through country-specific model runs, always starting in 1990. The total amount of AR per year was distributed between different FTs using the same proportions of FTs observed in the FM area. Based on a preliminary model assessment, we generally used the current YT library for AR [61]. This library was derived from the increment data reported by each country. These values represent the gross volume yield of each stand (while the YTs derived by the standing volume include the impact of past silvicultural treatment) and therefore are more suitable for young stands, generally younger than 20 years (i.e. for AR), where in general no silvicultural treatments are applied. We assumed that the harvest rate was entirely satisfied by the FM area (with Portugal the only exception, see [26]), apart from a modest amount of harvest provided by D. The possible amount of harvest provided by AR (i.e. post-1990 forest) was generally very small [61]. For AR we estimated the maximum potential (and theoretical) amount of harvest provided assuming a common set of silvicultural practices for all the countries, with a 15 % commercial thinning applied to broadleaved forests 15-years or older and a 20 % commercial thinning applied to coniferous forests 20-years or older [61]. The only case where the harvest from AR was relevant was Portugal. EU summary of the main input data A summary, at the EU level, of the main input data applied by our study is reported in Fig. 8, including: the total harvest rate applied by CBM at the EU level, further distinguished between industrial roundwood (IRW) and fuelwood (FW) and between coniferous and broadleaved species (panel a); total area affected by storms and ice (panel b); total area affected by fire (panel c). The harvest rate was mainly derived by a specific study on the HWP at EU level [30] where, for each country, FAOSTAT data (further distinguished between IRW and FW) were collected, compared with other data sources (i.e., Forest Resource Assessment, National Inventory Report, NFIs, ecc.) and eventually corrected in order to account for possible inconsistencies (i.e., due to the bark fraction or different methodological approaches).Fig. 8 a Total harvest demand (in millions m3) applied by CBM at the EU level, further distinguished between industrial roundwood (IRW) and fuelwood (FW) and between coniferous and broadleaved species; b total area affected by storms and ice (in kha); c total area affected by fire (in kha) Since the amount of IRW used in the present study is consistent with [30], we derived the IRW C sink by this last study, based on the IPCC production approach [11]. In this method, estimates of net emissions are derived from a stock change calculations applied to products derived from domestic harvest, i.e., imported HWP are excluded in the national estimates. Abbreviations ARafforestation and reforestation Ccarbon CBMCarbon Budget Model CP1First Commitment Period CP2Second Commitment Period Ddeforestation Dbhdiameter at breast height DOMdead organic matter EUEuropean Union FMforest management FTforest type FWfuelwood GHGgreenhouse gas GHGIgreenhouse gas inventory HWPharvested wood product IPCCIntergovernmental Panel on Climate Change IRWindustrial Roundwood KPKyoto Protocol LULUCFland Use, Land-Use Change and Forestry MTmanagement type NFINational Forest Inventory NIRNational Inventory Report UNFCCCUnited Nations Framework Convention on Climate Change YTyield table 1 Since this paper reports the results of a long-term study, based on data available until 2014, further comparisons with the data reported by the 2015 GHGI is not appropriate. 2 In the area-based comparison, the FM area of the EU GHGI was decreased by the same amount decreased for CBM runs, i.e. excluding unproductive forests and overseas territories, see “Methods” for further details. 3 The uncertainty raises to 25–50 % when analyzed at country level. 4 According to FAO, the net annual increment is defined as the average periodic increment of volume of all trees measured at a certain minimum Dbh, including the increment of trees which have been felled and excluding the increment of trees which have died during the reference period [62]. Authors’ contributions RP carried out the data analysis, in collaboration with JWM. GG and WAK helped in the design of the study and the interpretation of results and together with RP wrote the manuscript, in collaboration with RAV. All authors read and approved the final manuscript. Acknowledgements This paper was prepared in the context of the Contract n. 31502, Administrative Arrangement 070307/2009/539525/AA/C5 between JRC and DG CLIMA. Further information was collected in the context of the AA 071201/2011/611111/CLIMA.A2. The analysis performed for each country was in general based on data publicly available and on additional information collected at country level, in collaboration with many colleagues and experts for each country. We especially thank our colleagues, Giulia Fiorese, Viorel Blujdea and Tibor Priwitzer, who provided useful comments and suggestions. We also thank four anonymous reviewers, who provided useful comments and suggestions to improve the manuscript. The views expressed are purely those of the authors and may not in any circumstances be regarded as stating an official position of the European Commission or Natural Resources Canada. Competing interests The authors declare that they have no competing interests. ==== Refs References 1. Alkama R Cescatti A Biophysical climate impacts of recent changes in global forest cover Science 2016 351 600 604 10.1126/science.aac8083 26912702 2. Naudts K Chen Y McGrath M Ryder J Valade A Otto J Lussayert S Europe’s forest management did not mitigate climate warming Science 2016 351 597 600 10.1126/science.aad7270 26912701 3. Lundmark T Bergh J Hofer P Lundström A Nordin A Poudel BC Sathre R Taverna R Werner F Potential roles of Swedish forestry in the context of climate change mitigation Forests 2014 5 557 578 10.3390/f5040557 4. URL (last access April 2016). http://science.sciencemag.org/content/351/6273/597.e-letters. 5. Knauf M Köhl M Mues V Olschofsky K Frühwald A Modeling the CO2-effects of forest management and wood usage on a regional basis Carbon Balance Manag 2015 10 13 10.1186/s13021-015-0024-7 26097501 6. Nabuurs G-J, Delacorte P, Ellison D, Hanewinkel M, Lindner M, Nesbit M, Ollikainen M, Savaresi A. A new role for forests and the forest sector in the EU post-2020 climate targets. From Science to Policy 2. European Forest Institute; 2015. 7. Mubareka S, Jonsson R, Rinaldi F, Fiorese G, San Miguel J, Sallnas O, Baranzelli C, Pilli R, Lavalle C, Kitous A. An integrated modelling framework for the forest-Based bioeconomy. Earthzine, 2014. URL (last access March 2015). http://earthzine.org/2015/01/19/an-integrated-modelling-framework-for-the-forest-based-bioeconomy/. 8. UNFCCC. Decision 2/CMP.7 Land use, land-use change and forestry, 2012. URL (last access March 2016). http://unfccc.int/resource/docs/2011/cmp7/eng/10a01. 9. Grassi G den Elzen MGJ Hof AF Pilli R Federici S The role of the land use, land use change and forestry sector in achieving annex I reduction pledges Clim Change 2012 115 3–4 873 881 10.1007/s10584-012-0584-4 10. UNFCCC. Adoption of the Paris Agreement. Report No. FCCC/CP/2015/L.9/Rev.1, 20155. URL (last access March 2016): http://unfccc.int/resource/docs/2015/cop21/eng/l09r01.pdf. (UNFCCC, 2015). 11. IPCC. In: Hiraishi T, Krug T, Tanabe K, Srivastava N, Baasansuren J, Fukuda M, Troxler TG, editors. Intergovernmental panel on climate change. Revised supplementary methods and good practice guidance arising from the Kyoto Protocol, 2013. Switzerland; 2014. 12. Proposal for a Regulation of the European Parliament and the Council on the inclusion of greenhouse gas emissions and removals from land use, land use change and forestry into the 2030 climate and energy framework. URL (last access July 2016). http://ec.europa.eu/transparency/regdoc/rep/1/2016/EN/1-2016-479-EN-F1-1.PDF. 13. EU NIR. Annual European Community greenhouse gas inventory 1990–2012 and inventory report 2014. Submission to the UNFCCC Secretariat. European Environment Agency, Technical report No 09/2014. URL (last access February 2016). https://www.google.it/search?q=Annual+European+Community+greenhouse+gas+inventory+1990%E2%80%932012+and+inventory+report+2014&ie=utf-8&oe=utf-8&gws_rd=cr&ei=fwi7VoSoJIXAOoCTroAF. 14. Neumann M Moreno A Mues V Härkönen S Mura M Bouriaud O Lang M Achten MJ Thivolle-Cazat A Broisz K Merganič J Decuyper M Alberdi I Astrup R Mohren F Hasenauer H Comparison of carbon estimation methods for European forests For Ecol Man 2016 361 397 420 10.1016/j.foreco.2015.11.016 15. Blujdea VNB Abadviñas R Federici S Grassi G The EU greenhouse gas inventory for the LULUCF sector: I. Overview and comparative analysis of methods used by EU member states Carbon Manag 2016 6 5 6 16. Ciais P Schelhaas MJ Zaehle S Piao SL Cescatti A Liski J Luyssaert S Le-Maire G Schulze E-D Bouriaud O Freibauer A Valentini R Nabuurs GJ Carbon accumulation in European forests Nat Geosci 2008 1 425 492 10.1038/ngeo233 17. Karjalainen T Pussinen A Liski J Nabuurs G-N Eggers T Lapveteläinen T Kaipainen T Scenario analysis of the impacts of forest management and climate change on the European forest sector carbon budget For Policy Econ 2003 5 141 155 10.1016/S1389-9341(03)00021-2 18. Pretzsch H Grote R Reineking B Rötzer TH Seifert ST Models for forest ecosystem management: a European perspective Ann Bot-London 2008 101 1065 1087 10.1093/aob/mcm246 19. Nabuurs GJ Schelhaas MJ Pussinen A Validation of the European forest information scenario model (EFISCEN) and a projection of Finnish forests Silva Fenn 2000 34 2 167 179 10.14214/sf.638 20. Böttcher H Kurz WA Freibauer A Accounting of forest carbon sink and sources under a future climate protocol-factoring out past disturbance and management effects on age-class structure For Ecol Man 2008 11 669 686 21. Ťupek B Zanchi G Verkerk G Churkina G Viovy N Hughes J Lindner M A comparison of alternative modelling approaches to evaluate the European forest carbon fluxes For Ecol Man 2010 260 241 251 10.1016/j.foreco.2010.01.045 22. Groen T Verkerk PJ Böttcher H Grassi G Cienciala E Black K Fortin M Köthke M Lehtonen A Nabuurs G-J Petrova L Blujdea V What causes differences between national estimates of forest management carbon emissions and removals compared to estimates of large-scale models? Environ Sci Policy 2013 33 222 232 10.1016/j.envsci.2013.06.005 23. Seidl R Schelhaas MJ Rammer W Verkerk PJ Increasing forest disturbances in Europe and their impact on carbon storage Nat Climate Change 2014 4 806 810 10.1038/nclimate2318 24. Verkerk PJ Anttila P Eggers J Lindner J Asikainen A The realizable potential supply of woody biomass from forests in the European Union For Ecol Man 2011 261 2007 2015 10.1016/j.foreco.2011.02.027 25. Kurz WA Dymond CC White TM Stinson G Shaw CH Rampley G Smyth C Simpson BN Neilson E Trofymow JA Metsaranta J Apps MJ CBM-CFS3: a model of carbon-dynamics in forestry and land-use change implementing IPCC standards Ecol Model 2009 220 480 504 10.1016/j.ecolmodel.2008.10.018 26. Pilli R Grassi G Kurz WA Viñas RA Guerrero N Modelling forest carbon stock changes as affected by harvest and natural disturbances. I. Comparison of model results for forest management with EU countries’ estimates Carbon Balance Manag 2016 11 5 10.1186/s13021-016-0047-8 27340427 27. Climate action, consultation on addressing greenhouse gas emissions from agriculture and LULUCF in the context of the 2030 EU climate and energy framework, 2016. URL (last access July 2016). http://ec.europa.eu/clima/consultations/articles/0026_en.htm. 28. Gardi C Jeffery S Saltelli A An Estimate of potential threats levels to soil biodiversity in EU Glob Change Biol 2013 19 1538 1548 10.1111/gcb.12159 29. Panagos P Hiederer R Van Liedekerke M Bampa F Estimating soil organic carbon in Europe based on data collected through an European network Ecol Ind 2013 24 439 450 10.1016/j.ecolind.2012.07.020 30. Pilli R Fiorese G Grassi G EU mitigation potential of harvested wood products Carbon Balance Manag 2015 10 6 10.1186/s13021-015-0016-7 26457115 31. FAO. Global forest resources assessment 2010. Main Report. FAO Forestry Paper, 2010, 163. URL (last access February 2016). http://www.fao.org/docrep/013/i1757e/i1757e.pdf. 32. Forest Europe, UNECE and FAO. State of Europe’s Forests 2015. Status and Trends in Sustainable Forest Management in Europe, 2015. URL (last access March 2016). http://www.foresteurope.org/docs/fullsoef2015.pdf. 33. Didion M Blujdea V Grassi G Hernández L Jandl R Kriiska K Lehtonen A Saint-André L Models for reporting forest litter and soil C pools in national greenhouse gas inventories: methodological considerations and requirements Carbon Manag 2016 34. Verkerk PJ Lindner M Zanchi G Zudin S Assessing impacts of intensified biomass removal on deadwood in European forests Ecol Indic 2009 11 27 35 10.1016/j.ecolind.2009.04.004 35. Baritz R Seufert G Montanarella L Van Ranst E Carbon concentrations and stocks in forest soils of Europe For Ecol Man 2010 260 262 277 10.1016/j.foreco.2010.03.025 36. Liski J Perruchoud D Karjalainen T Increasing carbon stocks in the forest soils of Western Europe For Ecol Man 2002 169 163 179 10.1016/S0378-1127(02)00306-7 37. Goodale CL Apps MJ Birdsey RA Field CB Heath LS Houghton RA Jenkins JC Kohlmaier GH Kurz WA Liu SR Nabuurs GJ Nilsson S Shvidenko AZ Forest carbon sinks in the Northern Hemisphere Ecol Appl 2002 12 891 899 10.1890/1051-0761(2002)012[0891:FCSITN]2.0.CO;2 38. Lal R Forest soils and carbon sequestration For Ecol Man 2005 2020 1–3 242 258 10.1016/j.foreco.2005.08.015 39. Achat DL Fortin M Landmann G Ringeval B Augusto L Forest soil carbon is threatened by intensive biomass harvesting Sci Rep 2015 5 15991 10.1038/srep15991 26530409 40. Schulp C Nabuurs G-J Verburg PH de Waal R Effect of tree species on carbon stocks in forest floor and mineral soil and implications for soil carbon inventories For Ecol Man 2008 256 3 482 490 10.1016/j.foreco.2008.05.007 41. Gasparini P Di Cosmo L Forest carbon in Italian forests: stocks, inherent variability and predictability using NFI data For Ecol Man 2015 337 186 195 10.1016/j.foreco.2014.11.012 42. Ortiz C Liski J Gärdenäs A Lehtonen A Lundblad M Stendahl J Ågren G Karltun E . Soil organic carbon stock changes in Swedish forest soils—a comparison of uncertainties and their sources through a national inventory and two simulation models Ecol Model 2013 251 221 231 10.1016/j.ecolmodel.2012.12.017 43. UNECE/FAO. The European forest sector outlook study II. 2010–2030. United Nations, Geneva, 2011. URL (last access February 2016). https://www.unece.org/fileadmin/DAM/timber/publications/sp-28.pdf. 44. Pilli R Grassi G Kurz WA Smyth CE Bluydea V Application of the CBM-CFS3 model to estimate Italy’s forest carbon budget, 1995–2020 Ecol Model 2013 266 144 171 10.1016/j.ecolmodel.2013.07.007 45. Böttcher H Verkerk PJ Gusti M Havlík P Grassi G Projection of the future EU forest CO2 sink as affected by recent bioenergy policies using two advanced forest management models GCB Bioenergy 2012 4 6 773 783 10.1111/j.1757-1707.2011.01152.x 46. European Union. EU Reference Scenario 2016. Energy, transport and GHG emissions. Trends to 2050. European Union; 2016. 47. Paul KI Polglase PJ Richard GP Predicted change in soil carbon following afforestation or reforestation, and analysis of controlling factors by linking a C accounting model (CAMFor) to models of forest growth (3PG), litter decomposition (GENDEC) and soil C turnover (RothC) For Ecol Manag 2003 177 485 501 10.1016/S0378-1127(02)00454-1 48. KP CRF Tables, Submission 2014. URL (last access June 2014). http://unfccc.int/national_reports/annex_i_ghg_inventories/national_inventories_submissions/items/7383.php. 49. Nabuurs G-J Lindner M Verkerk PJ Gunia K Deda P Michalak R Grassi G First signs of carbon sink saturation in European forest biomass Nat Climate Change 2013 3 792 796 10.1038/nclimate1853 50. UNFCCC, decision 2/CMP.7 on land use land use change and forestry, 2011. URL (last access February 2016). http://unfccc.int/resource/docs/2011/cmp7/eng/10a01.pdf. 51. Rüter S. Projections of net‐emissions from harvested wood products in European Countries. Work report of the institute of wood technology and wood biology, Johann Heinrich von Thünen‐Institute (vTI), Hamburg; 2011. Report No: 2011/1. 52. AWG-KP, Ad Hoc working group on further commitments for Annex I Countries under Kyoto Protocol. Submissions for Forest Management Reference Level, 2011. URL (last access March 2015). http://unfccc.int/bodies/awg-kp/items/5896.php. 53. Pan Y Birdsey RA Fang J Houghton R Kauppi P Kurz WA Phillips OL Shvidenko A Lewis SL Canadell JG Ciais P Jackson RB Pacala SW McGuire AD Piao S Rautiainen A Sitch S Hayes D a large and persistent carbon sink in the world’s forests Science 2011 333 988 993 10.1126/science.1201609 21764754 54. Smyth CE Stinson G Neilson E Lemprière TC Hafer M Rampley GJ Kurz WA Quantifying the biophysical climate change mitigation potential of Canada’s forest sector Biogeosciences 2014 11 3515 3529 10.5194/bg-11-3515-2014 55. Gardiner B, Blennow K, Carnus JM, Fleischer P, Ingemarson F, Landmann G, Lindner M, Marzano M, Nicoll B, Orazio C, Peyron JL, Reviron MP, Schelhaas MJ, Schuck A, Spielmann M, Usbeck T. Destructive storms in European forests: past and forthcoming impacts. Final report to European commission—DG Environment, 2010. URL (last access March 2015). http://www.efiatlantic.efi.int/portal/databases/forestorms/. 56. Bottero A Garbarino M Long JN Motta R The interacting ecological effects of large-scale disturbances and salvage logging on montane spruce forest regenereation in the western European Alps For Ecol Man 2013 292 19 28 10.1016/j.foreco.2012.12.021 57. Lindner M Maroschek M Netherer S Kremer A Barbati A Garcia-Gonzalo J Seidl R Delzon S Corona P Kolström M Lexer MJ Marchetti M Climate change impacts, adaptive capacity, and vulnerability of European forest ecosystems For Ecol Man 2010 259 698 709 10.1016/j.foreco.2009.09.023 58. Pilli R Grassi G Cescatti A Historical analysis and modeling of the forest carbon dynamics using the carbon budget model: an example for the Trento Province (NE, Italy) Forest@ 2014 11 20 35 10.3832/efor1138-011 59. Pilli R Calibrating CORINE Land Cover 2000 on forest inventories and climatic data: an example for Italy Int J Appl Earth Obs 2012 9 59 71 10.1016/j.jag.2012.04.016 60. Boudewyn P, Song X, Magnussen S, Gillis MD. Model-based, Volume-to-Biomass Conversion for Forested and Vegetated Land in Canada. Canadian Forest Service, Victoria, Canada, 2007 (Inf. Rep. BC-X-411). URL (last access, March 2015). http://cfs.nrcan.gc.ca/publications/?id=27434. 61. Pilli R Grassi G Moris JV Kurz WA Assessing the carbon sink of afforestation with the carbon budget model at the country level: an example for Italy iFor Biogeosci For 2014 8 410 421 10.3832/ifor1257-007 62. FAO, FRA 2015—terms and definitions. Forest resources assessment working paper 180, 2015. URL (last access July 2016). http://www.fao.org/docrep/017/ap862e/ap862e00.pdf.

PMC005xxxxxx/PMC5002050.txt

==== Front Curr Cardiol RepCurr Cardiol RepCurrent Cardiology Reports1523-37821534-3170Springer US New York 2756632877110.1007/s11886-016-0771-4Psychological Aspects of Cardiovascular Diseases (A Steptoe, Section Editor)Psychosocial Factors in Diabetes and Cardiovascular Risk Hackett Ruth A. ruth.hackett.09@ucl.ac.uk Steptoe Andrew a.steptoe@ucl.ac.uk Department of Epidemiology and Public Health, University College London, 1-19 Torrington Place, London, WC1E 6BT UK 27 8 2016 27 8 2016 2016 18 10 95© The Author(s) 2016 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Type 2 diabetes is a chronic disease that is increasing in prevalence globally. Cardiovascular disease is a major cause of mortality and morbidity in diabetes, and lifestyle and clinical risk factors do not fully account for the link between the conditions. This article provides an overview of the evidence concerning the role of psychosocial stress factors in diabetes risk, as well as in cardiovascular complications in people with existing diabetes. Several types of psychosocial factors are discussed including depression, other types of emotional distress, exposure to stressful conditions, and personality traits. The potential behavioral and biological pathways linking psychosocial factors to diabetes are presented and implications for patient care are highlighted. Keywords Type 2 diabetesPsychosocial factorsCardiovascular diseaseDepressionStressUniversity College London (UCL)issue-copyright-statement© Springer Science+Business Media New York 2016 ==== Body Introduction Diabetes is a growing problem that poses a major public health challenge globally. According to reports by the International Diabetes Federation (IDF) and the World Health Organization (WHO), over 8 % of the world population (415–420 million people) currently have diabetes, with prevalence expected to rise to 10.4 % (642 million) by 2040 [1, 2]. In the USA, an estimated 9.3 % of the population (29.1 million people) have diabetes [3]. Worldwide, type 2 diabetes is the most common form of diabetes accounting for 90 % of cases [1, 4]. Throughout this review, the term “diabetes” will be used to refer to type 2 diabetes unless indicated otherwise. Diabetes is the fourth or fifth leading cause of mortality in most high-income countries and as such represents a significant burden to public health systems [5]. In 2015, health spending on diabetes represented 12 % (USD 673 billion) of global health expenditure and it may account for up to 20 % of national health-care budgets in some countries [1]. Additionally, the indirect costs of diabetes such as a reduced labor force and lowered economic productivity are considerable [6]. Cardiovascular Disease in People with Diabetes Cardiovascular disease (CVD) is a major cause of mortality and morbidity in individuals with diabetes. Results from a meta-analysis of 102 prospective studies indicate that people with diabetes have a 2-fold excess risk of CVD compared with controls. This association is independent of traditional CVD risk factors [7]. The risk associated with raised blood glucose is continuous, such that intermediate categories of glucose disturbance or “prediabetes” are associated with heightened CVD risk [7, 8]. Contemporary evidence from a longitudinal population study of 1.9 million people provides more evidence that there is a strong association between diabetes and incident CVD [9]; however, it is worth noting that the magnitude of the reported relationship varied between different CVD subtypes in this study. In keeping with wider population time trends, CVD rates have fallen over the past decades among people with diabetes [10]. However, the reduction has been less than in the rest of the population; thus, the heightened risk of CVD in people with diabetes persists [11•, 12]. In addition to heightened CVD risk, diabetes is associated with CVD mortality. According to a review of 97 prospective studies, individuals with diabetes die 6 years earlier on average than their counterparts without the condition, and approximately 58 % of this survival difference is attributable to excess vascular deaths [13]. Similar estimates of CVD-driven excess mortality are presented in the latest IDF and WHO reports [1, 2]. In an authoritative pooled analysis of 91 longitudinal cohort studies, diabetes, stroke, and MI were found to have equivalent associations with all-cause mortality [14]. Outcomes following a cardiovascular event are substantially worse in people with diabetes. In a trial of 13,608 ST-segment elevation MI patients, participants with diabetes had significantly higher rates of recurrent nonfatal MI and cardiovascular death than patients without diabetes [15]. Interestingly, diabetes is approximately a third more strongly related to fatal than nonfatal MI [13]. Similarly, in individuals with heart failure, diabetes is an independent predictor of repeat hospitalization and cardiac death [16]. Diabetes increases the risk of recurrent stroke [17] and attenuates both cognitive and functional recovery [18]. Myocardial revascularization procedures are also challenging in individuals with diabetes; compared to patients without the condition, people with diabetes have a substantially increased risk of mortality and adverse clinical outcomes following these procedures [19, 20]. CVD and Diabetes: A Link Beyond Traditional Risk Factors Lifestyle factors (smoking, poor diet, physical inactivity, excess alcohol consumption), clinical factors (obesity, hypertension, raised cholesterol), and psychosocial stress have been identified as modifiable risk factors for CVD development [5, 21]. Psychosocial stress factors can be divided into negative emotional disorders (e.g., depression and anxiety), personal traits (e.g., anger or hostility), and external stressors (exposure to stressful conditions). The evidence relating sociodemographic, behavioral, and clinical risk factors with CVD risk in people with diabetes was reviewed in this journal in 2015 [22]. This article did not cover the role of psychosocial factors in diabetes and their association with CVD risk. Furthermore, there is evidence that the link between CVD and diabetes is not fully accounted for by traditional risk factors. In meta-analyses showing greatly increased CVD morbidity and mortality in diabetes, the reported associations were robust to adjustment for clinical and behavioral risk factors [7, 13]. With regards to intervention and prevention, intensive programs targeting lifestyle factors such as diet, physical activity, and weight management have been shown to prevent diabetes onset in people with and without prediabetes [23–25]. The Diabetes Prevention Programs are recognized as some of the most effective lifestyle interventions for preventing chronic disease. Since these interventions modify CVD risk factors such high body mass index (BMI) and blood pressure [26], they in turn should have an impact on CVD outcomes. However, lifestyle interventions to reduce CVD in people with diabetes have been largely disappointing. These trial results were reviewed in 2015 in a joint report from the American Heart Association and American Diabetes Association [11•]. Evidence from prospective intervention studies such as the LookAHEAD trial [26] and the ACCORD trial [27, 28] suggests that the modification of behavioral risk factors such as weight loss, blood pressure, and cholesterol control does not significantly lower risk of adverse CVD outcomes in people with diabetes. Another survey of nearly 100 studies that looked at physical activity interventions and subsequent CVD in people with diabetes concluded that although effects have been seen in small studies, large randomized control trials (RCTs) have not found a protective effect [29]. Some positive findings regarding diet have been reported in the Spanish PREDIMED RCT. This study found that participants with diabetes who were randomized to a Mediterranean diet had a 30 % reduced risk of CVD over 4.8 years of follow-up [30]. However, these results must be interpreted with caution as a similar effect of dietary change on CVD outcomes were not reported in the LookAHEAD trial or the Diabetes Prevention Programs [23–25]. It may be that the Mediterranean diet is superior to the low-calorie diet used in other trials, but further studies are required to test this assertion. In light of this evidence, it appears that increased risk of CVD in people with diabetes is not fully explained by traditional risk factors. Therefore, the current review focuses on psychosocial factors in diabetes. Specifically, we evaluate the role of psychosocial factors in diabetes development, as well as the evidence for the involvement of psychosocial factors translating diabetes into increased CVD risk. Psychosocial Factors in Cardiovascular Disease There is accumulating evidence that psychosocial stress plays a role in the pathogenesis of CVD [31, 32, 33•]. Numerous systematic reviews and prospective analyses [34, 35] are in agreement that chronic stressors and psychosocial factors predict future coronary heart disease in initially healthy populations independently of standard risk factors. Considering the strong links between CVD and diabetes, it is not surprising that there has been interest in investigating whether psychosocial stress plays a role in diabetes and whether this is related to CVD risk in this population. The following sections review evidence from prospective studies that have examined the links between different psychosocial stress factors and diabetes. The conceptual framework in Fig. 1 indicates that psychosocial factors potentially impact on diabetes and cardiovascular risk in several ways: (1) influencing lifestyle risk factors for diabetes such as adiposity and physical activity, (2) affecting the development of diabetes directly through mechanisms such as glucose dysregulation and inflammation, and (3) shaping the processes through which diabetes stimulates cardiovascular complications. These distinctions are important both in etiological understanding and because of their implications for disease management.Fig. 1 Conceptual framework Psychosocial Factors and Diabetes Risk Emotional Distress Depression is the most commonly researched factor in studies of diabetes. Results from two meta-analyses of longitudinal studies indicate that depression is associated with 37–60 % increased risk of developing diabetes [36, 37]. Prospective evidence also suggests that elevated depressive symptoms, as well as clinical depression, are related to subsequent incidence of diabetes [38, 39]. The associations reported in these studies remained significant after controlling for diabetes risk factors such as BMI, family history of diabetes, smoking, physical activity, diet, and alcohol consumption. It should be emphasized that these studies do not prove causality, and alternative noncausal explanations are plausible [40•]. One possibility is that diabetes and depression share common etiological factors such as physical inactivity or inflammation that may not be completely eliminated by statistical adjustments. Alternatively, preclinical diabetes may increase the chances that an individual reports depression, resulting in a reverse causal process that is detailed later in this review. Fewer studies have investigated whether anxiety is associated with diabetes development. Engum [41] investigated the association in a Norwegian population cohort of 37,291 people. Over a 10-year follow-up, individuals who reported symptoms of anxiety at baseline had an increased risk of developing diabetes (odds ratio (OR) = 1.5; 95 % confidence interval (CI) 1.3–1.8) [41]. A limitation of this study was that anxiety and depression symptoms were not investigated separately; therefore, the effect of anxiety on diabetes independent of depression could not be assessed. A similar association between anxiety alone and incident diabetes (OR = 1.6; 95 % CI 1.2–2.1) was found in the Netherlands Study of Depression and Anxiety (NESDA) [42]. However, not all longitudinal research has found an association between anxiety and subsequent diabetes [43, 44], and there may be sex differences that are poorly understood [45, 46]. Psychological distress encompasses a range of comorbid psychological factors, such as depressive and anxiety symptoms, general stress as well as sleep disturbance. In a UK study of 9514 people, psychological distress at baseline was associated with incident diabetes over 18 years of follow-up adjusting for age, sex, education, and income. However, the relationship did not remain significant after additional adjustment for health-related factors [47]. In the UK Whitehall II cohort, psychological distress did not predict diabetes in the overall sample, but in a subsample of participants at high risk of diabetes (prediabetes at baseline and >40 on the Framingham diabetes risk score), distress was associated with a 40.9 % increased risk of diabetes independent of a range of covariates [48]. No sex interaction effects were found in the UK studies, but an earlier Swedish study of 5227 individuals who were normoglycemic at baseline reported an association between distress and subsequent diabetes in male but not in female participants [49]. The reasons for the mixed findings in the studies are unclear. It may be that initial health moderates the relationship and psychological distress accelerates progression to diabetes only in high-risk individuals. Another issue is that distress in of itself may be too broad a measure and that only particular aspects of it are related to future diabetes. Exposure to Life Stress Chronic exposure to external stressors has also been implicated in diabetes onset. To date, the majority of research has explored the relationship between work stress and incident diabetes [50, 51••]. Job strain, which is the combination of high job demands and low control at work, is a widely studied work stress construct [52]. A large meta-analysis investigating the link between job strain and diabetes development pooled results from 13 prospective European cohort studies. Over a mean follow-up of 10.3 years, job strain was associated with a 1.15-fold (95 % CI 1.06–1.25) increased risk of incident diabetes [51••]. This association was independent of a range of covariates and extends previous pooled cross-sectional associations [53]. A substantial number of studies of the relationship between long work hours and diabetes have also been carried out [50]. It appears that that working 55 h or more a week also increases the risk of developing diabetes but only in low socioeconomic status (SES) groups (RR 1.29, 95 % CI 1.06–1.57). Perceived stress is a broader conceptualization of psychosocial stress exposure that has been implicated in diabetes development. Novak et al. [54] investigated the relationship between perceived permanent stress (self-reported stress related to work or home life that was ongoing for a year or more) and incident diabetes in a sample of 7251 men. Over the 35-year follow-up period, men with permanent stress had a greatly increased risk of diabetes (HR 1.52; 95 % CI 1.26–1.82) compared with those reporting no or periodic stress. This relationship was not accounted for by conventional diabetes risk factors. Similarly, Japanese men with high levels of “stress in daily life” have been found to have a greater risk of incident diabetes over a 3-year follow-up [55]. However, findings from other prospective studies with both male and female participants have been equivocal. In the Copenhagen City Heart Study involving 7066 participants, men who reported daily emotional stress were two times more likely to develop diabetes than those with low levels of stress over a 10-year follow-up period (OR = 2.36; 95 % CI 1.22–4.59), but no associations were found for women [56]. Conversely, a study of 3759 Australian men and women found no relationship between perceived stress and the development of abnormal glucose tolerance in men over 5 years of follow-up, but detected an association in women [57]. It is unclear why some studies have found sex differences, but sample size might offer one possible explanation. The largest study to date investigated the relationship between perceived stress and subsequent diabetes development in 55,826 Japanese men and women over a 10-year follow-up period [58]. In this analysis, high levels of perceived stress were found to increase the risk of diabetes onset in both men and women, but effects were stronger among male (OR 1.39; 95 % CI 1.15–1.65) than female participants (OR 1.25; 95 % CI 1.01–1.56) after adjustment for known diabetes risk factors. Another study assessed the association between perceived stress and diabetes in 22,567 participants (71 % men) from a French workforce cohort. Over 5.3 years of follow-up, no association between perceived stress and future diabetes was observed in the full sample [59]. However, perceived stress was significantly associated with new onset diabetes in participants of low occupational status (OR 1.39; 95 % CI 1.02–1.90). The authors suggest that low occupational status might reflect a greater work stress. If this is the case, these results would map onto previous meta-analytic findings indicating that long working hours are only associated with incident diabetes in those of low SES [50]. Considering the current evidence as a whole, there appears to be an association between perceived stress and increased risk of diabetes in initially healthy populations. However, sex or SES (as measured by occupational status) may moderate this association. As in the case of depression, these longitudinal observational studies do not prove causality. Early Life Adversity Early life adversity has not been widely investigated as a risk factor for future diabetes onset, though it appears to be a significant issue for health-related processes such as telomere length and inflammation in adult life [60, 61]. A review published in 2015 summarized the existing evidence on diabetes [62•], analyzing 7 prospective and cross-sectional studies with data on 87,251 participants. People who reported an adverse childhood experience had a 32 % increased odds of diabetes. Looking at different types of childhood stress, the strongest association was found for neglect (OR 1.92), followed by sexual abuse (OR 1.39) and physical abuse (OR 1.30). A limitation of this meta-analysis is that some of the studies included involved retrospective accounts of childhood adversity which may not be accurate, whereas other studies were analyses of life course data. The studies varied in diabetes measurement, with both self-reported and objective measures being used. There are unanswered questions in this area such as whether there is a critical period in exposure to early life stress or whether there is a dose-response relationship between the frequency or duration of the stress and diabetes risk. Prenatal stress has also been suggested to play a role in the development of diabetes. In a large Danish cohort study, participants who were exposed to prenatal stress (n = 45,302 out of nearly 1.9 million) because their mothers experienced bereavement in their prenatal life were found to have an elevated risk of future diabetes (incidence rate ratio 1.31; 95 % CI 1.01–1.69). This association was independent of parental diabetes and other conventional diabetes risk factors [63]. Personality Traits Personality factors are not well researched in relation to diabetes. Hostility is a trait that is typically conceptualized as a negative cynical attitude toward others, with a propensity for anger or aggression [64]. This trait has been associated prospectively with raised fasting glucose [65] and cross-sectionally with insulin resistance [66, 67], glycated hemoglobin (HbA1c) [67, 68], and prevalent diabetes [69]. Additionally, angry temperament has been investigated in relation to diabetes development. In a cohort study of 11,615 individuals who were disease-free at baseline, individuals with an angry temperament had a 1.34 increased hazard of incident diabetes over the 6-year follow-up period [70]. This association has also been investigated in the MESA cohort [43] with a smaller sample size of 5598 participants but a longer follow-up of 11.4 years. Participants reporting high levels of trait anger at baseline had a 48 % greater risk of developing diabetes than those with low anger (HR 1.48; 95 % CI 1.04–2.12) independent of demographic factors, exercise, diet, alcohol use, and smoking. However, the association was attenuated following adjustment for waist circumference. So although there have not been many studies investigating anger and hostility, the current evidence suggests that these characteristics may also be associated with an increased risk of diabetes in later life. Potentially Protective Psychosocial Factors The vast majority of work has investigated relationships between negative psychosocial stress factors and future diabetes. To our knowledge, only two studies have investigated associations with potentially protective positive factors. Crump et al. studied the relationship between resilience to stress in adolescence and diabetes in later life [71••]. The study used an impressive sample of over 1.5 million Swedish military conscripts who were assessed for stress resilience during a semistructured interview at 18 years. The participants were followed up for an average of 25.7 years using national medical records. Participants with the lowest stress resilience had a 51 % increased hazard of future diabetes compared to those with high levels of resilience (HR 1.51; 95 % CI 1.46–1.57) independent of a range of diabetes risk factors. These findings suggest that personal resilience to stress might be an important factor in the development of diabetes. The other study looked at life satisfaction, emotional vitality, and optimism in 7800 individuals from the Whitehall II cohort [72]. Over 13 years of follow-up, these well-being factors were not associated with incident diabetes as assessed by self-reported and glucose tolerance testing. The authors performed subanalyses to assess whether associations varied by type of diabetes diagnosis. Individuals with high life satisfaction (OR = 0.85; 95 % CI = 0.76–0.95) and emotional vitality (OR = 0.86; 95 % CI = 0.77–0.97) were less likely to report doctor-diagnosed diabetes. It is uncertain why associations differed for doctor-diagnosed and screen-detected diabetes. Psychosocial Factors and Cardiovascular Complications in People with Diabetes The research detailed in previous sections had primarily focused on pathway 2 outlined in Fig. 1. Additionally, psychosocial factors may be relevant to the risk of cardiovascular complications in people with diabetes (pathway 3). Depression in Diabetes Depression is common in people with diabetes [73–76]. A meta-analysis of 10 studies with 51,331 individuals estimated the prevalence of depression to be almost doubled in people with diabetes compared to those without the condition (17.6 vs. 9.8 %; OR = 1.6; 95 % CI 1.2–2.0) [75]. The vast majority of the studies have been conducted in Western countries, but a study of 213,797 people in 47 countries from around the world has shown that people with diabetes had a 2-fold greater prevalence of depressive symptoms than those without diabetes (OR 2.36; 95 % CI 1.91–2.92) [73]. As well as increased prevalence in people with diabetes, longitudinal studies indicate that diabetes diagnosis is a risk factor for incident depression [37, 77, 78]. An analysis pooling data from 16 longitudinal studies involving 497,223 participants with an average follow-up of 5.8 years indicated that people with diabetes have a 25 % increased risk of developing depression compared with controls without diabetes [78]. More recent longitudinal studies have confirmed this relationship. For example, an analysis of the English Longitudinal Study of Aging (ELSA) cohort found that diabetes was associated with a doubling in the odds of depressive symptoms over 4 years of follow-up [42], while a larger study of 37,043 Swedish twins with diabetes showed 33 % increased risk of major depression [41]. However, in both of these studies, the association was only found in younger participants with diabetes (≤55 or 64) and not in older individuals. The reasons are not known. It might be that middle-aged people with diabetes perceive their condition as a disease of old age that should not have happened to them and become more despondent following diagnosis. Alternatively, it could be that younger patients have more competing priorities (e.g., employment, raising children, and financial commitments such as mortgages) and find diabetes care a greater burden to manage in comparison with older patients who are likely to have a grown-up family and be retired. There is some evidence from smaller cross-sectional studies that diabetes management might be more challenging for younger than older adults [79, 80] but stronger prospective evidence from larger samples is required to explore these possibilities further. It should be pointed out that the link between depression and diabetes is not unique to this condition. Similar associations have been observed for other common diseases such as osteoarthritis, coronary heart disease, chronic lung diseases, and stroke [81, 82]. Nevertheless, comorbid depression in diabetes is a considerable threat to quality of life in people with diabetes [83, 84] and is of clinical importance since it has been associated with suboptimal glycemic control [85, 86] and nonadherence to treatment regimens [87, 88], as well as with the microvascular [89, 90] and macrovascular complications of diabetes. This review focuses on the risk of macrovascular complications. Depression and Complications An increasing body of literature indicates that depression exacerbates the risk of macrovascular complications in people with diabetes. Prospective studies have found a relationship between comorbid depression and CVD in diabetes samples [91, 92]. For instance, in the Pathways Epidemiological Follow-up Study of over 4000 participants with diabetes, individuals with depression had a 24 % increased risk of adverse macrovascular complications including MI, stroke, cardiovascular procedures, and cardiac death over 5 years of follow-up [92]. This association was independent of prior complications, sociodemographic characteristics, health behaviors, and diabetes self-care variables. Another study followed a cohort of 345,949 individuals who were free of CVD at baseline over a 7-year period. Results showed that participants with diabetes alone and those with major depression alone had a 30 % increased risk of MI, whereas those with a double diagnosis (both diabetes and major depression) had an 82 % excess risk of subsequent MI compared to participants without either condition [93]. Another study published in 2016 examined the association between depressive symptoms or perceived stress at baseline and risk of CVD in 22,003 US adults [94]. Over almost 6 years of follow-up, people with diabetes who reported elevated depressive symptoms or perceived stress had a significantly increased incidence of stroke (HR 1.57; 95 % CI 1.05–2.33) and acute CVD (HR 1.57; 95 % CI 1.02–2.40). These associations were independent of demographics factors but were attenuated when controlling for lifestyle factors. The majority of studies have focused on white samples in Western countries. To address this issue, Ting et al. [95•] conducted a 7.4-year follow-up study in a sample of 7835 Chinese patients with diabetes who were free of CVD at baseline. In this sample, diagnosed depression predicted CVD [HR = 2.18] and the majority of the risk appeared to be driven by stroke [HR = 3.55]. The reported associations remained significant after adjustment for a wide range of conventional CVD risk factors. As might be expected from these findings, people with both depression and diabetes have greater CVD mortality as well as morbidity rates. An analysis of 16 prospective studies indicated that comorbid depression in diabetes is associated with a 39 % increased risk of cardiovascular death and a 46 % higher risk of all-cause mortality [97]. Anxiety in Diabetes In comparison with the research on comorbid depression, there is relatively little research on other psychosocial factors and their relationship to CVD risk in people with diabetes. However, there is some evidence that the prevalence of other psychosocial disorders is elevated in people with diabetes in comparison with the general population. Overviews of this research suggest that diabetes is associated with 20 % increased odds of having an anxiety disorder and 48 % increased odds of having elevated anxiety symptoms [96]. Comorbid anxiety is of clinical relevance to people with diabetes. An association between anxiety and glycemic control has also been reported in a number of studies [97], but to our knowledge, only one study to date has investigated the prospective relationship between anxiety and diabetes complications such as retinopathy, neuropathy, or CVD, and no associations were found [44]. However, the number of cases of diabetes and anxiety was limited in this study, so it may have been unpowered to detect such effects. Psychological Distress and Diabetes-Specific Distress Psychological distress has been linked with CVD morbidity and increased mortality rates in people with diabetes. In a study of 1533 individuals with diagnosed diabetes, distressed participants were found to have a 1.69 increased hazard of a CVD event and a 1.76-fold greater mortality rate compared to individuals without psychological distress over an average follow-up of 5.4 years [98]. This association was independent of CVD risk factors, and excluding participants who were receiving antidepressive treatment did not impact the results. It would be interesting in future studies to tease out what particular aspects of psychological distress are most strongly linked to CVD. As well as depression, anxiety, and general psychological distress, diabetes-related emotional distress, a stress condition specifically resulting from concerns and worries about diabetes and its management, is common in people with diabetes [99]. According to estimates from a study of 8596 adults with diabetes from 17 countries, 44.6 % of patients suffer from significant diabetes-related distress [100]. Diabetes-related distress is of clinical significance, since several longitudinal studies suggest that it is associated with poor glycemic control independently of depression [99, 101, 102]. No studies to date have investigated the association between diabetes-distress and macrovascular complications. Pathways Linking Psychosocial Factors and Diabetes The precise mechanisms linking psychosocial factors and diabetes remain to be elucidated. However, several potentially interrelated pathways that plausibly account for the link have been proposed. As noted in Fig. 1, one possibility is that the adverse relationship between psychosocial factors and diabetes may be mediated via behavioral pathways. Behavioral mechanisms include poor diet, physical inactivity, excess alcohol consumption, and smoking. Reduced adherence to self-care behaviors and cardioprotective medications such as blood pressure and lipid-lowering drugs could also play a role. Several lines of research support this mechanism. Comorbid depression in diabetes increases nonadherence to a range of behaviors including diet, medication usage, and exercise [87, 88]. For example, a study of 2759 individuals with diabetes found that participants with persistent or increasing depression symptoms had significantly poorer adherence to dietary and exercise regimens than their counterparts without depression over a 5-year follow-up period [88]. A cross-sectional study with a similar sample size of 2646 individuals with diabetes reported that physical inactivity doubled in the presence of depressive symptoms [103]. Considerably less research has investigated the behavioral pathways linking other psychosocial factors and diabetes. The Copenhagen City Heart Study of 7066 adults found that perceived stress was associated with physical inactivity and unsuccessful smoking cessation or alcohol reduction attempts over a 10-year follow-up as well as the development of overt diabetes among men [56]. Negative psychosocial factors that are common in diabetes may also decrease motivation for healthy lifestyle choices that in turn impact CVD risk. However, the association between diabetes and CVD is not fully explained by behavioral risk factors [7, 13], and results from RCTs suggest that the modification of behavioral risk factors does not significantly lower CVD outcomes in people with diabetes, despite the fact that lifestyle change has a marked effect on diabetes incidence [11•]. This offers the possibility that a direct biological pathway could link psychosocial factors with CVD risk in people with diabetes. Several different psychobiological mechanisms could link psychosocial factors with both CVD and diabetes. Depression, anxiety, and diabetes-related distress are associated with suboptimal glycemic control [85, 86, 97, 102, 104], and there is evidence that the risk of CVD increases in line with the degree of hyperglycemia [7, 105]. Long-term exposure to psychosocial stress factors can result in chronic allostatic load which involves dysregulation of multiple biological systems including neuroendocrine, cardiovascular, metabolic, and inflammatory pathways [106]. Cross-sectional evidence suggests that people with diabetes demonstrate characteristics of high allostatic load in response to a stressful experience [107•]. Evidence from the same sample suggests that hostility (a psychosocial stress factor) can further exaggerate the disturbances in stress-related processes in people with diabetes [108]. Many of the different pathways involved in stress-related allostatic load have been linked with both diabetes and CVD. Dysregulation of the stress hormone cortisol is associated with both acute and chronic stress factors [109], and prospective evidence has linked changes in daily cortisol secretion with CVD mortality [110], as well as new onset diabetes [111]. Another potential biological mechanism linking psychosocial factors to both CVD and diabetes is inflammation. Elevated inflammatory cytokine concentrations are observed in people reporting high levels of psychosocial stress [112]. Both diabetes [113] and CVD have been characterized as inflammatory conditions [114]. In initially healthy populations, heightened inflammatory cytokines levels have been associated with new onset CVD [115], and it is plausible that raised cytokine levels could increase the risk of CVD in diabetes populations, too. The complexity of the link between psychosocial factors and CVD risk in people with diabetes is reflected in the numerous pathways that may be involved. It is likely that the proposed behavioral and biological pathways do not act in isolation but rather that they are interrelated. For example, depression in diabetes might impact complications through poorer adherence to diet and exercise regimes. Future research in this area may benefit from investigating the synergies between behavioral and biological pathways in linking psychosocial factors to CVD risk in people with diabetes. Implications for Patient Care Lifestyle interventions to prevent diabetes have been successful [23–26]. However, interventions to modify behavioral risk factors to prevent CVD in people with diabetes have been largely disappointing. This poses the question whether modifying psychosocial stress, another risk factor for CVD, would have a therapeutic impact. To date, the majority of research in this area has investigated whether the treatment of depression in diabetes is beneficial. A Cochrane review in 2012 included 19 RCTs investigating both psychological and pharmacological interventions for depression in patients with diabetes [116]. Psychological intervention studies showed a beneficial effect on short-, medium-, and long-term depression severity and had a good impact on depression remission compared to usual care. However, the effect of psychological intervention on glycemic control was mixed and inconclusive. With regards to the pharmacological interventions, there was a moderate effect of antidepressant medication on short-term depression severity and depression remission, and interestingly, the pharmacological trials significantly improved glycemic control in the short term as well. But no study to date has assessed the relationship between depression treatment and glycemic control in the longer term. Taking the evidence together, it appears that depression treatment is moderately effective in diabetes, but only pharmacological trials have shown a consistent improvement in glycemic control. Mindfulness-based interventions for modifying psychosocial stress factors have also been tested in people with diabetes [117•]. They have been found in several studies to have psychological benefits, lowering depression, anxiety stress, and diabetes-distress symptoms in people with diabetes. However, the evidence for the effectiveness of these interventions on glycemic control is mixed. Out of the seven studies that assessed HbA1c as a marker of glycemic control, four interventions lowered HbA1c levels, but the three largest studies reported no change in HbA1c [117•]. Mindfulness-based intervention in diabetes is a new field, and much of the research is exploratory. It may be that the short follow-up periods of many studies were not sufficient to observe significant changes in HbA1c. There is limited evidence that the treatment of psychosocial factors can reduce the risk of adverse outcomes in people with diabetes [118]. The Prevention of Suicide in Primary Care Elderly Collaborative Trial (PROSPECT) RCT was used to investigate whether depression management would decrease mortality in diabetes [119]. Depressed people with diabetes who were assigned to the intervention group (an individualized case management approach) had significantly lower mortality rates than controls over a 5-year follow-up (HR 0.49; 95 % CI 0.24–0.98). However, this study has been criticized with regards to study design and analysis, with suggestions that the methods may not have been appropriate [120]. The impact of mindfulness interventions on CVD outcomes in people with diabetes has not yet been examined [117•]. In sum, there is little evidence as yet that the treatment of psychosocial factors in diabetes has a benefit on CVD outcomes. However, pharmacological interventions have been shown to improve glycemic outcomes in the short term and hyperglycemia is linearly associated with increased CVD risk [7, 105]. Additionally, both psychological and pharmacological treatments as well as mindfulness-based interventions appear to have beneficial effects on psychosocial factors in people with diabetes. Despite the limited effectiveness of these treatments on overt CVD outcomes, there have been calls that the psychological well-being in people with diabetes should be a priority for its own sake [121]. Since lifestyle interventions have been shown to be effective for preventing diabetes, perhaps targeting individuals before onset rather than after diagnosis might be the optimal strategy. Conclusion There is an emerging body of evidence that psychological stress factors play a role in the pathogenesis of diabetes. A variety of negative psychosocial factors have (for the most part) been shown to increase the risk of diabetes in initially healthy populations. There is less research on the involvement of psychosocial factors in CVD risk in people with existing diabetes. To date, most of the research in this area has been on depression, with evidence that a double diagnosis of diabetes and depression increases the risk of CVD in this population. Most of the studies are observational, so causal conclusions are difficult to draw. The mechanisms through which psychosocial stress factors increase diabetes risk and affect outcomes in people with existing diabetes are yet to be fully understood. It is likely that both behavioral and biological pathways are involved. Interventions have been shown to have a beneficial effect on psychosocial factors in people with diabetes, but the evidence for effects on glycemic control is mixed, and research on interventions impacting CVD outcomes in people with diabetes is lacking. Despite the limited evidence for effects on physiological outcomes, it can be argued that improving the psychological well-being of people with diabetes should be a priority in its own right. This article is part of the Topical Collection on Psychological Aspects of Cardiovascular Diseases Ruth A. Hackett and Andrew Steptoe are supported by the British Heart Foundation. Compliance with Ethical Standards Conflict of Interest Ruth A. Hackett and Andrew Steptoe declare that they have no conflicts of interest. Human and Animal Rights and Informed Consent This article is based on published papers and does not contain any examination with human or animal subjects performed by any of the authors. ==== Refs References

PMC005xxxxxx/PMC5002051.txt

==== Front Maxillofac Plast Reconstr SurgMaxillofac Plast Reconstr SurgMaxillofacial Plastic and Reconstructive Surgery2288-81012288-8586Springer Berlin Heidelberg Berlin/Heidelberg 276353917910.1186/s40902-016-0079-8ResearchS100 and p65 expression are increased in the masseter muscle after botulinum toxin-A injection Park Young-Wook 1http://orcid.org/0000-0001-5088-2732Kim Seong-Gon 82-33-640-2468kimsg@gwnu.ac.kr 1Jo You-Young 21 Department of Oral and Maxillofacial Surgery, College of Dentistry, Gangneung-Wonju National University, 7 Jukhyun-gil, Gangneung, 210-702 South Korea 2 Sericultural and Apicultural Materials Division, National Academy of Agricultural Science, Suwon, South Korea 26 8 2016 26 8 2016 12 2016 38 1 334 7 2016 22 8 2016 © The Author(s). 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Background The purpose of this study was to compare the expression levels of p65 and S100 in the rat masseter muscle after the injection of different concentrations of botulinum toxin-A (BTX-A). Methods We injected either 5 or 10 U of BTX-A into both masseter muscle of rats. As a control group, the same volume of saline was injected. After 14 days, the animals were sacrificed. Subsequently, a biopsy and immunohistochemical staining of the samples were performed using a p65 or S100 antibody. Results The cross-sectional area of each myofibril was significantly reduced by BTX-A injection (P < 0.001). The expression of p65 and S100 increased significantly with increasing concentrations of BTX-A (P < 0.001). Conclusions The injection of BTX-A into the masseter muscle induced muscle atrophy. Subsequently, p65 and S100 expression in myoblasts were increased for the protection of muscle cells. Keywords Botulinum toxin-Ap65S100ApoptosisMasseter musclehttp://dx.doi.org/10.13039/501100003627Rural Development AdministrationPJ01121404Kim Seong-Gon issue-copyright-statement© The Author(s) 2016 ==== Body Background Botulinum toxin is a neurotoxin produced by the anaerobic bacterium Clostridium botulinum [1, 2]. This toxin selectively hydrolyzes the receptor that is required for the binding of the synapse vesicle and the membrane on the pre-synapse of the neuromuscular junction; as a result, the toxin blocks the release of acetylcholine [3–6], causing muscle weakness, paralysis, and atrophy [7]. Botulinum toxin-A (BTX-A) is a type of botulinum toxin that was approved by the US Food and Drug Administration (FDA) in 1989. Since that time, BTX-A has been used for cosmetology, focal dystonia, facial spasm, hyperhidrosis, and the treatment of muscle hypertrophic disorder [8, 9]. In addition, in the oral and maxillofacial area, BTX-A is injected into masticatory muscles such as the masseter muscle or the temporal muscle for esthetic and therapeutic purposes [10–12]. Previous studies primarily investigated the effect of BTX-A on muscles and nerves [13]. However, muscle paralysis or weakness impacts on the bone, resulting in a reduction of the affected bone, even if BTX-A is injected into the muscle [14, 15]. Therefore, more studies are needed to explain the mechanism by which BTX-A affects the masseter muscle. BTX-A injection into the masseter muscle induces muscle atrophy [16]. Several marker proteins have the potential to increase their expression during the skeletal muscle atrophy process [17]. S100 is a well-known calcium-binding protein [18, 19]. The terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay has been used to assess apoptosis in many publications [20]. After burn injury, muscle atrophy is accompanied by increasing S100 levels and TUNEL-positive cells in the motor nerve [21]. The NF-kB pathway is important in cellular apoptosis. The expression of caspases and key components of the NF-kB pathway (p65) is increased during skeletal muscle atrophy [22]. Therefore, the expression levels of S100 and p65 in the masseter muscle may change after BTX-A injection. However, S100 expression levels have not been studied before in the context of BTX-A injection. The purpose of this study was to compare the expression levels of p65 and S100 in the rat masseter muscle after the injection of different concentrations of BTX-A. Methods Experimental animals and housing conditions Fifteen Sprague-Dawley rats (age 10–12 weeks, body weights 250–300 g) were purchased from Samtako (Osan, Korea). Each rat was individually housed and allowed an adaptation period for 10 days. The Institutional Animal Care and Use Committee of Gangneung-Wonju National University approved this experiment (GWNU 2015-25). Experimental design The animals were divided into three groups: the control group and two experimental groups. Five rats were used in each group. In one experimental group, 5 U of a BTX-A solution was injected into both sides of the masseter muscle. In the other experimental group, 10 U of a BTX-A solution was injected on both sides of the masseter muscle. The control group was injected with 0.1 cc of saline on both sides of the masseter muscle. Fourteen days after injection, all rats were sacrificed, and samples containing the masseter muscle and mandible were obtained for histological examination. After the harvested tissues were fixed in a formalin solution for 1 day, decalcification was performed with a 5 % nitric acid solution, and specimens containing the masseter muscle and mandible were generated. For following evaluation, the specimens were cut as horizontal plane. The height for sectioning ramus was determined as the central cut between the zygomatic arch plane and the lower mandibular border plane. The specimens were stained with hematoxylin and eosin (H&E) to evaluate the cortical bone thickness of the mandibular ramus. We photographed the histological view at 14 days post-surgery and measured the ramal cortical bone thickness of the rats using size measuring software (SigmaScan-Pro®; SPSS Science, Chicago, IL, USA). The cross-sectional area of myofibrils was also measured using the same software. Immunohistochemical staining was performed for p65 and S100. Both antibodies were murine monoclonal antibodies purchased from Santa Cruz Biotech (Santa Cruz, CA, USA). The primary antibody dilutions were as follows: S100, 1:30; and p65, 1:50. Subsequent procedures were in accord with a previous publication [23]. To quantify the intensity of the immunohistochemical reaction, the intensity of positive staining was evaluated in five random fields of the masseter muscle at a ×200 magnification using a computer-assisted image analysis program. The staining intensity in immunohistochemistry experiments was shown as the mean intensity value (0: no stain, 255: highest stain). Counterstaining procedure was omitted to ensure that the intensity value would be solely attributable to the positive immunohistochemical reaction. Statistical analysis All of the results were statistically analyzed using one-way analysis of variance (ANOVA), followed by a post hoc test (Bonferroni’s method). The significance level was set as P < 0.05. Results A horizontal cut of the mandibular ramus is shown in Fig. 1a–c for saline-treated, 5-U BTX-A-treated, and 10-U BTX-A-treated specimens, respectively. The average thicknesses of the cortical bone were measured as follows: the saline-treated control group had a thickness of 0.18 ± 0.05 mm, the 5-U BTX-A-treated group had a thickness of 0.15 ± 0.05 mm, and the 10-U BTX-A-treated group had a thickness of 0.10 ± 0.03 mm (Table 1). The cortical bone thickness of the mandibular ramus of rats was reduced in the BTX-A-treated groups in comparison to the saline-treated control group. However, the difference observed among the groups was not statistically significant (P = 0.055).Fig. 1 Cortical bone of the rat mandibular ramus. The arrows indicate the mandibular ramus. a Saline-treated group; b 5-U BTX-A-treated group; c 10-U BTX-A-treated group (H&E stain, original magnification ×20) Table 1 The cortical bone thickness and the cross-sectional areas of masseter muscle after BTX-A injection Group Cortical bone thickness Cross-sectional area Average (mm) P value Average (μm2) P value Saline-treated 0.18 ± 0.05 – 621.68 ± 87.74 – 5-U BTX-A-treated 0.15 ± 0.05 NS 495.42 ± 110.15 0.037 10-U BTX-A-treated 0.10 ± 0.03 NS 263.05 ± 115.31 <0.001 The P value was calculated by post hoc test and compared with the saline-treated group BTX-A botulinum toxin-A A cross section of the masseter muscle is shown in Fig. 2a–c for saline-treated, 5-U BTX-A-treated, and 10-U BTX-A-treated specimens, respectively. The average cross-sectional areas of myofibrils were 621.68 ± 87.74 μm2, 495.42 ± 110.15 μm2, and 263.05 ± 115.31 μm2 for the saline-treated control, the 5-U BTX-A-treated group, and the 10-U BTX-A-treated group, respectively (Table 1). The cross-sectional area of myofibrils was significantly reduced in the BTX-A-treated groups in comparison to the saline-treated control group (P < 0.001). The post hoc test revealed differences between the group treated with 10 U BTX-A and the other groups, with significantly lower values in the 10-U BTX-A group than in the saline-treated control and the 5-U BTX-A-treated group (P < 0.001 for both groups). When we compared the 5-U BTX-A-treated group to the saline-treated control, the observed difference was statistically significant (P = 0.037).Fig. 2 Cross section of the masseter muscle. a Saline-treated group; b 5-U BTX-A-treated group; c 10-U BTX-A-treated group (H&E stain, bar = 20 μm) The immunohistochemical findings for p65 and S100 are shown in Figs. 3 and 4. The immunohistochemical findings demonstrated that the expression of p65 and S100 were significantly higher in the 10-U BTX-A-treated groups than in the saline group (Figs. 3d and 4d; P < 0.001 and P = 0.002 for p65 and S100, respectively). The mean intensity values for p65 were 81.02 ± 1.53, 82.93 ± 2.11, and 106.46 ± 10.86 for the saline, 5-U BTX-A, and 10-U BTX-A treatments, respectively. The post hoc test revealed differences between the group treated with 10 U of BTX-A and the other groups, with significantly higher values in the 10-U BTX-A group than in the saline-treated control and the 5-U BTX-A-treated group (P < 0.001 for both groups). The mean intensity values for S100 were 78.98 ± 4.36, 85.31 ± 3.09, and 96.46 ± 8.98 for the saline, 5-U BTX, and 10-U BTX treatments, respectively. The post hoc test revealed differences between the group treated with 10 U BTX-A and the other groups, with significantly higher values in the 10-U BTX-A group than in the saline-treated control and the 5-U BTX-A-treated group (P = 0.002 and P = 0.038 for the saline- and 5-U BTX-A-treated groups, respectively).Fig. 3 Immunohistochemical staining for p65. a Saline-treated group; b 5-U BTX-A-treated group; c 10-U BTX-A-treated group (without counterstaining, bar = 20 μm). d Measurement of the average intensity of staining (*P < 0.05) Fig. 4 Immunohistochemical staining for S100. a Saline-treated group; b 5-U BTX-A-treated group; c 10-U BTX-A-treated group (without counterstaining, bar = 20 μm). d Measurement of the average intensity of staining (*P < 0.05) Discussion BTX-A has been used in a variety of oral and maxillofacial applications. BTX-A is primarily injected into masticatory muscles, such as the masseter muscle or the temporal muscle, for esthetic and therapeutic purposes in the maxillofacial area [3, 6]. A small number of studies that investigated the effects of BTX on masticatory muscles have been reported [7, 17]. However, short-term molecular-level changes in the masseter muscle after BTX-A injection have not been reported previously. In this study, p65 and S100 were increased in a dose-dependent manner in the rat masseter muscle 14 days after the BTX-A injection (Figs. 3 and 4). As the muscle underwent atrophic changes (Fig. 2), the thickness of the ramal cortical bone decreased 14 days after the BTX-A injection (Fig. 1). To our knowledge, this is the first report of dose-dependent early change in the masseter muscle after BTX-A injection. In this study, the thickness of the cortical bone in the mandibular ramus was decreased 14 days after BTX-A injection in comparison to the saline-treated control (Fig. 1). However, the difference observed among the groups was not statistically significant (Fig. 1d, P > 0.05). Previous publication demonstrates significant difference in bone thickness at 4 and 8 weeks after BTX-A injection [24]. This lack of significance might be due to the small sample size and short follow-up period. The thinner cortex of the mandibular ramus in the BTX-injected group might be similar to the disuse atrophy that occur secondary to masseter muscle atrophy. A number of previous studies reported that if the bone does not receive continuous stimulation from muscles, the bone will atrophy [8, 14]. When the patient having osteoporosis receives BTX-A injection, the optimal dosage of BTX-A should be carefully monitored. BTX-A injection results in an immediate reduction of electromyographic signals in the masseter muscle [5]. In this study, BTX-A-injected myofilaments exhibited a reduced size in comparison to the saline-injected control (Fig. 2). Generally, muscle atrophy is induced by systemic illnesses that contribute to muscle wasting. In this study, BTX-A injection reduced the size of myofibrils; this change should be interpreted as muscle atrophy (Fig. 2). A component of NF-kB (p65) is activated during cytokine-induced myotubule atrophy [25]. In addition, serum S100 induces myoblast apoptosis via the stimulation of reactive oxygen species [26]. In this study, the expression of NF-kB and the S100 was increased in the BTX-A injected groups in comparison to the saline-treated control. When the apoptosis of cells occurs, the expression of NF-kB is increased [14, 27]. When the calcium concentration in the tissue is increased, the expression of S100 also is increased [28, 29]. The observed increase in both p65 and S100 might be caused by muscle atrophy or apoptotic stress after BTX-A injection. When muscle apoptosis occurs, intracellular calcium ions may be released into the extracellular space [30]. As S100 is a calcium-binding protein, increased calcium release may increase the expression of S100. However, S100 protects against myoblast apoptosis [31]. Thus, increased expression of S100 might occur due to the protection of myoblasts from apoptosis after BTX-A injection. BTX-A induced myoblast apoptosis is also confirmed in recent publication [32]. Conclusions In this study, the injection of BTX-A into the masseter muscle induced muscle atrophy. Subsequently, p65 and S100 expression in myoblasts were increased for the protection of muscle cells. Acknowledgements This work was carried out with the support of “Cooperative Research Program for Agriculture Science and Technology Development (Project No. PJ01121404),” Rural Development Administration, Republic of Korea. The authors appreciated Jang-Ha Lee and Min-Keun Kim for their help in the animal experiment. Authors’ contributions KSG conducted most of the experiment, and JYY designed the experiment. KSG and PYW wrote the manuscript and performed a critical review on the experimental process. All authors read and approved the final manuscript. Competing interests The authors declare that they have no competing interests. Ethics approval and consent to participate All procedures were conducted according to the guidelines of laboratory animal care and were approved by the Gangneung-Wonju National University for animal research (GWNUA-2015-25). ==== Refs References 1. Burgen A Dickens F Zatman L The action of botulinum toxin on the neuro-muscular junction J Physiol 1949 109 10 23 10.1113/jphysiol.1949.sp004364 15394302 2. Schiavo G Benfenati F Poulain B Rossetto O Polverino de Laureto P DasGupta BR Tetanus and botulinum-B neurotoxins block neurotransmitter release by proteolytic cleavage of synaptobrevin Nature 1992 359 832 835 10.1038/359832a0 1331807 3. Su JZ Cai ZG Yu GY Microvascular autologous submandibular gland transplantation in severe cases of keratoconjunctivitis sicca Maxillofac Plast Reconstr Surg 2015 37 5 10.1186/s40902-015-0006-4 25664315 4. Navarrete AL Rafferty KL Liu ZJ Ye W Greenlee GM Herring SW Botulinum neurotoxin type A in the masseter muscle: effects on incisor eruption in rabbits Am J Orthod Dentofacial Orthop 2013 143 499 506 10.1016/j.ajodo.2012.10.023 23561411 5. Park SY Park YW Ji YJ Park SW Kim SG Effects of a botulinum toxin type A injection on the masseter muscle: an animal model study Maxillofac Plast Reconstr Surg 2015 37 10 10.1186/s40902-015-0010-8 25938091 6. Seok H Lee SW Kim MK Kim SG Park YW Park SW Correction of post-traumatic lower lip asymmetry using botulinum toxin type A J Korean Assoc Maxillofac Plast Reconstr Surg 2013 35 256 259 7. Ansved T Odergren T Borg K Muscle fiber atrophy in leg muscles after botulinum toxin type A treatment of cervical dystonia Neurology 1997 48 1440 1442 10.1212/WNL.48.5.1440 9153487 8. Kün-Darbois J-D Libouban H Chappard D Botulinum toxin in masticatory muscles of the adult rat induces bone loss at the condyle and alveolar regions of the mandible associated with a bone proliferation at a muscle enthesis Bone 2015 77 75 82 10.1016/j.bone.2015.03.023 25857689 9. Tsai CY Lei YY Yang LY Chiu WC Changes of masseter muscle activity following injection of botulinum toxin type A in adult rats Orthod Craniofac Res 2015 18 202 211 10.1111/ocr.12095 26083593 10. Blitzer A Brin MF Greene PE Fahn S Botulinum toxin injection for the treatment of oromandibular dystonia Ann Otol Rhinol Laryngol 1989 98 93 97 10.1177/000348948909800202 2916831 11. Kim HS Yun PY Kim YK A clinical evaluation of botulinum toxin-A injections in the temporomandibular disorder treatment Maxillofac Plast Reconstr Surg 2016 38 5 10.1186/s40902-016-0051-7 26855937 12. von Lindern JJ Niederhagen B Appel T Bergé S Reich RH Type A botulinum toxin for the treatment of hypertrophy of the masseter and temporal muscles: an alternative treatment Plast Reconstr Surg 2001 107 327 332 10.1097/00006534-200102000-00004 11214045 13. Duchen L Strich SJ The effects of botulinum toxin on the pattern of innervation of skeletal muscle in the mouse Q J Exp Physiol Cogn Med Sci 1968 53 84 89 4297234 14. Arora R Yates C Gary BD McClellan S Tan M Xi Y Panepoxydone targets NF-kB and FOXM1 to inhibit proliferation, induce apoptosis and reverse epithelial to mesenchymal transition in breast cancer PLoS One 2014 9 6 e98370 10.1371/journal.pone.0098370 24896091 15. Warden SJ Galley MR Richard JS George LA Guildenbecher EA Judd AM Reduced gravitational loading does not account for the skeletal effect of botulinum toxin-induced muscle inhibition suggesting a direct effect of muscle on bone Bone 2013 54 98 105 10.1016/j.bone.2013.01.043 23388417 16. von Lindern JJ Niederhagen B Bergé S Appel T Type A botulinum toxin in the treatment of chronic facial pain associated with masticatory hyperactivity J Oral Maxillofac Surg 2003 61 774 778 10.1016/S0278-2391(03)00153-8 12856249 17. Moon YM Kim YJ Kim MK Kim SG Kweon HY Kim TW Early effect of Botox-A injection into the masseter muscle of rats: functional and histological evaluation Maxillofac Plast Reconstr Surg 2015 37 46 10.1186/s40902-015-0049-6 26753166 18. Ilg EC Schäfer BW Heizmann CW Expression pattern of S100 calcium‐binding proteins in human tumors Int J Cancer 1996 68 325 332 10.1002/(SICI)1097-0215(19961104)68:3<325::AID-IJC10>3.0.CO;2-7 8903474 19. Mischke D Korge BP Marenholz I Volz A Zieqler A Genes encoding structural proteins of epidermal cornification and S100 calcium-binding proteins form a gene complex (“epidermal differentiation complex”) on human chromosome 1q21 J Invest Dermatol 1996 106 989 992 10.1111/1523-1747.ep12338501 8618063 20. Heatwole VM (1999) TUNEL assay for apoptotic cells. Immunocytochemical Methods and Protocols. Springer, Heidelberg, p 141–148 21. Wu SH Huang SH Cheng KI Chai CY Yeh JL Third-degree hindpaw burn injury induced apoptosis of lumbar spinal cord ventral horn motor neurons and sciatic nerve and muscle atrophy in rats BioMed Res Int 2015 2015 372819 25695065 22. Magnusson-Lind A Davidsson M Silajdžić E Hansen C McCourt AC Tabrizi SJ Skeletal muscle atrophy in r6/2 mice-altered circulating skeletal muscle markers and gene expression profile changes J Huntingtons Dis 2014 3 13 24 25062762 23. Kim HC Song JM Kim CJ Yoon SY Kim IR Park BS Combined effect of bisphosphonate and recombinant human bone morphogenetic protein 2 on bone healing of rat calvarial defects Maxillofac Plast Reconstr Surg 2015 37 16 10.1186/s40902-015-0015-3 26161381 24. Vegger JB Brüel A Thomsen JS Vertical trabeculae are thinned more than horizontal trabeculae in skeletal-unloaded rats Calcif Tissue Int 2015 97 516 526 10.1007/s00223-015-0035-0 26163234 25. Yamaki T Wu CL Gustin M Lim J Jackman RW Kandarian SC Rel A/p65 is required for cytokine-induced myotube atrophy Am J PhysiolCell Physiol 2012 303 C135 C142 10.1152/ajpcell.00111.2012 26. Sorci G Riuzzi F Agneletti AL Marchetti C Donato R S100B causes apoptosis in a myoblast cell line in a RAGE‐independent manner J Cell Physiol 2004 199 274 283 10.1002/jcp.10462 15040010 27. Seo BN Ryu JM Yun SP Jeon JH Park SS Oh KB Delphinidin prevents hypoxia-induced mouse embryonic stem cell apoptosis through reduction of intracellular reactive oxygen species-mediated activation of JNK and NF-kB, and Akt inhibition Apoptosis 2013 18 811 824 10.1007/s10495-013-0838-2 23584725 28. Heizmann CW The calcium signal is transmitted into an intracellular response via the families of calcium-binding proteins which are thought to be involved in the regulation of many cellular activities. Calcium Transport and Intracellular Calcium Homeostasis 2013 Lyon NATO Advanced Research Workshop 301 29. Mishra SK Siddique HR Saleem M S100A4 calcium-binding protein is key player in tumor progression and metastasis: preclinical and clinical evidence Cancer Metastasis Rev 2012 31 163 172 10.1007/s10555-011-9338-4 22109080 30. Yu SP Canzoniero LM Choi DW Ion homeostasis and apoptosis Curr Opin Cell Biol 2001 13 405 411 10.1016/S0955-0674(00)00228-3 11454444 31. Riuzzi F Sorci G Donato R S100B stimulates myoblast proliferation and inhibits myoblast differentiation by independently stimulating ERK1/2 and inhibiting p38 MAPK J Cell Physiol 2006 207 461 470 10.1002/jcp.20580 16419039 32. Moon YM Kim MK Kim SG Kim TW Apoptotic action of botulinum toxin on masseter muscle in rats: early and late changes in the expression of molecular markers Springerplus 2016 5 991 10.1186/s40064-016-2680-9 27398270

PMC005xxxxxx/PMC5002052.txt

==== Front Carbon Balance ManagCarbon Balance ManagCarbon Balance and Management1750-0680Springer International Publishing Cham 6010.1186/s13021-016-0060-yResearchComparison of national level biomass maps for conterminous US: understanding pattern and causes of differences Neeti N. neeti@teriuniversity.ac.in 1Kennedy R. rkennedy@coas.oregonstate.edu 21 Department of Natural Resources, TERI University, New Delhi, India 2 College of Earth, Ocean, Atmospheric Sciences, Oregon State University, Corvallis, OR USA 26 8 2016 26 8 2016 12 2016 11 1927 1 2016 17 8 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.Background As Earth observation satellite data proliferate, so too do maps derived from them. Even when two co-located maps are produced with low overall error, the spatial distribution of error may not be the same. Increasingly, methods will be needed to understand differences among purportedly similar products. For this study, we have used the four aboveground biomass (AGB) maps for conterminous US generated under NASA’s Carbon Monitoring System. We have developed systematic approach to (1) assess both the absolute accuracy of individual maps and assess the spatial patterns of agreement among maps, and (2) investigate potential causes of the spatial structure of agreement among maps to gain insight into reliability of methodological choices in map making. Results The comparison of the four biomass maps with FIA based total biomass estimates at national scale suggest that all the maps have higher biomass estimate compared to FIA. When the four maps were compared among each other, the result shows that the maps S and K have more similar spatial structure whereas the maps K and W have more similar absolute values. Although the maps K and W were generated using completely different methodological workflow, they agree remarkably. All the maps did well in the dominant forest type with maximum agreement between them. The comparison of difference between maps S and K with regional maps suggests that these maps were able to capture the disturbance and not so much regrowth pattern. Conclusions The study provides a comprehensive systematic approach to compare and evaluate different real data products using examples of four AGB maps. Although ostensibly the four maps map the same variable, they have different spatial distribution at different scale. Except the 2003 map, one can use other maps at the coarser spatial resolution. Finally, the disparate information available through different maps indicates a need for a temporal framework for consistent monitoring of carbon stock at national scale. Keywords CMSAGBFIACarbonissue-copyright-statement© The Author(s) 2016 ==== Body Background Because forests provide important ecosystem services and play a role in the global carbon cycle, forest characteristics such as biomass, tree height, and percent forest cover have been mapped using a variety of remote sensing techniques [1–7]. As data increase in availability, and as mapping techniques proliferate, many similar products are becoming available for the same region—often at different spatial scales and derived using different techniques. In an ideal world, multiple mapped estimates of the same biophysical variable should largely agree in the same location and time. Presumably, all credible maps are verified against a more reliable source to estimate overall error, but the spatial distribution of errors may not be the same for different maps. Thus, when two maps of the same quantity are compared at a given location, they may disagree. From an end-user perspective, this is a problem: different maps may have quite different implications for carbon accounting, for example, but users are given no guidance about which map to choose. From a scientific perspective, this may be an opportunity: patterns of disagreement among maps may provide insight into how datasets and techniques perform under varying conditions. Reconciling maps will become an increasingly important activity within NASA’s Carbon Monitoring System (NASA-CMS) [8]. NASA-CMS is a broad initiative to apply NASA’s synoptic view of Earth systems to the monitoring of carbon. NASA-CMS activities run the gamut from local-scale mapping of carbon state [9, 10] to global scale mapping of carbon flux [11]. Often, projects at different scales produce similar products, or products that could be compared against other projects through simple manipulation (e.g., multiple estimates of biomass over time could be compared against estimates of flux). As a mix of both regional scale and global scale projects, NASA-CMS will be increasingly faced with disparate estimates of carbon states and fluxes at different scales. Recognizing that different estimates of carbon-related maps co-occur, several researchers have recently reported on comparisons among different biomass maps [12–14]. In some cases, comparisons among maps were made with fine-resolution reference data (e.g., [13, 14]), but such data are not always available. In other cases, maps were compared at the same spatial resolution (e.g., [12]) but this precludes comparisons generated at different resolution. In fact, there are often situations where we need to identify a map which is most close to the truth in the absence of fine-resolution reference data while understanding various methodological and input grain size differences between various available maps. Indeed, the notion of multiple maps may imply that one is better, or that one more closely matches truth, but such a uniform truth is rarely feasible, and therefore we need spatial distribution of uncertainty. In practice, the assessment of spatially-distributed map data can be considered an effort to paint a picture of spatial uncertainty which is carried out by describing relative error among maps or measurements using a suite of complementary quantitative measures [15]. However, there exist several challenges in assessing a map and comparing different maps. First, the unit of analysis in a raster-based map is a square pixel that is arbitrary when we compare to any phenomenon in the landscape. Second, raster based maps are often generated at different scales, so comparison requires cell by cell alignment while ensuring the information in individual maps remain same. Moreover, most of the techniques developed to compare real variable maps can be used only for pairwise comparison, limiting the ability of simultaneous comparison of multiple maps to assess the spatial pattern of similarity and dissimilarity. As the NASA-CMS matures, mapped estimates of both carbon biomass and flux will be available at different scales for the same location, requiring quantitative comparisons among many maps. Indeed, NASA-CMS has already produced one forest biomass map for the conterminous United States, but already several versions existed developed by other groups. For users, guidance must be given about the relative merits of these maps under different conditions, and under what conditions a given map may be avoided. For developers and scientists, however, comparisons should be structured to leverage the differences in methods to provide insight into possible improvements or best practices. Thus, using these several forest biomass maps as a test case, we report on strategies and methods both to describe differences among maps for users, and to evaluate possible sources of disagreement. We therefore had two broad objectives:Descriptive: Assess both the absolute accuracy of individual maps and assess the spatial patterns of agreement among maps. Evaluative: Investigate potential causes of the spatial structure of agreement among maps to gain insight into reliability of methodological choices in map making. To achieve these objectives, we introduce a comprehensive and systematic approach to compare multiple maps generated at different scales (extent and spatial resolution). This comprehensive approach not only evaluates accuracy of individual maps while describing the similarity and dissimilarity between the spatial structure of various maps, but also systematically explores effects of scale and various causes of change. Methods Our systematic approach consists of several steps (Fig. 1). The foundational step is a thorough assessment of key map-making steps. This provides a context for the descriptive phase, and helps guide hypotheses to be tested in the evaluative phase.Fig. 1 Overall systematic approach in reconciling maps. a Descriptive phase involves assessing accuracy of individual maps and both pairwise and simultaneous comparison of multiple maps, b evaluation phase involves detailed assessment of statistical approach to understand implications of identified sources of differences Next, the descriptive phase seeks to quantify how well individual maps agree with reference data and with each other. Two key strategies are important. First, in addition to evaluating pairwise differences, as is done in other comparative studies, we advocate simultaneous comparison across all maps to understand underlying spatial patterns of agreement and potentially isolate outlier maps. Second, we aggregate our measures of agreement and disagreement to ecologically coherent mapping regions, recognizing that some maps may perform better in particular ecological contexts and that users may only be interested in this regional scale. Finally, the evaluative phase seeks to test specific sources of error in the map making process. Here, we must draw inference from comparisons without the benefit of manipulation. Thus, rather than simply developing a range of comparative metrics, this phase places comparisons into specific test of methodological contrast among maps. The key is to hypothesize how differences in methodology would lead to specific differences among pairs of maps, and then test those hypotheses. It is the combination of metric and map combination that allows greater inference: the observation of an effect in a single map is likely uninformative, but hypothesis-driven comparison of that effect between maps may provide greater insight. Assessment of map-making steps We examined the four national-scale maps of forest biomass that were available at the beginning of our study (Fig. 2). Papers detailing map production are given by Saatchi et al. [16], Kellndorfer et al. [17], Blackard et al. [18], and Wilson et al. [19]. For parsimony, we hereafter refer to these maps by their first letter, i.e. S, K, B, and W, respectively. All maps are generally considered usable, either through self-reported accuracies or through community use. Map W is reported to have strong agreement with plot based estimates of biomass (agreement coefficient ~0.99) and to have a strong goodness of fit (within 90 % confidence interval). The average absolute error for the map B ranges between 40 and 60 metric tons per hectare, except for the higher-biomass areas of the Pacific Northwest (163 metric tons per hectare). The accuracies for map S and K are not known, but they, like all of the maps, are already being used as input in different studies [10, 20, 21].Fig. 2 Aboveground biomass (AGB) density maps for conterminous US a Saatchi et al. [16] ‘S’, b Kellndorfer et al. [17] ‘K’, c Blackard et al. [18] ‘B’, d Wilson et al. [19] ‘W’ Like many spatial mapping exercises in remote sensing, the general approach for producing AGB maps is to extrapolate high quality training data from a small sample of locations to a large, contiguous space. Typically, values measured at the training samples are linked with data values from geospatial datasets at the locations where they intersect. Statistical models built at those locations are then used to extrapolate to the rest of the map. Maps can differ in how they handle training data, which geospatial datasets are used, and what statistical models are built. We identified six methodological sources of disagreement in how the four biomass maps were generated (Table 1). For each, we developed expectations for how that difference may manifest itself in the resultant maps. The first source is the forest mask used to define the total area for which biomass is estimated. Conservative forest masks would lead to lower estimates of total biomass, but a more liberal forest mask would include a higher proportion of marginal forest and thus likely reduce average biomass density. The second source of potential disagreement between the maps is spatial resolution. Map resolution interacts with the fundamental spatial structure of the landscape: If the former is finer than the latter, the map can adequately capture the range of variability of the landscape. A third contrast among maps is the remote sensing source data used for extrapolation. Maps using optical data may be less able to capture high biomass than maps using radar data, as optical data are known to saturate at lower biomass compared to radar (longer wavelength such as L- and P-band) [22–25]. The sensitivity of L- band SAR data for biomass estimation increases if used along with forest height generated using InSAR [25]. The fourth source of difference is the statistical technique used for extrapolation (parametric vs. non-parametric). Non-parametric techniques could be expected to perform better at extreme values (low and higher biomass region) than parametric techniques, especially if the extreme values deviate greatly from a normal distribution. The fifth source of difference is year of mapping. Maps produced in different years would be expected to differ both because of intervening disturbance and growth, and because the pool of training data could be different.Table 1 Sources of disagreement among the four national AGB maps AGB maps Forest mask Grain size (m) Spatial predictors Extrapolation technique (parametric/non-parametric) Year of generation of map Allometric equation Saatchi et al. [16] (S) NLCD 2006 93 F(MODIS, Landsat, L-band PALSAR, GLAS, topography) Maximum Entropy (Parametric) 2005 Component ratio method Kellndorfer et al. [17] (K) NLCD 2001 30 F(C-band InSAR, Landsat, NLCD canopy density, structure, SRTM) Regression tree (Random forest, Non-parametric) 2000 Regional method Blackard et al. [18] (B) In house 250 F(MODIS, climate variables variables, topography) Regression tree (Cubist, Non-parametric) 2003 Regional method Wilson et al. [19] (W) No Forest mask 250 F(MODIS, climate variables, topography, level III ecoregions) Phenological gradient nearest neighborhood (Semi-parametric/Semi-nonparametric) 2009 Component ratio method The sixth area of potential disagreement rests not with the methodologies to extrapolate, but with the training data themselves. In the case of forest AGB in the conterminous US, the training data come from the US Forest Service’s Forest Inventory and Analysis (FIA) program. At each of thousands of field plots, field crews measure details of trees using a regular sampling and mensuration protocol. These raw tree data can then be converted to estimates of plot-level biomass using allometric equations that relate tree characteristics to known biomass measured destructively at a sample of locations (and often as an entirely separate effort). Allometric equations varied across the four maps studied here, with some using regional-specific equations and others a more nationally-consistent component ratio method (CRM) [26]. The CRM approach has been shown to lead to lower biomass estimates then the regional approach [26], and thus maps based on the CRM would thus be expected to show an overall bias toward lower values. Descriptive comparisons Descriptive comparisons provide users guidance about which maps are more accurate, and where on the landscape (at both national and regional scales) the maps agree and disagree when analyses carried out at pixel level. Comparison with ground reference data Because the FIA program is tasked with providing defensible estimates of forest resources at the national scale, estimates from FIA plot data are the de facto standard against which any forest resource maps must be compared. Although FIA plot data are used as inputs in various points of the mapping process for all four national maps tested here, this does not guarantee that summarized estimates of biomass will agree, since maps extrapolate the FIA information differently. Comparison with FIA data requires appropriate use of those data. The goal of the FIA sampling design is to provide good estimates at aggregated administrative levels: The smallest unit is typically the US county or parish, but state-level comparisons are more common and robust, especially when the analyses is carried out at the MODIS scale [27]. Additionally, normal users do not have access to actual plot locations, and thus data are typically available and readily usable only at county or larger scales. Thus, for our comparisons we compared plot and map data at the aggregation level of the state. Map data aggregation was a simple matter of summing biomass density estimates to the state level, utilizing each map’s own forest mask for the aggregation footprint. For plot data, we acquired state level FIA AGB data from the FIA 2013 database [28] at the appropriate time step for the FIA collection strategy of the state. For forests in the eastern US, we used the time frame of 2003–2007, as that was the most consistent time frame without double counting any trees across different states. For forests in the western US, which are on a 10 year repeat cycle, we used data from 2000 onwards for the west side as that was the most consistent data available through the website across all the states at the time of analyses. The estimates have already converted the plot-level tree measurements to the plot scale, and then used sample-design considerations to scale the plot-level data to the state level. Three considerations of the FIA data are relevant for later comparisons. First, the FIA uses the CRM approach to calculate biomass from tree measurements. Second, the inventory cycles occur at 5 and 10 year cadences in the east and west, respectively, meaning that an estimate at a given time will be a different mix of older and more recent plots in the east and the west. Finally, the extrapolation scaling factors are based on forestland masks developed by the US Forest Service (and also used in the B and W maps). Once biomass had been aggregated, we produced three descriptive products. First, we summed forest biomass among all maps and all plots at the conterminous US scale. Second, we used simple regression of aggregated map biomass at the state level against FIA plot biomass (n = 48 states in conterminous US). Finally, we calculated the Wilmott’s index of agreement (d): d=1-∑i=1n(Xi-Yi)2(|Xi-X̲|+|Yi-X̲|)2X where Xi was the biomass estimate for state i estimated from FIA plot data and Yi was the same estimated from the maps [29]. The X variable is considered to be the truth variable, which is appropriate here because FIA plot data are the national standard for forest monitoring. Wilmott’s index of agreement is symmetric, bounded and does not over penalize for disagreement. High agreement is indicated by values close to 1.0. There exist many other indices for such pair wise comparison such as mean square error (MSE) and its root (RMSE) to more specialized metrics [e.g., agreement coefficient (AC)] to comparing distribution. The indices such as MSE, RMSE have been critiqued for being unbounded and asymmetric [30] while the AC is found to highly sensitive to outlier. The comparison of distributions could simply yield the uninteresting finding of differences with little insight on the reason behind the differences. Spatial structure of agreement and disagreement is critically useful for understanding whether disagreement is potentially related to specific issues. Comparison of multiple maps After descriptive comparison against a trusted data source (FIA based estimates), the next step was to show where maps agreed and disagreed. We used principal components analysis (PCA) for simultaneous comparison among all maps. PCA is a spectral decomposition technique commonly used in remote sensing to remove the redundancy from multi-spectral images, but it can also be used to identify the dominant pattern common among various maps [31]. The six orientation modes (O, P, Q, R, S, T) commonly used for PCA differ in their definition of statistical variables and observations [32]. In this case, statistical observations are samples in space and statistical variables are various continuous maps, therefore, R-mode PCA is used for the analysis. Implementation of the PCA took place in the R statistical package [33], and required some basic preparation of the datasets. First, because the maps reported biomass in different unit systems and different map projections, we aligned all biomass map values to the same system at the pixel scale, and then aggregated (taking mean of x by x window size followed by nearest neighborhood resampling, x = 8 for map S, x = 2) the finer-scale maps (S and K) to the 240 m grain size of the coarser maps to make them comparable to the other two maps. The coarser resolution maps (B and W) were resampled from 250 m to 240 m resolution by using nearest neighborhood approach. Second, we clipped all maps to the forest mask area common to all forest masks, as “no-data” would not be informative in the PCA analysis. Finally, we stacked all four maps into a single, four-layer image. Once these preparatory steps had been taken, we ran the PCA analysis on both the entire conterminous US data (at the 240 m pixel scale) and at the scale of each of the 66 mapping regions [34]. The regional scale analysis is useful because the PCA statistical space is defined by the range of variation in the whole dataset, and thus the broad US-wide comparison may obscure patterns that would be relevant to users at the regional scale. Results of the PCA were interpreted in two ways. First, spatial patterns in the first and second axis images show where on the landscape the maps agree and disagree. Because PCA axis 1 identifies the vector through the multivariate space that explains the most variation, it can be interpreted as the overall pattern of biomass agreement across all maps. PCA axis 2 is orthogonal to the first axis, and thus indicates the dominant sources of disagreement among the maps. Second, the correlation coefficients of each map contributing to those two axes can give insight into which maps are agreeing or disagreeing with the other maps. A map with high correlation coefficient on the first axis is one whose spatial patterns agree with the other maps; a map with high correlation coefficient on the second axis is one whose spatial patterns disagree with the other maps. Scale of agreement If maps generally agree with truth data when aggregated to the mapping region scale, but show patterns of disagreement at the pixel scale, then a natural question from users is whether intermediate scales of resolution may make maps more comparable. To evaluate this question, we conducted pair-wise map comparison to analyze the impact of spatial resolution on map agreement. As with the PCA analysis (‘‘comparison of multiple maps’’ section), we first ensured that all maps were clipped to the same footprint and were at the same starting pixel resolution (acknowledging that the 30 m and 90 m products were already degraded for the first comparison), and then we aggregated from the starting 240 m resolution to pixel resolutions of 480, 720 and 960 m. For each of those resolutions, we evaluated paired map agreement with the Wilmott’s index of agreement at mapping region scale, as all the pixels in a mapping region are expected to have similar ecological condition. Critically, because no map could be considered the truth map, we conducted this analysis on all pairwise comparisons. High median d-scores across all mapping regions suggest consistent strong agreement, while high range of d-scores suggests spatial patterns of variability across mapping regions. Evaluative comparisons Evaluative comparisons were designed to test whether patterns of agreement and disagreement could provide insight into sources of error in map production. Because each map was produced using a different suite of methods (Table 1), each map needed to be compared individually to the other maps. To test the six different possible sources of differences, we used a range of different pairwise map comparisons, both on the original data and on maps of differences between pairs of maps. Additionally, for certain tests we brought in ancillary data on forest type, topographic position, and regional biomass time series. We first describe the basic data manipulations, and then describe how these were used to evaluate the six different sources of map difference. Data manipulations Cumulative biomass The simplest data manipulation was simple summing of biomass across all pixels in a given map to estimate total forest biomass at the national scale. Total biomass at the national scale had already been calculated for the descriptive phase (see section on ‘‘cumulative biomass’’ section). Mapping region scale biomass density Within each mapping region, we calculated the cumulative frequency distribution of biomass density values of a map for all pixels in each map’s forest mask. From these, we identified the 10th-, 50th- and 90th percentile values for the mapping region. We also calculated mean biomass density by mapping region. Difference maps For all pairwise combinations of the four national-scale maps, we developed maps of biomass difference at the pixel scale. As with the PCA analysis, the maps were first clipped to a common forest mask, and if the pixel sizes of the two maps were different, the higher resolution map was resampled and aggregated to the lower resolution map cell size (see details on aggregation above). Differencing was achieved using simple image algebra on a cell by cell basis. Spatial patterns in the difference maps were then related to spatial patterns of ancillary data. Ancillary data We sought to understand whether spatial patterns in map differences were related to spatial patterns in other geospatial datasets. For these, we obtained the spatial predictor data layer and manipulated the map as necessary (using mean/mode aggregation and/or nearest neighborhood resampling) to ensure cell alignment among maps. We obtained through the ORNL website [35] a forest type land cover map with cells of 250 m resolution, for the nominal mapping year 2001. We utilized all forest cover types for analysis. Separately, we obtained a forest age map with cell size of 1 km for the nominal mapping year 2006 [36] and grouped pixels into young (<40 years old), immature (40–80 years old), mature (80–140 years), and old (>140 years). Finally, for topographic variables, we obtained SRTM elevation at 30 m resolution and derived slope. We grouped pixels according to three elevation categories (<500 m, 500–1500 m, and >1500 m), and, where appropriate, two slope categories: relatively benign (0 to <30 degree slope) and extreme (30–50 degree slope). Within mapping regions of interest, we summarized the difference maps according to the zones delineated by the groups in those ancillary data sources. To assess impact of year of map generation, we required ancillary biomass map data where technique of production was held constant across time. At present no such map exists at the national scale, and thus we used a regional-scale map product [37]. That product generated yearly biomass estimates from 1990 to 2010 for the high-biomass states of Washington, Oregon, and California. Because the core of that approach was a change detection approach at the Landsat scale, those maps explicitly represent the effects of growth and disturbance on biomass. We refer to these maps as KennXXXX, where XXXX represented the year of map. We differenced KennXXXX maps for the years corresponding to each of the map pairs in the national efforts, and compared distributions of the differences in the consistently-produced maps to the differences in distributions in the national maps. Note that we not using this map as a truth dataset, but simply as a means of holding constant the means of production across time to isolate the impacts of growth and disturbance. Evaluative tests The data manipulations were then used in evaluative tests of each of the six identified sources of map error. We emphasize that evaluative tests utilize metrics that by themselves could be uninformative—they key is to compare the metrics among map pairs, driven by hypotheses about how map production could lead to differences. Forest mask Each map used a slightly different mask to identify potential forested pixels. If the forest mask were the only factor affecting map results, we would expect that maps with a conservative forest mask would show a lower total biomass when aggregated to the national scale. To test, we simply summed biomass nationally for all four maps and evaluated relative to the forest mask area of each map. Spatial resolution Spatial resolution determines the scale of pattern which can be captured by a given map. If spatial patterns of forest biomass vary meaningfully at a scale finer than that of a given map, that map will be unable to capture the high and low biomass values, and thus would be expected to have a compressed range of biomass relative to the actual landscape [24]. When comparing among maps of different resolution, the impact of resolution will matter if the scale of meaningful variation is intermediate between resolutions of the maps. To test, we compared 10th and 90th percentile values by mapping region (‘‘mapping region scale biomass density’’ section) for all pairs of maps. If pixel resolution were a driving factor, we would expect to see compressed ranges in both the W and B maps relative to the S and K maps. Source sensor data Because the signal retrieved from a passive optical sensor typically saturates at biomass levels lower than that of an active sensor such as radar (e.g., L-Band SAR, C-band InSAR) [26, 27], we might expect that maps involving radar would be able to track biomass better in high-biomass mapping regions even though there may be difference between various radar data depending on frequency and polarization. To test, we compared the 90th percentile values by mapping region for all pairs of maps. If sensor source were a driving factor, we would expect the W and B maps to show a compressed upper end relative to the S and K maps. Similarly, because older forests have more structurally complex canopies, we might expect an active-sensor approach to better capture biomass in older forests, or in generally higher-biomass forest types [38]. To test, we compared difference maps grouped by forest type and age for mapping regions where high biomass and older forests were more prevalent. If active sensors performed better in these systems, we might expect the S and K maps to show higher biomass (net positive difference) vs the B and W maps in older forests and in higher biomass type forests. Finally, because of the side-looking nature of radar, we might expect the radar-derived maps to perform more poorly in topographically complex areas [39]. To test, we compared difference maps grouped by elevation and slope categories (noted in section on the ancillary data in the method above). If error introduced by side-looking radar were an issue, we would expect the S and K maps to show greater difference with the optical maps in either high elevation or high slope areas. Extrapolation technique When linking the observed covariate (here, biomass) to predictor data (here, geospatial data), parametric regression techniques explicitly attempt to minimize variability in the covariate, often resulting compressed prediction ranges when extrapolation is later performed [40] low values are overestimated and high values are underestimated. Non-parametric approaches may not suffer from this extrapolation issue. To test, we examined the 10th and 90th percentiles of biomass distributions. If extrapolation technique were an issue, we would expect the W and S maps to show ranges compressed relative to the K and B maps. Nominal year of map National maps were produced in different years, and thus some differences between any pair of maps would be caused by intervening growth and disturbance in the period between the two maps. Thus, the observed difference between maps convolves both map production differences and real differences in the landscape. To test, we compared difference maps to differences in the KennXXXX map pairs for the same period. Because the KennXXXX maps explicitly capture disturbance and growth processes using a consistent technique, they serve as a proxy for the expected differences between maps for any given pair of years. If two maps agree but differ only in disturbance and growth, we would expect the distributions of biomass difference to be similar to that in the proxy maps. Substantial departures from that expected distribution would indicate that year of map production was not the only cause of difference in the maps. Allometric equations Because the CRM method used nationally has been shown to produce lower estimates of biomass than regionally-specific allometric equations, we would expect mean and median values of biomass to be lower in those maps using CRM approaches. To test, we compared median and mean values by mapping region. If allometric equations were a critical difference, we would expect the S and W maps to be consistently lower than the K and B maps. Results and discussion Descriptive comparisons Comparison with ground reference data All four maps estimate slightly higher national AGB than does the summed FIA plot data based estimates for conterminous US (Table 2). Of the four, map S’s estimate is most similar to FIA and map B’s most dissimilar.Table 2 Total forest area and biomass for conterminous US for the four AGB maps based estimates and FIA based estimates AGB map Forest area (Million Ha) Total biomass (Gt) K 414.8 27.4 B 260.47 29.6 S 350.23 26.7 W 486.14 27.2 FIA – 25.4 When regressing biomass on a state-by-state basis (Fig. 3), map S shows the closest alignment with the 1:1 line (slope: 1.02), but with greater scatter than both maps K and W (r2 = 0.98, 0.99). Map B remains substantially different at the state scale, particularly for medium to high biomass states. Wilmott’s d suggests that map W is most similar to the FIA estimates at state scale, followed closely by map S.Fig. 3 Comparison between FIA plots based estimates and the four AGB maps at state scale: a FIA vs S, b FIA vs K, c FIA vs B, d FIA vs W Comparison across multiple maps The national-scale PC1 image shows patterns of biomass that are generally in agreement among all four maps (Fig. 4a), with correlations strong between PC1 and all four source maps (Fig. 4c). As expected, forest biomass is highest in west-coast forests, lower in parts of the mountainous eastern interior forests and lowest in the interior west and north central states. Patterns of biomass disagreement (PC 2) are more spatially heterogeneous (Fig. 4b), and appear to be dominated by map B (Fig. 4d). The difference of map B is in the PCA analysis is consistent with its difference against the other maps in the state level comparison with FIA plot data (Fig. 3). Relative to PC comparisons at the national scale, mapping-region scale PC analyses can show more nuanced or even opposing patterns. Mapping region 1 (Washington state) and mapping region 52 (Wisconsin and Michigan state) provide useful examples. In mapping region 1 (Fig. 5), patterns of biomass agreement (PC 1, Fig. 5a, c) are as expected, with high biomass forests in the wetter western portions of the state and lower in the drier eastern portions of the state. Consistent with the national scale analysis, Map B dominates the disagreement vector (PC 2, Fig. 5b, d). However, in mapping region 52, Map W and map B dominate the disagreement vector (PC 2, Fig. 6b) even though the patterns of biomass agreement are largely as expected (Fig. 6a, c). The map W and map B has completely opposing patterns in the mapping region, and the spatial pattern is not at all related to the other two maps (Fig. 6b, d). Thus, the guidance for which map is less reliable may vary with mapping region, and points to the need for local users to develop local-scale comparisons.Fig. 4 Spatial pattern of agreement and disagreement among four AGB maps at pixel scale for conterminous US generated using PCA a Principal Component 1 (PC1) score. b Principal Component 2 (PC2) score. c Correlation between PC1 score and the four AGB maps. d Correlation between PC2 score and the four AGB maps Fig. 5 Spatial pattern of agreement and disagreement among four AGB maps at pixel scale in mapping zone 1 generated using PCA. a Principal Component 1 (PC1) score. b Principal Component 2 (PC2) score. c Correlation between PC1 score and the four AGB maps. d Correlation between PC2 score and the four AGB maps Fig. 6 Spatial pattern of agreement and disagreement among four AGB maps at pixel scale for mapping zone 52 generated using PCA. a Principal Component 1 (PC1) score. b Principal Component 2 (PC2) score. c Correlation between PC1 score and the four AGB maps. d Correlation between PC2 score and the four AGB maps Scale of agreement For all maps but Map B, the Wilmott’s measure of agreement between pair-wise maps suggests that agreement increases at coarser spatial resolution, potentially leveling off at the top end of the range of spatial resolution tested here (Fig. 7a). There is high consistency across all pair-wise comparisons at multiple spatial resolution (range of d values less than 0.4) except for the comparison with the map B. For Map B, similarity did not change with increasing resolution, and the range of disagreement was high (range of d values near 0.8). However, similar to mapping region results for PCA, the agreement between maps varies with regions (Fig. 7b). The comparisons of map S with the other three maps at 240 m resolution suggest that it has maximum agreement with the map K across all the mapping regions. The agreement of map S with B decreases in the eastern mapping regions (Fig. 7b).Fig. 7 Pairwise comparison between the four AGB maps a multiple resolution comparison at mapping zone scale where X1 = 240 m, X2 = 480, X3 = 720 m, X4 = 960 m; X = AGB map, b Comparison at 240 m at pixel level within each mapping zone Evaluative comparisons Given that these maps are being compared post hoc, no experimental manipulations of method can be designed to unambiguously identify causes of change. However, by designing evaluative tests to compare specific maps against each other under specific hypotheses of change, we can paint a richer picture of possible sources of difference. Thus, we first describe the results of the metric comparisons here, and then follow with an assessment of how those comparisons should be interpreted on a pairwise basis to test potential methodological sources of error. Cumulative biomass The same cumulative biomass results conducted for the descriptive phase apply to the evaluative phase (Table 1; Fig. 3). Notably map S and map W agree well with FIA, and maps W and B have higher estimates. This contrast is likely related to the commonality of allometric equation (CRM approach) between the FIA and S and W maps. Mapping regions scale biomass density Pairwise comparisons between the four maps at low, medium and high biomass density quantiles at mapping region scale parse out differences among maps more closely. In general, maps K and W track the 1:1 line at low, median, and high biomass, suggesting agreement across the biomass distribution and across all mapping regions (Fig. 8d, h, k). Map S agrees with Maps K and W at the median of the biomass distribution (Fig. 8d), but is generally higher than Maps K and W at both high and low ends of the range (Fig. 8g, j). Map B is higher than all other maps in nearly all situations, and particularly appears to have relatively higher biomass at the low end of the biomass range (Fig. 8g–i).Fig. 8 Pairwise comparison (least square regression) of mean and different percentiles of four AGB maps at mapping zone scale: mean (first row), median (second row), 10th percentiles (third row), and 90th percentile (fourth row) and each column in a row represents a particular AGB map as x-axis (a, d, g, j: S map as X axis; b, e, h, k: K map as X axis; c, f, i, l: W map as X axis)  Difference maps and relation to ancillary data We evaluated difference maps in relation to ancillary data for all 66 mapping regions. At the US scale, the difference between the map S and K suggests that the map S has higher biomass estimates compared to the map K at higher biomass regions, and lower biomass at lower biomass regions (Fig. 9a). Again, we use mapping region 1 to illustrate core principles and findings. In mapping region 1, Map S has higher biomass estimates than does Map K in the wetter, western part of the region, but that pattern is generally flipped when compared to Maps B or W (Fig. 9 b–d). Forest characteristics (Fig. 10a, b) and topography (Fig. 10c, d) were then related on a pixel by pixel level to the difference map, and difference median and range summarized by forest type (Fig. 11a), forest age (Fig. 11b), elevation (Fig. 11c), and slope (Fig. 11d). As before, Map B was an outlier in most comparisons in this mapping region. In most of the comparison with map B, the median biomass difference is farther from 0, the range of difference is larger with higher biomass estimates across the mapping region.Fig. 9 Pairwise differences between four AGB maps: a Difference between the map S and the map K for conterminous US, pairwise difference between AGB density maps in mapping zone 1: b S & K, c S & B, d S & W e K & B f K & W, and g W & B Fig. 10 Forest and topographical characteristics for mapping zone 1: a forest type b forest age, c elevation, and d slope Fig. 11 Relationship between pairwise AGB difference maps and various forest and topographic characteristics in mapping zone 1: a with forest type, b with forest age, c with elevation, and d with slope Testing sources of disagreement Metrics of comparison described above are the core from which tests can be developed. Again, the goal of the evaluative comparisons is to use specific map-pair comparisons by metric to infer whether certain methodological choices are contributing to map disagreement. Forest mask If forest mask area were a fundamental reason for disagreement, we would expect maps with conservative forest area to show less biomass. But comparing total forest and total biomass (Section on cumulative biomass above, Table 2), there is little evidence that forest mask was an important factor driving the difference among maps. Indeed, the map with the highest biomass estimates (Map B) had the most conservative forest mask. Spatial resolution For tests of the impact of spatial resolution, we compared the high-resolution maps (S and K) with the coarser-resolution maps (W and B). If spatial resolution were an issue we would expect coarse resolution maps to show compressed data ranges at both high and low biomass values relative to the higher resolution maps. Map S appears to show this effect, as it shows greater range in both the 90th and 10th percentiles compared to the W map (Fig. 8j, g). However, map K did not show this effect in comparison to map W (Fig. 8h, k), suggesting that resolution alone does not explain the differences in the maps. Source sensor data If active sensors are able to map higher biomass without saturating, we would expect maps W and B to saturate at high biomass relative to both map S and K. Map S does appear to estimate higher biomass (90th percentile) than all other maps (Fig. 8j), but this effect does not occur with the other map created using an active sensor (Map K). However, this effect may be associated with the use of L-Band PALSAR in the map S. Map K uses C-band InSAR, which is less sensitive to the higher biomass compared to the higher wavelength [23]. Thus, our results are consistent with the notion that a specific type of active sensor may provide greater response to high biomass areas. This pattern can be partially corroborated by focusing on the high-biomass mapping region 1 (Fig. 11a). There, map S shows higher biomass estimates compared to K (that uses C-band radar data for height measurement) for high biomass forests (Douglas fir or Hemlock/Sitka spruce; Fig. 11a). However, the contrast does not hold for high biomass age classes (>80 years old; Fig. 11b). Moreover, when compared to optical-based maps, the map K map does not consistently show higher biomass estimates in high biomass forests (Douglas fir or Hemlock/Sitka spruce; Fig. 11a) nor in high biomass age classes (>80 years old; Fig. 10b). Moreover, we see no evidence for topographic impacts on radar-derived maps. If topographic conditions were an issue for side-looking radar, maps S and K might show dampened range in regions of higher topography. But instead, we find that the range of difference values between either the S or K maps and the other maps show no patterns with topographic condition (Fig. 11d). We acknowledge that these maps also began as higher resolution sources, and thus errors of topography may be offset by improvements from resolution. Extrapolation technique We see no evidence that extrapolation technique influenced differences among the maps. Again, we focus on the high and low end of the biomass range, and ask whether maps that use parametric or semi-parametric approaches (maps S and W) are compressed relative to counterparts that use nonparametric approachs (maps B and K). In this case, neither the S nor W map shows compressed ranges in the 10th or 90th percentiles of biomass distributions relative to the other maps (Fig. 8). This argues that extrapolation technique may not play a large role in determining differences among maps. Nominal year of map Because maps differed in nominal year represented, we would expect possible differences due to disturbance and growth of forests between map years. We tested against a regional scale map whose production focused on disturbance and growth, and that had consistent methods across time (Fig. 12). Although weak, evidence suggests that between 40 and 45 % of the total difference between national maps was consistent with growth and disturbance in the regional-scale map (Table 3).Fig. 12 Relationship between change in AGB (Mg/ha) in the national scale maps and the regional maps for disturbances and regrowth pattern for the states of WA, OR and CA over the years: a between 2000 and 2003, b between 2000 and 2005, c between 2000 and 2009, d between 2003 and 2005, e between 2003 and 2009, f between 2005 and 2009 Table 3 Total biomass estimates in million tonnes in three western states (OR, WA, and CA) over four years: Comparison between national map and regional map based estimates States 2000 (Mt) 2003 (Mt) 2005 (Mt) 2009 (Mt) WA K: 1564.88 B: 2091.94 S: 1812.82 W: 1732.58 Kenn: 1547.72 Kenn: 1537.92 Kenn: 1526.93 Kenn: 1509.06 OR K: 1838.67 B: 2564.11 S: 2150.93 W: 1957.12 Kenn: 1765.63 Kenn: 1734.37 Kenn: 1727.32 Kenn: 1708.72 CA K: 2106.35 B: 2305.67 S: 1818.43 W: 1974.56 Kenn: 1815.66 Kenn: 1808.30 Kenn: 1810.52 Kenn: 1798.77 Allometric equations Allometric equation impacts may have played a role in some of the differences among maps. Median and mean biomass estimates from Map B (using regional scale equations) were consistently higher than all other maps (Fig. 8a–f), and Map W’s estimates (using CRM equations) were lower than Map K’s. However, Map S did not fit the pattern, showing higher median values than both Map K and W. Conclusions There exist several AGB maps generated using different set of input data and methodology. Although these maps describe the same biophysical variable, they vary both in quantity and spatial pattern. We provide a systematic approach to describe and evaluate the differences between maps at multiple scales, while assessing the accuracy of individual mapping effort at an aggregated scale. This study emphasizes that comparison between maps need to be structured with specific hypothesis and tests. Moreover, we suggest that simple directives about which map to use are perhaps overly simplistic. Indeed, the answer for defining various tests depends on the need of the user and scale of comparison and therefore one needs to have set of methodology at different scale with different identified steps. The results from this systematic analysis on comparison of four national level AGB maps suggest that the absolute accuracy and spatial pattern of agreement vary with scale (both spatial resolution and spatial extent). Three (S, K and W) of the four maps largely agreed, and the two maps (K and W) generated with quite different methodological workflows agreed remarkably well. One map generated with high resolution, active sensor data appeared to capture a greater range of data. Finally, the map that did not agree at the pixel scale continued to disagree even with aggregation, indicating aggregation alone may not make maps similar. When compared with FIA based AGB summed at national scale—which is also basis for national level carbon accounting—all four maps slightly overestimate biomass. However, the total summed AGB for map S is most similar to FIA based estimates, and would therefore argue that the map S could be chosen among the four maps for national level analysis. However, the map W is most similar to FIA based estimates when aggregated to the state scale. The comparison among biomass maps at pixel level (240 m resolution) using R-mode PCA suggests that spatial structure of agreement varies at national and regional scale (spatial extent). Map B disagrees most with respect to the other maps when analyzed at national scale, but it is not always true when analyzed at the mapping region scale. Thus, one needs to look at both regional and broad scale differences before making decision about using one map over others for carbon accounting. The spatial structure of the maps S and K have maximum agreement at national scale PCA analysis, thus they agree well with each other as well as FIA based estimates. The spatial structure of agreement also varies with the spatial resolution. Except for map B, the agreement among other maps increases as spatial resolution coarsens. Maximum agreement occurs in the northwest (higher biomass) mapping region of the country. The changes in agreement with aggregation varied between maps, and were not same for the three maps (leaving map B aside). The structure of the agreement and disagreement among the AGB maps achieved by testing the full suite of potential sources of differences provides evidence in support of causes of differences. For example, all the maps did best in the dominant forest type of a given region, but variability was found in non-dominant forest types. The comparison of maps with regional efforts provided information about how well these maps were able to capture regrowth and disturbance pattern. For example, the comparison of the difference between the maps K and S with regional map suggest that the locations where there is lower biomass in S compared to K is mainly due to disturbance. However, the difference between the two could not capture regrowth well. Thus, sequences of maps capture both actual change and the combined effects of each map’s error, and thus argues against using sequences of these maps for spatial monitoring of biomass over time. Our strategy for testing the full suite of potential error sources shows how remarkably consistent some maps were. Notably, the W and K maps differed in nearly every methodological approach, yet agreed closely at the mapping region scale. The magnitude of W and K matches closely even though the spatial structure differs whereas the spatial structure of S and K agrees (simultaneous comparison using PCA) but magnitude differs (pairwise comparison) when compared at ecoregion scale. While most of the maps generally agree at broad spatial scales, spatial patterns of disagreement at the local scale are notable. PCA analyses at both national and mapping region scale clearly show that the disagreement (PC2) has spatial pattern, and the improvement of agreement with spatial aggregation (Fig. 11) corroborates the notion that pixel-level estimates vary considerably among maps. Thus, a user at the regional scale would be advised to evaluate local scale variation among all maps before choosing one to get insight on the sources of differences (error or physical change in the landscape). With the understanding of differences, one can use ensemble approach to have an accurate map of aboveground biomass map. Thus, this study provides guidance how to approach comparison of multiple maps systematically by designing specific steps for various hypothesis for describing and evaluating the spatial pattern of differences between maps. Abbreviations NASA-CMSNational Aeronautics and System Administration Carbon Monitoring System MSEmean square error RMSEroot mean square error AGBabove ground biomass FIAForest Inventory and Analysis CRMComponent Ratio Method PCAPrincipal Components Analysis ORNLOak Ridge National Laboratory Authors’ contributions The study was designed by NN and RK. NN performed the analysis and produced the figures. NN and RK wrote the manuscript and have equal contribution. Both authors read and approved the final manuscript. Acknowledgements This work was part of NASA funded CMS project (Grant number NNX12AP76G, PI-Robert Kennedy) and a USDA NIFA funded project (Grant number 2011-67003-20458, PI-Robert Kennedy). We are thankful to Sassan Saatchi, Sangram Ganguly, Barry Tyler Wilson, Rachael Riemann and Josef Kellndorfer for providing their datasets and clarifying questions regarding the datasets. Competing interests The authors declare that they have no competing interests. ==== Refs References 1. Saatchi SS Houghton RA Alvala RCDS Soares JV Yu Y Distribution of aboveground live biomass in the Amazon basin Glob Change Biol 2007 13 4 10.1111/j.1365-2486.2007.01323.x 2. DeFries RS Hansen MC Townshend JRG Janetos AC Loveland TR A new global 1-km dataset of percentage tree cover derived from remote sensing Glob Change Biol 2000 6 2 10.1046/j.1365-2486.2000.00296.x 3. Powell SL Cohen WB Healey SP Kennedy RE Moisen GG Pierce KB Ohmann JL Quantification of live aboveground forest biomass dynamics with landsat time-series and field inventory data: a comparison of empirical modeling approaches Remote Sens Environ 2010 114 5 10.1016/j.rse.2009.12.018 4. Turner DP Cohen WB Kennedy RE Alternative spatial resolutions and estimation of carbon flux over a managed forest landscape in western Oregon Landscape Ecol 2000 15 5 10.1023/A:1008116300063 5. Simard M Zhang K Rivera-Monroy VH Ross MS Ruiz PL Castañeda-Moya E Twilley RR Rodriguez E Mapping height and biomass of mangrove forests in Everglades National Park with SRTM elevation data Photogramm Eng Remote Sens 2006 72 3 10.14358/PERS.72.3.299 6. Ni X Zhou Y Cao C Wang X Shi Y Park T Choi S Myneni RB Mapping forest canopy height over continental China using multi-source remote sensing data Remote Sensing. 2015 7 7 7. Yang T Wang C Li GC Luo SZ Xi XH Gao S Zeng HC Forest canopy height mapping over China using GLAS and MODIS data Science China Earth Sciences. 2015 58 1 8. http://carbon.nasa.gov. Accessed 1 Jan 2016. 9. Zhang G Ganguly S Nemani RR White MA Milesi C Hashimoto H Wang W Estimation of forest aboveground biomass in California using canopy height and leaf area index estimated from satellite data Remote Sensing Environ. 2014 151 44 10.1016/j.rse.2014.01.025 10. Huang W Swatantran A Johnson K Duncanson L Tang H Dunne JO Hurtt G Dubayah R Local discrepancies in continental scale biomass maps: a case study over forested and non-forested landscapes in Maryland, USA Carbon Balance Manag 2015 10 1 10.1186/s13021-014-0011-4 25632297 11. Liu J Bowman K Lee M Henze D Bousserez N Brix H Carbon monitoring system flux estimation and attribution: impact of ACOS-GOSAT XCO2 sampling on the inference of terrestrial biospheric sources and sinks Tellus B. 2014 12. Mitchard ET Saatchi SS Baccini A Asner GP Goetz SJ Harris NL Brown S Uncertainty in the spatial distribution of tropical forest biomass: a comparison of pan-tropical maps Carbon Balance Manag 2013 8 1 1 10.1186/1750-0680-8-10 23305341 13. Avitabile V Herold M Henry M Schmullius C Mapping biomass with remote sensing: a comparison of methods for the case study of Uganda Carbon Balance Manag 2011 6 1 1 10.1186/1750-0680-6-7 21362201 14. Huang W Swatantran A Johnson K Duncanson L Tang H Dunne JO Hurtt G Dubayah R Local discrepancies in continental scale biomass maps: a case study over forested and non-forested landscapes in Maryland, USA Carbon Balance Manag 2015 10 1 1 10.1186/s13021-014-0011-4 25632297 15. Riemann R Wilson BT Lister A Parks S An effective assessment protocol for continuous geospatial datasets of forest characteristics using USFS forest inventory and analysis (FIA) data Remote Sens Environ 2010 114 10 10.1016/j.rse.2010.05.010 16. Saatchi, S, Yifan Y, Fore A, Neumann M, Chapman B, Nguyen M, et al. CMS US forest biomass. 2005. (Received data through personal communication). 17. Kellndorfer J, Walker W, LaPoint L, Bishop J, Cormier T, Fiske G, Kirsch K, Westfall J. NACP Aboveground Biomass and Carbon Baseline Data (NBCD 2000), USA. 2000. Data Set. http://daac.ornl.gov. 18. Blackard JA Finco MV Helmer EH Holden GR Hoppus ML Jacobs DM Lister AJ Moisen GG Nelson MD Riemann R Mapping US forest biomass using nationwide forest inventory data and moderate resolution information Remote Sens Environ 2008 112 4 10.1016/j.rse.2007.08.021 19. Wilson BT Woodall CW Griffith DM Imputing forest carbon stock estimates from inventory plots to a nationally continuous coverage Carbon Balance Manag 2013 8 1 10.1186/1750-0680-8-1 23305341 20. Raciti SM Hutyra LR Newell JD Mapping Carbon Storage in Urban Trees with Multi-Source Remote Sensing Data: Relationships between Biomass, Land Use, and Demographics in Boston Neighborhoods Sci Total Environ. 2014 500 501 21. Krankina ON DellaSala DA Leonard J Yatskov M High-biomass forests of the Pacific Northwest: who manages them and how much is protected? Environ Manag 2014 54 1 10.1007/s00267-014-0283-1 22. Steininger MK Satellite estimation of tropical secondary forest above-ground biomass: data from Brazil and Bolivia Int J Remote Sens 2000 21 6 7 10.1080/014311600210119 23. Lu D The potential and challenge of remote sensing-based biomass estimation Int J Remote Sens 2006 27 7 24. Lu D Chen Q Wang G Liu L Li G Moran E A survey of remote sensing-based aboveground biomass estimation methods in forest ecosystems Int J Digit Earth. 2016 9 1 63 105 10.1080/17538947.2014.990526 25. Saatchi S Marlier M Chazdon RL Clark DB Russell AE Impact of spatial variability of tropical forest structure on radar estimation of aboveground biomass Remote Sens Environ 2011 115 11 2836 2849 10.1016/j.rse.2010.07.015 26. Heath LS,Hanson MH, Smith JE, Smith WB, Miles PD. Investigation into Calculating Tree Biomass and Carbon in the FIADB Using a Biomass Expansion Factor Approach. In 2008 Forest Inventory and Analysis (FIA) Symposium October 21–23, 2008; Park City, UT. Edited by: McWilliams W, Moisen G, Czaplewski R. USDA Forest Service, Rocky Mountain Research Station. Proc. RMRS-P-56CD. US Department of Agriculture, Forest Service, Rocky Mountain Research Station. Fort Collins, CO, 2009. 27. Zhang X Kondragunta S Estimating Forest Biomass in the USA Using Generalized Allometric Models and MODIS Land Products Geophys Res Lett 2006 33 9 28. http://www.fia.fs.fed.us/tools-data/. Accessed on Sep 01, 2012. 29. Willmott CJ Ackleson SG Davis RE Feddema JJ Klink KM Legates O’donnell J Rowe CM Statistics for the evaluation and comparison of models J Geophys Res Oceans 1985 90 C5 10.1029/JC090iC05p08995 30. Ji L Gallo K An agreement coefficient for image comparison Photogramm Eng Remote Sens 2006 72 7 10.14358/PERS.72.7.823 31. Neeti N Eastman JR Characterizing implications of two-dimensional space–time orientations for principal component analysis of geographic time series Int J Remote Sens 2015 36 1 10.1080/01431161.2014.994717 32. Cattell RB. The data box: its ordering of total resources in terms of possible relational systems. 67–128. RB Cattell. In: Handbook of Multivariate Experimental Psychology, 1966. 33. Team, R Core. R: A language and environment for statistical computing. Vienna, Austria. 2014. http://www.R-Project.Org. 34. Homer CG, Gallant, A. Partitioning the conterminous United States into mapping zones for Landsat TM land cover mapping. 2001, USGS Draft White Paper available at http://landcover.usgs.gov. 35. Ruefenacht B Finco MV Nelson MD Czaplewski R Helmer EH Blackard JA Holden GR Lister AJ Salajanu D Weyermann D Conterminous US and alaska forest type mapping using forest inventory and analysis data Photogramm Eng Remote Sens 2008 74 11 10.14358/PERS.74.11.1379 36. Pan Y Jing MC Birdsey R McCullough K He L Deng F Age structure and disturbance legacy of North American forests Biogeosciences. 2011 8 3 10.5194/bg-8-715-2011 37. Kennedy RE, Ohmann JL, Gregory MJ, Roberts HM, Yang Z, Cohen W, et al. In: Prep an empirical, integrated forest carbon monitoring system. http://geotrendr.ceoas.oregonstate.edu/data/ 38. Fernandez-Ordonez Y, Jesus S-R, Leblon B. Forest Inventory using Optical and Radar Remote Sensing. In: Advances in Geoscience and Remote Sensing. InTech. 2009. doi:10.5772/8330 39. Bayer T Winter R Schreier G Terrain influences in SAR backscatter and attempts to their correction IEEE Trans Geosci Remote Sens 1991 29 3 10.1109/36.79436 40. Cohen WB Maiersperger TK Gower ST Turner DP An improved strategy for regression of biophysical variables and landsat ETM+ data Remote Sens Environ 2003 84 4 10.1016/S0034-4257(02)00173-6

PMC005xxxxxx/PMC5002066.txt

==== Front J Med Internet Res J Med Internet Res JMIR Journal of Medical Internet Research 1439-4456 1438-8871 JMIR Publications Toronto, Canada v18i8e223 27520283 10.2196/jmir.5664 Original Paper Original Paper Web-Based Video-Coaching to Assist an Automated Computer-Tailored Physical Activity Intervention for Inactive Adults: A Randomized Controlled Trial Eysenbach Gunther Whitehead Lisa Irwin Brandon Bock Beth Torous John Alley Stephanie PhD http://orcid.org/0000-0001-9666-5071 1Physical Activity Research Group School of Human, Health and Social Sciences Central Queensland University Building 77 Bruce Highway Rockhampton, 4701 Australia 61 749232263 61 749232265 s.alley@cqu.edu.au Jennings Cally PhD 2http://orcid.org/0000-0001-9017-7077 Plotnikoff Ronald C PhD 3http://orcid.org/0000-0002-3763-8273 Vandelanotte Corneel PhD 1http://orcid.org/0000-0002-4445-8094 1 Physical Activity Research Group School of Human, Health and Social Sciences Central Queensland University Rockhampton Australia 2 Alberta Center for Active Living Faculty of Physical Education and Recreation University of Alberta Edmonton, AB Canada 3 Priority Research Centre for Physical Activity and Nutrition Faculty of Health and Medicine University of Newcastle Callaghan Australia Corresponding Author: Stephanie Alley s.alley@cqu.edu.au 8 2016 12 8 2016 18 8 e22329 2 2016 19 4 2016 1 6 2016 13 6 2016 ©Stephanie Alley, Cally Jennings, Ronald C Plotnikoff, Corneel Vandelanotte. Originally published in the Journal of Medical Internet Research (http://www.jmir.org), 12.08.2016. 2016 https://creativecommons.org/licenses/by/2.0/ This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included. Background Web-based physical activity interventions that apply computer tailoring have shown to improve engagement and behavioral outcomes but provide limited accountability and social support for participants. It is unknown how video calls with a behavioral expert in a Web-based intervention will be received and whether they improve the effectiveness of computer-tailored advice. Objective The purpose of this study was to determine the feasibility and effectiveness of brief video-based coaching in addition to fully automated computer-tailored advice in a Web-based physical activity intervention for inactive adults. Methods Participants were assigned to one of the three groups: (1) tailoring + video-coaching where participants received an 8-week computer-tailored Web-based physical activity intervention (“My Activity Coach”) including 4 10-minute coaching sessions with a behavioral expert using a Web-based video-calling program (eg, Skype; n=52); (2) tailoring-only where participants received the same intervention without the coaching sessions (n=54); and (3) a waitlist control group (n=45). Demographics were measured at baseline, intervention satisfaction at week 9, and physical activity at baseline, week 9, and 6 months by Web-based self-report surveys. Feasibility was analyzed by comparing intervention groups on retention, adherence, engagement, and satisfaction using t tests and chi-square tests. Effectiveness was assessed using linear mixed models to compare physical activity changes between groups. Results A total of 23 tailoring + video-coaching participants, 30 tailoring-only participants, and 30 control participants completed the postintervention survey (83/151, 55.0% retention). A low percentage of tailoring + video-coaching completers participated in the coaching calls (11/23, 48%). However, the majority of those who participated in the video calls were satisfied with them (5/8, 71%) and had improved intervention adherence (9/11, 82% completed 3 or 4 modules vs 18/42, 43%, P=.01) and engagement (110 minutes spent on the website vs 78 minutes, P=.02) compared with other participants. There were no overall retention, adherence, engagement, and satisfaction differences between tailoring + video-coaching and tailoring-only participants. At 9 weeks, physical activity increased from baseline to postintervention in all groups (tailoring + video-coaching: +150 minutes/week; tailoring only: +123 minutes/week; waitlist control: +34 minutes/week). The increase was significantly higher in the tailoring + video-coaching group compared with the control group (P=.01). No significant difference was found between intervention groups and no significant between-group differences were found for physical activity change at 6 months. Conclusions Only small improvements were observed when video-coaching was added to computer-tailored advice in a Web-based physical activity intervention. However, combined Web-based video-coaching and computer-tailored advice was effective in comparison with a control group. More research is needed to determine whether Web-based coaching is more effective than stand-alone computer-tailored advice. Trial Registration Australian New Zealand Clinical Trials Registry (ACTRN): 12614000339651; http://www.anzctr.org.au/TrialSearch.aspx?searchTxt=ACTRN12614000339651+&isBasic=True (Archived by WebCite at http://www.webcitation.org/6jTnOv0Ld) motor activity health promotion chronic disease e-counseling Internet ==== Body pmcIntroduction Physical activity improves physical and mental health and significantly lowers the risk of noncommunicable disease, including cardiovascular disease, diabetes, and cancer [1]. Australian guidelines recommend 150-300 minutes of moderate-intensity activity each week, over 5 days, to receive health benefits and reduce the risk of noncommunicable disease [2]. Despite this, less than 50% of Australians meet these recommendations [3]. As such, there is a need for effective and affordable physical activity interventions with a broad reach. Health behavior change interventions delivered via the Web have the potential to reach a large audience at low cost, due to the majority of Australians (92%) having access to the Web [4]. Furthermore, they are convenient for participants and enable the content to be delivered in a nonconfrontational manner [5-7]. Many studies have found Web-based physical activity interventions to be effective in the short term. For example, reviews of Web-based physical activity interventions have found more than half of controlled trials to report positive physical activity outcomes [8,9]. However, problems with low engagement and high dropout rates lead to small and short-term behavior changes [8,10]. Furthermore, many interventions struggle to reach those most in need of increasing their physical activity [8,10-12]. Personalized health advice through coaching sessions or computer-tailored feedback engages participants and improves the effectiveness of Web-based health behavior interventions [13-15]. Both Web-based coaching sessions and computer-tailored advice provide support similar to traditional face-to-face coaching sessions, at a lower cost with fewer geographical restrictions [13,16]. Coaching is defined as facilitating health behavior change through interactions between a health professional (coach) and a client [17]. Web-based coaching sessions are most similar to traditional face-to-face coaching as they provide personal interaction. Coaching in Web-based physical activity interventions improves participants’ perceptions of their social support, which is associated with greater levels of behavior change [18,19]. The Social Cognitive Theory stipulates that the acquisition of a new behavior is influenced by the individuals’ cognitive factors (eg, attitude), behavioral factors (eg, skills), and contextual factors including reinforcement, instructions, and social norms [20]. Including coaching in a Web-based intervention addresses the otherwise overlooked social contextual factors that play an important role in behavior change [20]. Advances in Internet technology and broadband capacity now allow the option of delivering coaching sessions via free Web-based video-calling programs, which allow participants to view the coach and engage in a verbal discussion. The popular video-calling programs Skype, Google Hangouts, and FaceTime are encrypted, which ensures privacy of participant information [21]. Video-coaching facilitates higher engagement, feelings of accountability, and social support, and reduces the risk of misunderstandings compared with emails and instant messaging [22,23]. Video-coaching has been found to be effective in producing changes in parenting behavior [24], to have a high feasibility for smoking cessation [25], and a high feasibility for supporting in-home rehabilitation in the elderly [26]. However, no studies have used video-coaching as part of a Web-based physical activity program. Despite the potential of Web-based video-coaching, its reliance on the time of a behavior change expert leads to higher implementation costs compared with fully automated computer-tailored advice. Computer tailoring can deliver personalized advice at a low cost by using a computer-based expert system to automatically deliver feedback to participants’ responses to a Web-based questionnaire [13]. Computer-tailored physical activity advice is preferred by participants, leads to greater attention [27], and improved health behavior outcomes compared with generic health advice [14]. Although the effectiveness of computer tailoring is well established and it has the benefit of providing personalized advice to large numbers at low cost, it is unknown whether it could be more effective with an element of human support. To our knowledge, no health behavior interventions have combined computer-tailored advice with Web-based video-coaching. This approach may improve intervention outcomes by utilizing the benefits of both methods. Providing computer-tailored advice can limit the time required from a video-coach, therefore limiting costs, as well as reducing reliance on the knowledge and expertise of the coach. A brief coaching session can add an element of social support, as well as further explanation, personalization, and interpretation of theory-based computer-tailored advice received by participants at an earlier time. It is unknown whether brief video-coaching sessions to reiterate computer-tailored physical activity advice are feasible in terms of retention, adherence, engagement, and satisfaction. It is also unknown whether they lead to improved physical activity and quality of life compared with stand-alone computer-tailored advice. Therefore, this study explores the feasibility and effectiveness of a brief Web-based coaching session in addition to computer-tailored advice for inactive adults. The first aim of the study was to determine the feasibility of brief video-coaching, when used to discuss previously received computer-tailored physical activity advice, in a stand-alone Web-based intervention for inactive adults. Feasibility was determined by adherence and satisfaction of the coaching sessions and comparing intervention retention, adherence, website engagement, and satisfaction of the tailoring + video-coaching and tailoring-only groups. The second aim was to test the effectiveness of the video-coaching sessions in terms of physical activity and quality of life outcomes. It was hypothesized that computer tailoring in combination with video-coaching would result in greater retention, adherence, engagement, and satisfaction with the intervention, compared with a computer-tailored–only group and a waitlist control group. It was also hypothesized that computer tailoring and video-coaching would result in greater improvements in quality of life and physical activity compared with a computer-tailored–only group and a waitlist control group. Methods Research Procedure A detailed account of the methods can be found in the protocol paper [28] and a consort eHealth checklist for the paper can be found here (Multimedia Appendix 1) [29]. The recruitment methods, participant eligibility, protocol, intervention description, measures, and data analysis are summarized below. Recruitment Print advertising and Web advertising were used to recruit participants from a number of Australian metropolitan and regional cities (Sydney, Melbourne, Perth, Brisbane, Rockhampton, Bundaberg, Mackay, and Townsville). Print advertising included newspaper advertisements and articles, posters and leaflets displayed in health clinics, and leaflets distributed to peoples’ homes. The Web advertising included links displayed on community websites and paid advertisements on Google and Facebook. Ethics approval was received from the Central Queensland University Human Research Ethics Committee (H13/04-044), before recruitment took place from March 2014 to January 2015. This study is registered with the Australian New Zealand Clinical Trials Registry (ACTRN12614000339651). Participants People were eligible to participate if they were English-speaking Australian adults (older than 18 years). Participants were excluded if they were pregnant, at risk of injury or ill health from their increasing physical activity (as assessed by the Physical Activity Readiness Questionnaire), or if they were already meeting the physical activity recommendations (as assessed via a single item asking if participants participated in 30 minutes of physical activity on most days). It is likely that the intervention attracted participants with a high Internet literacy. Protocol Information about the study, including the affiliation with Central Queensland University, was available on the landing page of the intervention website (Multimedia Appendices 2 and 3). To assess individuals’ eligibility, how they heard about the program, and collect contact details, a screening questionnaire was delivered through the intervention website. Eligible participants were randomly assigned based on a sequence (not concealed) of random numbers between 1 and 3 to one of three study arms: tailoring + video-coaching, tailoring-only, or waitlist control. This was done in blocks of 15 participants. SA generated the random allocation sequence and assigned participants to groups. Participants remained blinded to their condition until after completing all baseline measures. Participants began the intervention on the Monday following their recruitment. The consent form and then baseline questionnaire were administered through the intervention website for all groups. Upon completing the baseline questionnaire, the intervention groups received module 1 of their personalized advice, whereas the control group received nothing. The intervention “My Activity Coach” delivered 1 module of computer-tailored advice every 2 weeks over 8 weeks (4 modules in total). During the weeks where no new modules were received, participants in the tailoring + video-coaching group received a brief coaching session through a Web-based video-calling program (eg, Skype) to reiterate the advice received in their previous module. Participants in the tailoring-only group received an email reminding them of the tailored advice they received in the previous module to ensure both intervention groups received the same number of contacts. Participants in the waitlist control group were given the opportunity to participate in the intervention without coaching after they completed the final questionnaire. Questionnaires were administered through the intervention website immediately after the end of the intervention (week 9) and 6 months after the end of the intervention. Participants had to log in to complete each survey, which ensured they were only completed once. The week 9 questionnaires were collected from June 2014 to March 2015. It was not possible to blind researchers to participants’ group assignment after they had completed the baseline questionnaire. Participants were blinded to their group assignment only when completing their baseline questionnaire. The consent sheet explained the 2 interventions and therefore it is possible that participants worked out whether they were in the intervention of interest or comparator. Participants who completed all surveys went in the draw to win 1 of 30 pedometers, 6 Fitbits, and 3 heart rate monitors. Because many individuals began but failed to complete the screening questionnaire, we tested conducting the screening questionnaire by phone (after receiving an ethics amendment from the Central Queensland University’s Human Research Ethics Committee). This was done for 15 prospective participants and discontinued because of failure to increase screening completions. The Web-based booking system for coaching participants was changed halfway through the trial as the first booking system was discontinued. No other changes to the protocol were carried out during the trial. Intervention The 8-week “My Activity Coach” Web-based intervention delivered a new module of tailored advice to participants every 2 weeks [30]. Each module required participants to complete a brief Web-based questionnaire about their physical activity and psychosocial correlates of physical activity. Feedback was then provided based on their responses to the questionnaire (Multimedia Appendices 4 and 5). Participants received up to 4 reminder emails and a reminder phone call when they did not complete the survey required for each module. The tailored advice was based on behavior change theory (Theory of Planned Behavior [31]) and communication theory (Elaboration Likelihood Model [32]). Each module began with a graph including bars to represent participants’ current physical activity, their physical activity during the previous modules, as well as the minimum and optimal physical activity recommendations. Module 1, titled “Are you active enough,” explained the physical activity recommendations and health benefits of physical activity tailored to their body mass index (BMI), age, and level of physical activity. The module ended with a suggested goal (based on their current activity level) to work toward until the next module. Module 2, titled “Let’s set some goals,” provided participants with information on goal setting and action planning. Module 3, titled “Physical activity and your environment,” delivered tailored information on using participants’ social and physical environments to increase their physical activity. Module 4, titled “Staying active,” addressed relapse prevention. Participants also received tailored advice on their perceived benefits and barriers to being active and self-efficacy to become more active throughout the modules. The modules and intervention website were adapted from an earlier 2-module Web-based intervention with computer-tailored advice for inactive adults [33]. Focus groups were conducted to inform development of this prior intervention [34]. Updates were conducted by a website developer who also created the original intervention website. An action-planning tool became available to participants after module 2. The tool allowed participants to create an action plan for up to 4 activities (specifying where, when, for how long, and with whom they will be active over the following 2 weeks). Participants could print a calendar-based overview of their action plan (Multimedia Appendix 6). The coaching group’s 10- to 15-minute biweekly video-coaching sessions were conducted through a Web-based video-calling program of participants’ choice (eg, Skype, Google Hangouts, Yahoo Messenger, and FaceTime). During the session the activity coach commented on the tailored advice participants received in the module from the previous week, answered any questions participants had, and provided encouragement, support, and accountability. Measures Participants’ demographics including sex, age, BMI, household income (less than AUD $31,200, AUD $31,200-$77,999, more than AUD $78,000), education (less than secondary, secondary, further education), and employment (full time, part time or casual, and not in paid employment) were assessed in the baseline survey. Participants were also asked if they used a video-calling program (Skype, Google Hangouts, FaceTime, other, none). Completion of the coaching sessions, the length of the coaching sessions, and reasons for missed coaching sessions were recorded by the coach. Intervention retention and adherence were assessed by recording participants’ completion of the research surveys and intervention modules, respectively, and website engagement was measured through Google Analytics. Google Analytics recorded the number of website visits and time spent on the website for each participant. Intervention participants’ satisfaction with the intervention was assessed at the end of the intervention (week 9). Satisfaction with module questions (4 items), computer-tailored advice (14 items), website usability (13 items), overall program (5 items), and coaching (14 items) were all assessed. The items were specifically developed for this study, although based on previous research [35]. The items were on a 5-point Likert scale where participants were asked to rate their agreement (1=strongly agree to 5= strongly disagree) to statements about the intervention. All positively framed questions were reverse scored. For each category participants with a mean rating of 3.6 or higher (maximum = 5) were categorized as “satisfied.” Coaching participants were also asked if they completed a coaching session and if not, why not. All intervention participants were asked 4 open-ended questions about 3 topics: the advice, website, and overall program. Coaching participants who completed a coaching session were asked an additional 4 open-ended questions specifically relating to the coaching sessions. The 4 questions for each category (advice, website, program, and coaching) were (1) what did you like about the advice, website, program, or coaching; (2) what did you not like; (3) any recommendations for improvement; and (4) any other thoughts (Multimedia Appendix 7). Responses for all questions were thematically analyzed. The main outcome, weekly physical activity, was assessed at all time points via the Active Australia Survey, (AAS), which has a high percentage agreement with other physical activity measures (67%-75%) [36] and has a good test-retest reliability (kappa = .52) [37] including when self-administered [38]. Quality of life was measured at all 3 time points by the SF-12 version 2, which is valid [39] and reliable [40] including when self-administered on the Web [41]. Physical health and mental health component scores were calculated from the SF-12 version 2 following manual instructions [42]. Data Analysis Baseline demographics for participants in each trial arm (tailoring + video-coaching, tailoring-only, and control) are presented (Table 1). The demographics of completers versus dropouts, as well as coaching participants who did versus did not complete a coaching session, were compared using chi-square and t tests. To test feasibility of the coaching sessions, completion of the coaching sessions, the length of the coaching sessions, and reasons for missed coaching sessions were presented. Next, the 2 intervention groups, as well as coaching participants who did versus did not complete a coaching session, were compared on retention (dropouts vs completers) and adherence (completed 1-2 modules vs 3-4 modules) using a chi-square test, number of website visits and time spent on the website using t tests, and satisfaction scores (satisfied vs neutral or not satisfied) using a chi-square test. To test effectiveness, longitudinal data were analyzed using intention-to-treat principles. Physical activity, mental health score, and physical health score were each modeled using linear mixed models with time (baseline, week 9, and 6 months) as a repeated factor, fixed effects of time and group (control, tailoring-only, tailoring + video-coaching), and a time by group interaction (Table 2). Significance level was set to P<.05. Sample Size Sample size calculations demonstrated that a sample size of 300, or 100 in each study arm, was required to detect between-group differences in physical activity from baseline to postintervention using linear mixed models [43]. This calculation was based on the alpha level of ≤.05 (80% power) and a small effect size (0.43) and 25% attrition, which are common in similar interventions [10]. Results Flow of Participants Of the 239 randomly assigned participants, 154 completed the baseline questionnaire and at least one of the intervention modules. Of these, 84 participants completed the postintervention survey at week 9 (55% retention). A total of 59 participants completed the 6-month follow-up questionnaire (38% retention). There were no demographic differences between those who completed the week 9 survey and those who did not. The majority of participants were recruited through Facebook (63%), and small percentages were recruited through Google (8%), a newspaper article (6%), letterbox drops (5%), family or friend (5%), leaflets (5%), posters (4%), community websites (3%), and newspaper advertisements (1%). Figure 1 presents the flow of participants through the trial. Table 1 Baseline characteristics, physical activity, and quality of life by group assignment Variables Tailoring + video coaching Tailoring only Control Sex n=154, n (%) Male 16 (30) 14 (25) 7 (16) Female 37 (67) 42 (75) 38 (84) Employment n=151a, n (%) Full time 19 (37) 21 (38) 18 (40) Part time or casual 10 (19) 10 (19) 9 (20) Not in paid employment 23 (44) 23 (43) 18 (40) Education n=151a, n (%) Less than secondary 2 (4) 0 (0) 0 (0) Secondary 8 (15) 10 (19) 6 (13) Further education 42 (81) 44 (81) 39 (87) Income n=112b, n (%) More than AUD $78,000 21 (55) 14 (36) 21 (60) AUD $31,200-$77,999 11 (29) 17 (44) 6 (17) Less than AUD $31,199 6 (16) 8 (20) 8 (23) Uses Web-based video-calling n=151a, n (%) Yes 51 (63) 53 (68) 63 (79) No 30 (37) 25 (32) 17 (21) Age (years), n=154 55.26 (10.93) 52.18 (11.53) 55.18 (13.45) BMIc, n=150d, M(SD) 32.08 (7.43) 31.58 (7.43) 29.97 (6.75) Baseline total physical activity (minutes/week), n=151, M(SD) 189.52 (214.30) 152.87 (174.33) 160.44 (191.23) Week 9 total physical activity (minutes/week), n=83, M(SD) 387.83 (264.89) 315.17 (264.39) 211.00 (164.16) 6-month total physical activity (minutes/week), n=59, M(SD) 419.52 (77.109) 319.71 (164.77) 305.00 (315.86) Baseline mental health score, n=148, M(SD) 46.10 (1.30) 41.36 (12.27) 42.74 (1.78) Week 9 mental health score, n=82, M(SD) 48.97 (10.27) 30.03 (6.12) 45.19 (10.50) 6-month mental health score, n=59, M(SD) 48.16 (12.11) 43.90 (10.34) 44.94 (11.84) Baseline physical health score, n=146, M(SD) 46.90 (10.17) 51.92 (8.28) 48.62 (8.98) Week 9 physical health score, n=78, M(SD) 46.88 (12.05) 51.48 (6.95) 47.32 (9.37) 6-month physical health score, n=59, M(SD) 46.60 (10.22) 52.38 (5.47) 48.18 (10.46) aBaseline data (employment, education, video calling use and BMI) were lost for 3 participants. bA total of 39 respondents chose not to disclose their income. cBMI: body mass index. d BMI was lost for four participants. Figure 1 Participant flow through the intervention trial. *PARQ: Physical Activity Readiness Questionnaire. Sample Characteristics The majority of participants were female (117/154, 76.0%) and were on average 54 years of age. Just under half (64/151, 43%) were not in paid employment and the majority (125/151, 82.8%) had completed a higher education course. Less than half (n=60/151, 39.7%) of participants were physically active and on average participated in 168 minutes of physical activity. The average BMI was 31, which is in the obese range. The majority (n=106/151, 70.2%) used a video-calling program. Coaching Adherence and Satisfaction Coaching participants had low adherence to the coaching sessions. Just under half of coaching group completers (11/23, 48%) and less than a quarter of all coaching participants (11/52, 21%) completed 1 coaching session. There were no demographic differences between those who participated in a coaching session and those who did not. Of the coaching participants who did not do a coaching session, 8 wanted a second chance to do a session but did not book or show up the second time, 7 refused to do one, 4 were too busy, 7 had technical difficulties, 2 had injuries, and contact was lost with 13. Of those who participated in at least 1 coaching session, an average of 2.4 sessions were completed, the average coaching session length was 10.4 minutes, and 15% of the coaching calls were interrupted with technical difficulties. A total of 8 coaching participants (8/52, 15%) completed the coaching satisfaction questions. The majority were satisfied with the coaching sessions (n=5/8, 71.4%). In response to open-ended coaching questions, participants said the sessions held them accountable (n=3), appreciated the support (n=1), appreciated the information (n=1), liked the structure (n=1), and liked talking about exercise that suits them (n=1). However, some had technical problems (n=2) and would have preferred to use a phone (n=1). Intervention Adherence and Retention Retention did not differ between intervention groups (χ21= 4.7, P=.11). Participants who completed at least 1 coaching session had a higher percentage of week 9 survey completers (n=8/11, 73%), compared with other intervention participants (n=50/95, 53%), but this difference was not significant (χ21= 1.6, P=.21). Just under half (n=50/106, 47%) of participants completed at least 3 of the 4 intervention modules. Intervention adherence was similar for the tailoring + video-coaching group and the tailoring-only group (χ21= 2.1, P=.15); however, significantly more participants who participated in the coaching sessions completed at least 3 of the 4 intervention modules (n=9/11, 82%) compared with other intervention participants (n=41/95, 43%; χ21= 6.0, P=.01). Website Engagement The average website visits and minutes spent on the website for the intervention groups were 7.53 (SD 7.14) and 87.07 (SD 77.33) minutes, respectively. The average number of website visits was similar for the tailoring + video-coaching group and the tailoring-only group (t1,103= 0.05, P=.96). Average minutes spent on the website was higher for the tailoring + video-coaching group (mean 99.58 minutes, SD 95.71) than the tailoring-only group (mean 75.25 minutes, SD 52.90), but this was not significantly different (t1103= 1.60, P=.11). Participants who completed the coaching sessions spent a significantly longer time (t1103= 2.73, P=.02) on the intervention website (mean 174.64 minutes, SD 110.11) compared with other intervention participants (mean 77.84 minutes, SD 67.48). Participants who completed the coaching sessions also visited the website more frequently (mean 10.20, SD 3.85) compared with other intervention participants (mean 7.25, SD= 7.36), but this difference was not significant (t1103= 1.24, P=.22). Satisfaction More than two-thirds of the participants were satisfied with the overall program (n=36/53, 68%), while the majority of participants were satisfied with the website usability (n=40/53, 77%), the computer-tailored advice (n=38/53, 76%), and the module questions (n=48/53, 91%). There was no difference between intervention groups on program satisfaction scores (χ21= 0.14, P=.71). A higher percentage of those who participated in the coaching sessions (n=7/8, 88%) were satisfied with the program compared with other participants (n=29/45, 64%); however, this difference was not statistically significant (χ2= 1.66, P=.20). In response to open-ended questions on the overall program, participants mentioned that they liked the convenience (n=4), ease of use (n=4), the information (n=5), emails (n=2), found it motivating (n=5), and liked the accountability (n=2). Participants also mentioned that they would like more contact with a real person (n=7) and thought there were too many questions (n=2). In response to the questions on the advice, participants mentioned that it was easy to understand (n=12), was concise (n=4), laid out well (n=4), was nonjudgmental (n=2), and liked the personalization (n=2). However, some participants thought it was not personalized enough (n=11), did not like the Web-based format (n=4), and learned nothing new (n=2). In response to open-ended questions on the website, some participants mentioned that the website was easy to use (n=14), whereas others thought it was hard to use (n=5). Lastly, some participants thought the website could use more visuals and interesting links (n=6). Physical Activity Physical activity (minutes/week) improved from baseline to postintervention (week 9) and from baseline to follow-up (6 months) in all groups (Table 2,Figure 2). According to the linear mixed models analysis, the increase in physical activity from baseline to postintervention in the tailoring + video-coaching group in comparison with the control group was significant (Table 2). No significant difference was found between the intervention groups (Table 2). No significant differences were found between groups on physical activity changes from baseline to follow-up at 6 months. Figure 2 Mean physical activity at baseline, week 9 (post intervention), and 6 months (follow-up), unadjusted. Baseline n=151, postintervention n=83, and follow-up n=59. Note: 95% confidence intervals presented for the coaching and control groups only. Quality of Life Physical health scores remained relatively constant across each time point (Table 2). Mental health scores remained relatively constant in the control and tailoring + video-coaching group; however, mental health scores dropped in the tailoring-only participants at postintervention (Table 2). As such, the linear mixed models analysis indicated a significant difference in mental health between both intervention groups from baseline to postintervention (Table 2). Table 2 Physical activity, mental health, and physical health changes by group: results of intention-to-treat analysis using linear mixed models Outcome variables Baseline to postintervention Baseline to follow-up Estimatea P value Cohen’s d Estimatea P value Cohen’s d Physical activity (n=151) Tailoring + video-coaching versus control 140.94 (−254.01 to −27.87) .01 0.55 66.16 (−244.55 to 112.24) .46 0.19 Tailoring + video-coaching versus tailoring-only 35.39 (−148.50 to 77.71) .54 0.11 −25.16 (−211.74 to 161.43) .79 0.18 Mental health (n=148) Tailoring + video-coaching versus control 0.04 (−5.87 to 5.94) .99 0.22 0.36 (−5.65 to6.38) .90 0.22 Tailoring + video-coaching versus tailoring-only −13.18 (−19.10 to −7.26) .00 0.91 1.43 (−4.91 to 7.77) .65 0.34 Physical health (n=146) Tailoring + video-coaching versus control −2.97 (−7.16 to 1.22) .16 0.41 −2.07 (−6.34 to 2.20) .34 0.43 Tailoring + video-coaching versus tailoring-only −2.58 (−7.00 to 1.83) .25 0.33 0.16 (−4.32 to 4.65) .94 0.03 aEstimate of tailoring and video-coaching group’s change in the dependent variable in comparison with control and tailoring-only. Higher scores represent greater improvements than the comparison group. Discussion Feasibility The first aim of this study was to determine the feasibility of Web-based video-coaching sessions. The low participation and high satisfaction with the coaching sessions suggests that the majority of people participating in Web-based physical activity interventions are reluctant to talk to a coach using a video-calling program, but those who do find it worthwhile. The low participation in the coaching sessions could be explained by the high percentage of coaching participants (37%) who do not use video-calling software. This may be due to the older age of the participants who are less comfortable with technology [4]. Phone calls may have also been received better in this sample. The lower intervention adherence in the video-coaching + tailoring group also suggests that increasing participant burden through an additional video-coaching component may reduce feasibility. The high satisfaction of the video-coaching sessions adds to past research findings of high satisfaction of coaching sessions conducted over the phone [44]. An important reason for the high satisfaction could be the accountability provided through the coaching sessions, as this was the most frequently given positive feedback for the sessions. The two most common negative comments about the overall program were the lack of personal contact and that it was not personalized enough. Therefore, it is not surprising that those participants who participated in coaching sessions reported higher levels of program satisfaction. Participant satisfaction leads to improved engagement, which is important for intervention effectiveness [45]. In support of this, the participants who completed the coaching sessions were not only more satisfied with the program, but had significantly higher retention, higher adherence, and spent a significantly longer time on the website. This is in line with past research, demonstrating that personal contact in a Web-based intervention improves engagement [46,47]. However, it is possible that the coaching sessions are only effective in some people, and that the coaching sessions in this study may have retained a subsample of participants who were frequent Web users and therefore familiar with video-calling software or were more motivated to begin with. Due to the low participation in the coaching sessions, there were no significant differences in overall retention, adherence, engagement, and satisfaction between the 2 intervention groups. Web-based interventions with computer-tailored advice and coaching sessions may increase retention, adherence, engagement, and satisfaction, but only if they can convince participants to participate in the coaching sessions. Effectiveness The second aim of this study was to assess the effectiveness of the coaching sessions, by comparing physical activity and quality of life changes of the tailoring + video-coaching group with the tailoring-only and control groups. After the 8-week intervention period there was a significant treatment effect of the tailoring + video-coaching physical activity intervention on physical activity compared with no intervention. The improvement in physical activity compared with the control group (116 minutes per week, unadjusted) resulted in a moderate effect size and is considered clinically significant because of the large health effects seen from doing even small amounts of physical activity [48]. This finding is in line with findings that activity counseling over the phone [49], via email [50], and computer-tailored advice [13] improve participants’ physical activity in comparison with a control group. Few studies have specifically tested the effectiveness of counseling through video calls in physical activity interventions. Pilutti et al [51] found that video-coaching to promote physical activity in patients with multiple sclerosis was effective. However, the coaching group’s significant improvement in physical activity compared with the control group, in our study, could be due to the computer-tailored advice. The tailoring + video-coaching group participants improved their physical activity 27 minutes per week (unadjusted) more in comparison with the tailoring-only group; however, no significant between-group differences were found. The availability of human support may have improved the overall physical activity of the tailoring + video-coaching group. Participants who needed to discuss their computer-tailored advice were able to. Limited studies have compared the effectiveness of coaching in addition to computer-tailored advice. Van Hoye et al [52] compared physical activity self-monitoring with and without additional face-to-face coaching sessions. They found that the coaching group had significantly greater physical activity improvements than the self-monitoring only group. An earlier study tested the effectiveness of email coaching in addition to a basic Web-based weight loss intervention [53]. The email-coaching group had significantly greater weight loss outcomes than the Web-based intervention only group. However, the effectiveness of coaching found in these studies may be due to the minimal nature of the comparison interventions (generic physical activity information + self-monitoring). The physical activity advice given in our study was highly tailored to participants’ physical activity behavior, demographics, and psychosocial correlates of physical activity. This feedback might be enough to optimize physical activity outcomes (as demonstrated by a 172 minutes/week increase in physical activity in this group). The lack of significant differences between the intervention groups could also be due to low adherence to the coaching sessions, which may have reduced the effectiveness of the intervention in the tailoring + video-coaching group. The detailed computer-tailored advice may have discouraged participants to adhere to their coaching sessions as they were satisfied with the computer-tailored advice. Core intervention content may need to be delivered in the coaching sessions to promote higher adherence. Therefore, more research is needed to determine whether Web-based coaching is more effective than stand-alone computer-tailored advice. The physical activity levels in the tailoring + video-coaching and tailoring-only groups were maintained at 6 months. There were, however, no significant between-group differences in physical activity changes from baseline to 6 months. This was due to the control group participants increasing their activity from 9 weeks to 6 months after the intervention. The absence of between-group physical activity changes at 6 months after the intervention is not uncommon in physical activity interventions; however, it is usually due to the intervention group’s decline in physical activity rather than a physical activity increase in the control group [54]. There were no differences the physical health component of quality of life over time for any of the groups. This may be due to the sample not being large enough to detect subtle improvements or the overall high level of physical activity and physical health in the sample at baseline. Although previous studies have established a positive association between physical activity and quality of life [55], the effect of increased physical activity on quality of life is mainly seen in clinical samples that have a lower quality of life at baseline [56]. The significant reduction in quality of life in relation to mental health in the tailoring-only group from baseline to postintervention in comparison was unexpected, as this group’s participants did increase their physical activity at a clinically significant level and previous studies have associated this with an increase in quality of life [55]. Limitations Limitations of the research include self-reported physical activity, which may be subject to social desirability bias. Participants and researchers were not blinded to group assignment, which may have biased results. The sample was predominantly female, white, and educated. Therefore, the results may not be generalizable to males, other cultures, and low socioeconomic groups. A high percentage of the sample was not in paid employment when compared with the Australian population data (43% vs 33%) [57]. This may be due to the high percentage of females and the older age of the participants. It could also be due to the people not in paid employment having more time to participate. Therefore, the results may not be generalizable to people in paid employment. Furthermore, 40% of the participants were physically active at baseline despite the target group being inactive adults. This is not uncommon in physical activity studies [58], although results may not be generalizable to physically inactive Australians who are most in need of increasing their activity. The sample may have a high level of computer and Internet literacy as the majority of participants were reached through Facebook. Therefore, the findings may not be fully generalizable to people with a lower computer and Internet literacy. The additional phone calls the coaching participants received encouraging them to complete the coaching calls may have affected their physical activity levels. However, this is expected to be minimal due to the phone discussion being focused on scheduling the coaching call rather than their physical activity. Objective measurement of physical activity (eg, accelerometry) is needed to confirm the findings of this study. Although we were able to detect a difference in physical activity changes between the tailoring + video-coaching and tailoring-only groups, the analysis was underpowered to detect other differences between the groups as the required sample size of 100 per group was not met. The low retention is also a significant limitation. Low retention is common in Web-based programs, potentially due to the minimal face-to-face contact through either recruitment or the intervention. Similar retention rates have been observed in other Web-based health behavior interventions [59,60]. Retention is likely to be even lower outside of a randomized controlled trial setting that included reminder calls from researchers. Conclusions Combined Web-based video-coaching and computer-tailored advice was effective in comparison with a control group; however, only small nonsignificant improvements were seen when video-coaching was included in addition to computer-tailored advice in a Web-based physical activity intervention. Only a small percentage of participants adhered to Web-based video-coaching sessions, but those who participated were highly satisfied and more engaged in the intervention. Further research should investigate how adherence to Web-based coaching sessions can be improved. Project and personal support for SA was provided by the National Health and Medical Research Council of Australia (ID 1055804). RCP is supported by a Salary Award from the National Health and Medical Research Council of Australia. CV (ID 100427) is supported by a Future Leader Fellowship from the National Heart Foundation of Australia. Multimedia Appendix 1 CONSORT-EHEALTH checklist V1.6.2 [29]. Multimedia Appendix 2 Intervention landing page. Multimedia Appendix 3 Participant information sheet. Multimedia Appendix 4 Tailored advice with graph. Multimedia Appendix 5 Text-only tailored advice. Multimedia Appendix 6 Action plan output. Multimedia Appendix 7 Satisfaction survey. Abbreviations BMI body mass index Conflicts of Interest: None declared. ==== Refs 1 Lee I Shiroma EJ Lobelo F Puska P Blair SN Katzmarzyk PT Lancet Physical Activity Series Working Group Effect of physical inactivity on major non-communicable diseases worldwide: an analysis of burden of disease and life expectancy Lancet 2012 07 21 380 9838 219 29 10.1016/S0140-6736(12)61031-9 22818936 S0140-6736(12)61031-9 22818936 2 Department of Health, Australian Government 2014 2016-08-02 Canberra Australia’s Physical Activity & Sedentary Behaviour Guidelines for Adults (18-64 years) http://www.health.gov.au/internet/main/publishing.nsf/content/health-pubhlth-strateg-phys-act-guidelines 3 Australian Bureau of Statistics 2013 2016-08-02 Canberra ABS Australian Health Survey: First Results, 2011-12 http://www.abs.gov.au/ausstats/abs@.nsf/Lookup/4364.0.55.001main+features12011-12 4 Australian Communications and Media Authority 2015 2016-08-02 Canberra ACMA Australians embrace a digital life http://www.acma.gov.au/theACMA/engage-blogs/engage-blogs/researchacma/Australians-embrace-a-digital-life 5 Christopherson KM The positive and negative implications of anonymity in Internet social interactions: “On the Internet, Nobody Knows You're a Dog” Computers in Human Behavior 2007 11 23 6 3038 3056 10.1016/j.chb.2006.09.001 6 McConnon A Kirk SF Cockroft JE Harvey EL Greenwood DC Thomas JD Ransley JK Bojke L The Internet for weight control in an obese sample: results of a randomised controlled trial BMC Health Serv Res 2007 7 206 10.1186/1472-6963-7-206 18093289 1472-6963-7-206 18093289 7 Tate DF Finkelstein EA Khavjou O Gustafson A Cost effectiveness of internet interventions: review and recommendations Ann Behav Med 2009 08 38 1 40 5 10.1007/s12160-009-9131-6 19834778 19834778 8 Vandelanotte C Spathonis KM Eakin EG Owen N Website-delivered physical activity interventions a review of the literature Am J Prev Med 2007 07 33 1 54 64 10.1016/j.amepre.2007.02.041 17572313 S0749-3797(07)00164-X 17572313 9 Joseph RP Durant NH Benitez TJ Pekmezi DW Internet-Based Physical Activity Interventions Am J Lifestyle Med 2014 8 1 42 68 10.1177/1559827613498059 25045343 10 Davies CA Spence JC Vandelanotte C Caperchione CM Mummery WK Meta-analysis of internet-delivered interventions to increase physical activity levels Int J Behav Nutr Phys Act 2012 9 1 52 10.1186/1479-5868-9-52 22546283 1479-5868-9-52 22546283 11 Lewis B Williams D Dunsiger S Sciamanna C Whiteley J Napolitano M Bock B Jakicic J Getz M Marcus B User attitudes towards physical activity websites in a randomized controlled trial Prev Med 2008 11 47 5 508 13 10.1016/j.ypmed.2008.07.020 18773915 S0091-7435(08)00400-3 18773915 12 Short CE Vandelanotte C Dixon MW Rosenkranz R Caperchione C Hooker C Karunanithi M Kolt GS Maeder A Ding H Taylor P Duncan MJ Examining participant engagement in an information technology-based physical activity and nutrition intervention for men: the manup randomized controlled trial JMIR Res Protoc 2014 3 1 e2 10.2196/resprot.2776 24389361 v3i1e2 24389361 13 Lustria ML Cortese J Noar SM Glueckauf RL Computer-tailored health interventions delivered over the Web: review and analysis of key components Patient Educ Couns 2009 02 74 2 156 73 10.1016/j.pec.2008.08.023 18947966 S0738-3991(08)00469-2 18947966 14 Brouwer W Kroeze W Crutzen R de Nooijer NJ de Vries N Brug J Oenema A Which intervention characteristics are related to more exposure to internet-delivered healthy lifestyle promotion interventions? A systematic review J Med Internet Res 2011 13 1 e2 10.2196/jmir.1639 21212045 v13i1e2 21212045 15 Broekhuizen K Kroeze W van Poppel M Oenema A Brug J A systematic review of randomized controlled trials on the effectiveness of computer-tailored physical activity and dietary behavior promotion programs: an update Ann Behav Med 2012 10 44 2 259 86 10.1007/s12160-012-9384-3 22767052 22767052 16 Noar SM Benac CN Harris MS Does tailoring matter? Meta-analytic review of tailored print health behavior change interventions Psychol Bull 2007 07 133 4 673 93 10.1037/0033-2909.133.4.673 17592961 2007-09203-006 17592961 17 Olsen JM Health coaching: a concept analysis Nurs Forum 2014 07 49 1 18 29 10.1111/nuf.12042 24456550 24456550 18 Barrera M Toobert DJ Angell KL Glasgow RE Mackinnon DP Social support and social-ecological resources as mediators of lifestyle intervention effects for type 2 diabetes J Health Psychol 2006 05 11 3 483 95 10.1177/1359105306063321 16774900 11/3/483 16774900 19 Winett RA Anderson ES Wojcik JR Winett SG Bowden T Guide to health: nutrition and physical activity outcomes of a group-randomized trial of an Internet-based intervention in churches Ann Behav Med 2007 06 33 3 251 61 10.1080/08836610701358045 17600452 17600452 20 Bandura A Social foundations of thought and action: a social cognitive theory 1986 Englewood Cliffs, NJ Prentice-Hall 21 Skype 2016-08-02 Does Skype use encryption? Security and blocked accounts https://support.skype.com/en/faq/fa31/does-skype-use-encryption 22 Abbott JM Klein B Ciechomski L Best Practices in Online Therapy Journal of Technology in Human Services 2008 07 03 26 2-4 360 375 10.1080/15228830802097257 23 Fantus S Mishna F The Ethical and Clinical Implications of Utilizing Cybercommunication in Face-to-Face Therapy Smith College Studies in Social Work 2013 10 83 4 466 480 10.1080/00377317.2013.833049 24 Wade SL Oberjohn K Burkhardt A Greenberg I Feasibility and preliminary efficacy of a web-based parenting skills program for young children with traumatic brain injury J Head Trauma Rehabil 2009 24 4 239 47 10.1097/HTR.0b013e3181ad6680 19625863 00001199-200907000-00003 19625863 25 Whittaker R Dorey E Bramley D Bullen C Denny S Elley CR Maddison R McRobbie H Parag V Rodgers A Salmon P A theory-based video messaging mobile phone intervention for smoking cessation: randomized controlled trial J Med Internet Res 2011 13 1 e10 10.2196/jmir.1553 21371991 v13i1e10 21371991 26 Taylor A Morris G Pech J Rechter S Carati C Kidd MR Home Telehealth Video Conferencing: Perceptions and Performance JMIR Mhealth Uhealth 2015 3 3 e90 10.2196/mhealth.4666 26381104 v3i3e90 26381104 27 Brug J Oenema A Campbell M Past, present, and future of computer-tailored nutrition education Am J Clin Nutr 2003 04 77 4 Suppl 1028S 1034S 12663313 12663313 28 Alley S Jennings C Plotnikoff RC Vandelanotte C My Activity Coach - using video-coaching to assist a web-based computer-tailored physical activity intervention: a randomised controlled trial protocol BMC Public Health 2014 14 738 10.1186/1471-2458-14-738 25047900 1471-2458-14-738 25047900 29 Eysenbach G Consort-EHEALTHGroup CONSORT-EHEALTH: improving and standardizing evaluation reports of Web-based and mobile health interventions J Med Internet Res 2011 13 4 e126 10.2196/jmir.1923 22209829 v13i4e126 22209829 30 Alley S My Activity Coach 2014 2016-02-17 http://www.myactivitycoach.org.au/eng/pages/23 31 Ajzen I Attitudes, personality, and behaviour 1988 Illinois The Dorsey Press 32 Cacioppo J Petty RE The elaboration likelihood model of persuasion Advanced Consumer Research 1984 19 123 33 Vandelanotte C Duncan MJ Plotnikoff RC Mummery WK Do participants' preferences for mode of delivery (text, video, or both) influence the effectiveness of a Web-based physical activity intervention? J Med Internet Res 2012 14 1 e37 10.2196/jmir.1998 22377834 v14i1e37 22377834 34 Vandelanotte C Mummery WK Qualitative and quantitative research into the development and feasibility of a video-tailored physical activity intervention Int J Behav Nutr Phys Act 2011 8 1 70 10.1186/1479-5868-8-70 21718546 1479-5868-8-70 21718546 35 Yip MP Chang AM Chan J MacKenzie AE Development of the Telemedicine Satisfaction Questionnaire to evaluate patient satisfaction with telemedicine: a preliminary study J Telemed Telecare 2003 9 1 46 50 12641893 12641893 36 Brown W Bauman A Chey T Trost S Mummery K Comparison of surveys used to measure physical activity Aust N Z J Public Health 2004 04 28 2 128 34 15233351 15233351 37 Brown WJ Trost SG Bauman A Mummery K Owen N Test-retest reliability of four physical activity measures used in population surveys J Sci Med Sport 2004 06 7 2 205 15 15362316 15362316 38 Brown WJ Burton NW Marshall AL Miller YD Reliability and validity of a modified self-administered version of the Active Australia physical activity survey in a sample of mid-age women Aust N Z J Public Health 2008 12 32 6 535 41 10.1111/j.1753-6405.2008.00305.x 19076744 AZPH305 19076744 39 Burdine JN Felix MR Abel AL Wiltraut CJ Musselman YJ The SF-12 as a population health measure: an exploratory examination of potential for application Health Serv Res 2000 10 35 4 885 904 11055454 11055454 40 Ware J Kosinski M Keller SD A 12-Item Short-Form Health Survey: construction of scales and preliminary tests of reliability and validity Med Care 1996 03 34 3 220 33 8628042 8628042 41 Whitehead L Methodological issues in Internet-mediated research: a randomized comparison of internet versus mailed questionnaires J Med Internet Res 2011 13 4 e109 10.2196/jmir.1593 22155721 v13i4e109 22155721 42 Saris-Baglama RN Dewey CJ Chisholm GB Plumb E King J Kosinski M Bjorner JB Ware JE Qualitymetric 2010 2016-08-02 Lincon, RI QualityMetric Incorporated Quality metric health outcomes scoring software 4.0. Installation guide http://www.qualitymetric.com/download/InstallationGuide_ScoringSoftwareV4.pdf 43 Lu K Luo X Chen P Sample size estimation for repeated measures analysis in randomized clinical trials with missing data Int J Biostat 2008 4 1 Article 9 22462117 22462117 44 Nes Andréa AG van Dulmen SB Eide E Finset A Kristjánsdóttir OB Steen IS Eide H The development and feasibility of a web-based intervention with diaries and situational feedback via smartphone to support self-management in patients with diabetes type 2 Diabetes Res Clin Pract 2012 09 97 3 385 93 10.1016/j.diabres.2012.04.019 22578890 S0168-8227(12)00153-2 22578890 45 Mair F Whitten P Systematic review of studies of patient satisfaction with telemedicine BMJ 2000 06 3 320 7248 1517 20 10834899 10834899 46 Spittaels H De BI Implementation of an online tailored physical activity intervention for adults in Belgium Health Promot Int 2006 12 21 4 311 9 10.1093/heapro/dal030 16963783 dal030 16963783 47 Meyer C Ulbricht S Gross B Kästel L Wittrien S Klein G Skoeries BA Rumpf H John U Adoption, reach and effectiveness of computer-based, practitioner delivered and combined smoking interventions in general medical practices: a three-arm cluster randomized trial Drug Alcohol Depend 2012 02 1 121 1-2 124 32 10.1016/j.drugalcdep.2011.08.019 21924563 S0376-8716(11)00375-9 21924563 48 Warburton DE Nicol CW Bredin SS Health benefits of physical activity: the evidence CMAJ 2006 03 14 174 6 801 9 10.1503/cmaj.051351 16534088 174/6/801 16534088 49 Hekler EB Buman MP Otten J Castro CM Grieco L Marcus B Friedman RH Napolitano MA King AC Determining who responds better to a computer- vs. human-delivered physical activity intervention: results from the community health advice by telephone (CHAT) trial Int J Behav Nutr Phys Act 2013 10 109 10.1186/1479-5868-10-109 24053756 1479-5868-10-109 24053756 50 Tate DF Jackvony EH Wing RR A randomized trial comparing human e-mail counseling, computer-automated tailored counseling, and no counseling in an Internet weight loss program Arch Intern Med 2006 166 15 1620 5 10.1001/archinte.166.15.1620 16908795 166/15/1620 16908795 51 Pilutti LA Dlugonski D Sandroff BM Klaren RE Motl RW Internet-delivered lifestyle physical activity intervention improves body composition in multiple sclerosis: preliminary evidence from a randomized controlled trial Arch Phys Med Rehabil 2014 07 95 7 1283 8 10.1016/j.apmr.2014.03.015 24699237 S0003-9993(14)00222-6 24699237 52 Van HK Boen F Lefevre J The Impact of Different Degrees of Feedback on Physical Activity Levels: A 4-Week Intervention Study Int J Environ Res Public Health 2015 06 12 6 6561 81 10.3390/ijerph120606561 26067990 ijerph120606561 26067990 53 Tate DF Jackvony EH Wing RR Effects of Internet behavioral counseling on weight loss in adults at risk for type 2 diabetes: a randomized trial JAMA 2003 04 9 289 14 1833 6 10.1001/jama.289.14.1833 12684363 289/14/1833 12684363 54 Fjeldsoe B Neuhaus M Winkler E Eakin E Systematic review of maintenance of behavior change following physical activity and dietary interventions Health Psychol 2011 01 30 1 99 109 10.1037/a0021974 21299298 2011-02060-010 21299298 55 Bize R Johnson JA Plotnikoff RC Physical activity level and health-related quality of life in the general adult population: a systematic review Prev Med 2007 12 45 6 401 15 10.1016/j.ypmed.2007.07.017 17707498 S0091-7435(07)00302-7 17707498 56 Fong DY Ho J Hui BP Lee AM Macfarlane DJ Leung SS Cerin E Chan WY Leung IP Lam SH Taylor AJ Cheng K Physical activity for cancer survivors: meta-analysis of randomised controlled trials BMJ 2012 344 e70 22294757 22294757 57 Australian Bureau of Statistics 2016-08-02 Canberra ABS Education and Work http://www.abs.gov.au/ausstats/abs@.nsf/mf/6227.0 58 Vandelanotte C Stanton R Rebar AL Van Itallie AK Caperchione CM Duncan MJ Savage TN Rosenkranz RR Kolt GS Physical activity screening to recruit inactive randomized controlled trial participants: how much is too much? Trials 2015 16 446 10.1186/s13063-015-0976-7 26449210 10.1186/s13063-015-0976-7 26449210 59 Carr LJ Dunsiger SI Lewis B Ciccolo JT Hartman S Bock B Dominick G Marcus BH Randomized controlled trial testing an internet physical activity intervention for sedentary adults Health Psychol 2013 03 32 3 328 36 10.1037/a0028962 22823069 2012-19415-001 22823069 60 Skår S Sniehotta FF Molloy GJ Prestwich A Araújo-Soares V Do brief online planning interventions increase physical activity amongst university students? A randomised controlled trial Psychol Health 2011 04 26 4 399 417 10.1080/08870440903456877 20830646 926714239 20830646

PMC005xxxxxx/PMC5002067.txt

==== Front Biophys JBiophys. JBiophysical Journal0006-34951542-0086The Biophysical Society S0006-3495(16)30474-X10.1016/j.bpj.2016.06.028ProteinsModeling Functional Motions of Biological Systems by Customized Natural Moves Demharter Samuel 1Knapp Bernhard 2Deane Charlotte M. 2Minary Peter peter.minary@cs.ox.ac.uk1∗1 Department of Computer Science, University of Oxford, Oxford, UK2 Department of Statistics, University of Oxford, Oxford, UK∗ Corresponding author peter.minary@cs.ox.ac.uk23 8 2016 23 8 2016 111 4 710 721 1 2 2016 22 6 2016 © 2016 Biophysical Society.2016Biophysical SocietyThis is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).Simulating the functional motions of biomolecular systems requires large computational resources. We introduce a computationally inexpensive protocol for the systematic testing of hypotheses regarding the dynamic behavior of proteins and nucleic acids. The protocol is based on natural move Monte Carlo, a highly efficient conformational sampling method with built-in customization capabilities that allows researchers to design and perform a large number of simulations to investigate functional motions in biological systems. We demonstrate the use of this protocol on both a protein and a DNA case study. Firstly, we investigate the plasticity of a class II major histocompatibility complex in the absence of a bound peptide. Secondly, we study the effects of the epigenetic mark 5-hydroxymethyl on cytosine on the structure of the Dickerson-Drew dodecamer. We show how our customized natural moves protocol can be used to investigate causal relationships of functional motions in biological systems. Editor: Tamar Schlick. ==== Body Introduction Functional motions in biomolecules are central to many biological processes (1). Molecular simulations are often used as a tool to investigate these dynamics and interpret (2, 3) and/or refine (4) experimental data or inspire new experiments (5). Large improvements in computational resources and algorithms have been made since the first molecular simulation of a protein in 1977 (6, 7). Recent milestones include the 50-ns molecular dynamics (MD) simulation of the full satellite tobacco mosaic virus with 1,000,000 particles (8), the Folding@Home project that used >400,000 personal computers to study challenging problems such as protein folding (9), and a study that presented millisecond simulations to study the folding pathways of small fast-folding proteins (10). Despite these advances, the high dimensionality and complex energy surfaces still pose a challenge for simulations of large biomolecules (11, 12). In an effort to address these limitations there have been promising developments in dimensionality reducing methods that exploit the inherent modularity and collective motions in biomolecules (13, 14). For example, essential dynamics coarse-graining (ED-CG) identifies sites that reflect the essential dynamics of an atomistic molecular dynamics trajectory (15). Other methods based on elastic network models, principal component analysis, and normal mode analysis have also been successfully used to study functional motions in biomolecules (16, 17, 18). While these methods are not as physically accurate as MD simulations, their increased sampling efficiency makes them a valuable tool to generate new hypotheses that can be tested by experiments. One of the main challenges of these methods, however, is finding a set of degrees of freedom (DOFs) that describe the system accurately enough to draw biologically relevant conclusions (14). Thus, it is of value to have computationally cheap methods that allow for the easy manipulation of DOFs to test different hypotheses about the functional motions of biomolecules in silico. One method designed to address the dimensionality challenge is natural-move Monte Carlo (NMMC) (19, 20). NMMC is a conformational sampling method that exploits the modular nature of biomolecules to accelerate the exploration of structural landscapes to identify functionally relevant conformations. Instead of sampling the position of each atom in the system, groups of atoms or residues that are part of a shared structural region can be grouped and moved collectively. This gives rise to a conformational sampling strategy that considers the system as a collection of structural regions and exclusively samples their arrangements along the user-defined DOFs. Thus, this method reduces dimensionality by several orders of magnitude by sampling along generalized coordinates. While it does not directly reveal any kinetic information, it can rapidly generate ensembles of thermodynamically feasible structures that appear according to canonical probabilities using computational resources that are readily accessible. In a recent study we showed that NMMC yields comparable results to and is three orders-of-magnitude faster than conventional MD when simulating peptide detachment from class I major histocompatibility complex (MHC I) molecules (21). Traditionally NMMC is used to explore the conformational landscape along a particular set of DOFs chosen by the researcher. Several studies of DNA and protein systems have followed this approach (19, 20, 21, 22, 23, 24, 25). However, the initial choice of DOFs might not always be optimal. Additionally, if the objective is to investigate the causality of functional motions, it may be informative to perform NMMC simulations for a variety of sets of DOFs. Here we introduce a protocol based on customized natural moves (cNMs) to address the challenge of choosing suitable DOFs and to allow for the systematic investigation of hypotheses regarding functional motions in biomolecules. We use cNMs to modulate translations and rotations of segments as well as torsion and bend angles of bonds and compare different sets of cNMMC simulations to infer causal relationships in functional motions. We use two case studies to demonstrate its application. In the first, we investigate functional motions in the class II major histocompatibility complex (MHC II) and in the second, we study the structural effects of an epigenetic mark on a DNA model system. The MHC II is a transmembrane protein that presents potentially harmful peptides to CD4+ T-cells (26). The structure of the peptide-loaded MHC II binding groove is well documented (27); however, to date no structure has been solved for the peptide-free MHC II (28) due to its dynamic nature. Several studies suggest that the absence of peptide destabilizes the MHC II structure (29, 30). Using our protocol, we investigated the functional motions involved in the destabilization of the peptide-free MHC II complex. We designed multiple sets of cNMs and performed NMMC simulations to study the plasticity of the empty MHC II binding groove. Our simulations suggest that the β1 helix can assume a number of transitory states that cause a narrowing of the binding groove in the absence of peptide. We also demonstrate our protocol on the structure of the Dickerson-Drew dodecamer (DDD) that was modified with the 5-hydroxymethylcytosine (5hmC) epigenetic mark. 5hmC is generated by the catalytic activity of oxygenases on 5-methylcytosine (5mC). 5mC is known to increase dsDNA stability, which is consistent with its role in gene expression at CpG islands (31). Given the right sequence context, 5hmC can partly reverse the stabilizing effect of 5mC (32). Lercher et al. (33) have observed two configurations of this epigenetic mark on the DDD, one of which formed a hydrogen bond with the 3′-adjacent guanine. We hypothesized that this noncanonical hydrogen bond interferes with the canonical hydrogen-bonding of the neighboring basepair. To test this hypothesis, we defined a set of cNMs designed to amplify the structural effect of the epigenetic mark on the 3′-adjacent canonical basepair. Due to an increase of several orders of magnitude in computational efficiency compared to atomistic Cartesian sampling methods and its inherent capability to customize DOFs, the cNMMC protocol can facilitate multiple simulations of large biological assemblies. This allows for a side-by-side comparison of different sets of cNMs and enables the testing of hypotheses regarding functional motions in biological systems. Materials and Methods NMMC A molecular system can be defined as a collection of monomers or more formally a set, Ω={m1,…,mN}, where mi for i=1,…,N refers to the residues of a structure. Sequentially numbered residues can be grouped into chains, C={C1,…,CNc}, where Ci={mchi,mchi+1…,mcti} (e.g., chi and cti are indices of the head and tail residues of chain i) for i=1,…,Nc. The chain concept can be further generalized into a group of segments, S={S1,…,SNs}, where Si={mshi,mshi+1…,msti} for i=1,…,Ns (e.g., shi and sti are indices of the head and tail residues of segment i). After defining the segments, we can introduce the set of residues that make up the segments: (1) Ωs=∪i=1NsSi|Ωs⊆Ω. At each iteration in the simulation, the segments are moved in a Monte Carlo fashion along user-defined DOFs, which collectively are called natural moves; these may include translations and rotations of segments as well as torsion and bend angles within segments. After each propagation step, the set of atoms (or entire residues) connecting two segments are rearranged by a linear complexity chain closure algorithm (19). We will refer to this set as the molten zone (MZ). This allows for the reconstruction of chain breaks that may result from the movement of the segments. Thus, the residues outside of the segments form the set of MZ residues (2) Ωm=Ω∖Ωs=∪k=1NMZΩMZ(k), where NMZ is the number of MZs and ΩMZ(k) is the set of residues in the kth molten zone. Furthermore, we define ΩMZ={ΩMZ(1),…,ΩMZ(k)} as the set of molten zones. In this study, Ωs⊂Ω (Ωs is a proper subset of Ω), thus Ωm≠Ø for proteins. For nucleic acids we define each residue as a segment, therefore Ωm=Ø. In this case we use ∗ΩMZ to denote the set of molten zones ∗ΩMZ(j),j=1,…,k, where ∗ΩMZ(j) refers to a molten zone with a set of atoms that are used by the closure algorithm to connect chain breaks that may be caused by moving adjacent nucleotides independently. cNMs cNMs are natural moves that can be modified to investigate functional motions in biomolecules. cNMs include translations and rotations of segments, which may also exhibit internal flexibility such as torsion and bend angles of bonds. cNMs can be created by grouping two segments into one so that they move as a unified segment. Instead of two independent segments that are moved separately, there is now a single set of natural moves that describes the collective motion of both segments. This can be useful, for example, to test whether flexibility in an α-helix kink is important for a particular functional motion or to explore how different levels of collective motion may affect a structural mechanism. Customization may in addition occur at the level of internal flexibility of segments. When internal flexibility is disabled, segments are treated as rigid bodies. When internal flexibility is activated, torsion angles around bonds are changed along with the movement of segments. cNMs also allow us to selectively activate or deactivate sampling of torsional rotations around specific bonds. The protocol We devised a protocol based on cNMs to investigate functional motions in biomolecular structures. The key steps of the protocol: Step I—define a hypothesis; Step II—translate hypothesis into natural moves; Step III—activate/inactivate natural moves to generate test cases for the investigation of the hypothesis; and Step IV—perform conformational sampling on each test case and evaluate the results with respect to the hypothesis. The steps are described in more detail below. Step I: define a hypothesis This step, which is often left to the user, is based on observing experimental data in the literature and/or biological intuition about functionally relevant flexibility, the rigidity or collective motion of atoms, and secondary or even tertiary structure elements. While this step cannot be entirely generalized, it facilitates the engagement of applied scientists, who often have extensive knowledge of the system of interest, in the design of computational experiments so that their hypothesis can be tested. Step II: translate hypothesis into natural moves Based on this hypothesis, an initial set of natural moves can be defined that encapsulates all movements that the researcher specifies as important for the functional motion. These might be residues and/or larger segments as well as the torsion and bend angles of individual bonds. Step III: generate test cases After the initial set of natural moves has been defined, it is then possible to generate different sets of cNMs by selectively deactivating certain DOFs to study their effect on functional motions. Natural moves may be customized by modulating the DOFs that describe the movement of segments (translations and rotations) as well as their internal flexibility (torsion and bend angles of bonds). The relative movement of two segments may result in a chain-break that is closed by the rearrangement of atoms in the molten zone. When two segments are grouped into a bigger segment, the relative orientation of these segments to each other is maintained throughout the simulation. Here, we consider a molten zone ΩMZ(k), for k=1,…,NMZ to be active when the segments on either sides are moved independently and inactive otherwise. For clarity, we only present the case where MZs are made up of residues. For molten zones made up of atoms, we use ∗Ω instead of Ω. Formally we can introduce a function f:ΩMZ→{0,1},f(ΩMZ(k))=1 if ΩMZ(k) is enabled, which leaves corresponding Sj and Sj+1 as independent segments and f(ΩMZ(k))=0 if ΩMZ(k) is disabled, which fuses the two adjacent segments into one, e.g., S(j,j+1). Similarly, some segments may have internal flexibility such as bond torsion angles. Therefore, it is possible to introduce a set of torsion angles Ωϕ(l), for l=1,…,Nϕ and define a function g:Ωϕ→{0,1}, where g(Ωϕ(l))=1 if Ωϕ(l) is active and g(Ωϕ(l))=0 if Ωϕ(l) is inactive. The decomposition and the internal flexibility of segments in a structure may be represented by a vector D in which each element refers to the state of a specific MZ or torsion angle. D may be defined as: (3) D={f(ΩMZ(1)),…,f(ΩMZ(NMZ)),g(Ωϕ(1)),…,g(Ωϕ(Nϕ))}, where ΩMZ and Ωϕ refer to the respective molten zones and torsion angles and |D|=NMZ+Nϕ. Thus, given two functions, f and g, leads to a decomposition D(f,g), which we associate with a test case TD(f,g) (see Eq. 3 for the definition of D). Test cases are associated with a set of functions, e.g., fi,i=1,…,NT where NT is the number of test cases. If we have three molten zones, then there are 23=8 different test cases that result from eight functions with identical domains {ΩMZ(1),ΩMZ(2),ΩMZ(3)} and codomains {0,1}, but unique functional maps. A particular test case may allow for flexibility in an α-helix kink while another test case treats the helix as rigid. Similarly, a selected torsional rotation around a bond may be sampled freely in one test case, while in another test case the dihedral angle is maintained throughout the simulation. This capability allows the researcher to investigate causal relationships between structural features and biophysical mechanisms. Step IV: conformational sampling and evaluation Each test case implies a unique set of DOFs (cNMs) that can be sampled with NMMC. The resulting distributions can then be evaluated with respect to the initial hypothesis. Below we outline the method details for both of our case studies. Note that for reproducible results, replica simulations are needed (21, 28, 34). Implementation details Natural moves change the orientation, position, and internal state (dihedral and bond angles) of structural segments, which are connected by the remaining atoms (e.g., coarse-grained or actual) in the system. These remaining atoms bridge segments and constitute the molten zones (19). While Minary and Levitt (19) describes the anatomy of structural segments and MZs for both models (Cases 1 and 2) in detail, this article aims to provide a high-level practical annotation of natural moves using binary strings. (The implementation details on how to convert these binary strings into customized natural moves is described in “A Tutorial for the Customized Natural Move protocol”, which is available at http://www.cs.ox.ac.uk/mosaics/examples/functional_motions_cNMMC.html.) Simulation details All simulations were carried out with the MOSAICS software package (35). All distributions were plotted with matplotlib (36) and pandas (37) using a bandwidth of 0.1. Protein (MHC II) NMMC simulations were initiated from an x-ray structure of the MHC II (HLA-DR) in complex with HLA-DM at a resolution of 2.6 Å (PDB: 4GBX) (38). The structure was coarse-grained using a 3-point-per-residue protein model (22). We generated the MHC II model by removing the HLA-DM part of the structure file. To ensure extensive conformational sampling, we performed parallel tempering using six replicas at temperatures 300, 336, 376, 421, 472, and 529 K. We ran 15 independent repeats for each test case. Each repeat was run for 1,000,000 Monte Carlo iterations. These parameters were chosen so that the acceptance rates within each replica and the interreplica exchange rates were at least 0.25 and 0.1, respectively. All data were collected from the replica with a canonical temperature of 300 K. Distances were calculated with MDAnalysis (39) and the binding-groove surface area was calculated using differential geometric analysis as described in Hischenhuber et al. (40). DNA (Dickerson-Drew dodecamer) The Dickerson-Drew dodecamer (DDD) in configuration A (5-hydroxymethyl epigenetic marks point toward the O6 oxygen of the 3′-adjacent guanine (G/O6)) (33) at 1.3 Å resolution was used as the starting point. The missing hydroxyl hydrogens were added and oriented toward the 3′-adjacent G/O6. Hydrogens were added to the remaining atoms using pymol1.7’s h_add command (41). The 3′- and 5′-terminal basepairs were removed. An all-atom representation was used with the Amber99-bsc0 force field (42) and a dielectric dampening model (43). Using this model, we reproduced/predicted experimental nucleosome occupancy up to a resolution of a few nucleotides (44). Single-temperature natural-move Monte Carlo was performed at 300 K. We ran 30 independent repeats of 5,000,000 Monte Carlo iterations for each test case. Helical parameters were analyzed using x3DNA (45). Results Here we demonstrate a protocol, based on customized natural moves, which allows the user to design and perform multiple simulation test cases to investigate the causal relationship between different structural features and functional motions. The protocol is computationally cheap and has in-built customization capabilities, which makes the design and run-time of large numbers of customized simulations easily accessible for most research groups. Initially a hypothesis is defined regarding the functional motions of a particular biomolecule. The hypothesis is used to design natural moves that can be systematically turned on/off to test their involvement in a particular functional motion. In the following two case studies, we show how cNMs may help to understand the causality underlying functional motions in biomolecules. Case 1: the plasticity of the empty MHC II binding groove MHC IIs are transmembrane proteins expressed by antigen-presenting cells that are critical for the activation of the adaptive immune response in vertebrates (46, 47). Peptides derived from extracellular proteins that bind the MHC II binding groove inside the cell are transported to the surface and are recognized by receptors on the surface of CD4+ helper T-cells (48). While several MHC II crystal structures with high structural similarity have been solved in the presence of peptide (49), the MHC II structure devoid of peptide has not been solved to date (28). In the absence of peptide, the MHC class II binding groove can take on kinetically distinct forms that are either receptive or averse to peptide binding (29). The receptive state mainly exists straight after peptide dissociation and has a half-life of a few minutes after which the MHC II takes on a peptide averse state (30, 50, 51, 52). Structural changes in the binding groove have been implicated in this process (53, 54). In this case study we demonstrate how cNMs may be used to investigate the plasticity of the empty MHC II binding groove. Here we follow the general steps introduced in the Materials and Methods. Step I: define a hypothesis The literature suggests there are several structural features that may contribute to the plasticity of the empty binding groove. The C-terminal region of helix α1 has been shown to exhibit a distinct conformation in the absence of peptide by mass spectrometric mapping (54). This region is also part of the binding site for the peptide-loading chaperone HLA-DM, and undergoes a structural change upon binding HLA-DM (38). Therefore, we included this structural feature as an area of potential flexibility by introducing a molten zone at the C-terminal end of the α1 helix (α1-1 in Fig. 1 A). Residues β53–68 on helix β1 are part of epitopes for conformation-sensitive antibodies that are selective for the empty binding groove (53, 55). This region has been shown to undergo local structural changes by circular-dichroism spectroscopy (53). MD simulations and comparison of experimental MHC II structures revealed structural variability around a sharp kink in this region (49, 56, 57). Given these observations, we introduced a further MZ at the N-terminal kink of the β1 helix (β1-1 in Fig. 1 A). The second kink on the β1 helix has not been implicated in major structural changes. This is likely due to a disulphide bridge anchoring a conserved cysteine to the β-floor below. However, the segments on either side might still be influenced by flexibility in this kink, so a third molten zone was introduced at this point (β1-2 in Fig. 1 A). Thus, our hypothesis states that conformational flexibility in the three unstructured regions in the two helices (α1 and β1) contributes to the variability of binding groove width and area in the empty MHC II complex. Step II: translate hypothesis into natural moves Our hypothesis on binding groove flexibility provided us with a starting point for defining an initial set of segments, which can undergo three body rotations and translations. This resulted in an initial decomposition consisting of five segments (Fig. 1 B). We used secondary structure information to place MZs between these segments. In this coarse-grained protein case study we did not include any internal flexibility within the segments. Step III: generate test cases In this simple scenario, each of the three MZs may either be enabled or kept rigid, thereby splitting or grouping two neighboring segments. As a result, there are 23=8 different possible test cases that may be generated. For example, test case 010T refers to a system in which MZs ΩMZ(1) and ΩMZ(3) are deactivated and ΩMZ(2) is activated. This creates three regions (one in helix A and two in helix B), as shown in Fig. 1 C. Table 1 presents the remaining test cases. Note that we also introduced permanently activated MZs at the end of the helices to allow for the free movement of all the segments (Fig. S1 in the Supporting Material). Step IV: conformational sampling and evaluation Once the test cases were defined, we used NMMC (20) to generate the distributions seen in Fig. 2. Fig. 2 shows the binding-groove width as defined in Fig. 1 A and surface area distributions as calculated in Hischenhuber et al. (40) for all eight test cases. For clarity, the test cases are shown in two groups. The first group includes the test cases in which ΩMZ(2) was activated (Fig. 2, A and B). The resulting bimodal width and surface area distributions show that the binding groove readily transitions between a wide and a narrow conformation. Depending on the test case, the narrow population is more or less prominent. Test case 010T, for example, exhibits a distribution with clearly defined wide and narrow populations. Note that the distribution was shifted toward the wide population in test cases 110T and 111T when ΩMZ(1), i.e., the α1-1 kink was activated. The second group shows test cases in which ΩMZ(2) was deactivated (Fig. 2, C and D). Some narrowing of the binding groove can be observed for test cases 100T and 101T, but the effect on the surface area is minimal. Generally the binding groove remains in an open conformation when ΩMZ(2), i.e., the β1-1 kink is kept rigid (–0–T). Therefore, our customized natural move simulations suggest that the β1-1 kink plays a crucial role in facilitating a conformational change that results in the narrowing of the binding groove. Biological discussion All MHC class II structures with bound peptide that have been solved to date are structurally highly similar. In the absence of peptide, the MHC II is thought to undergo conformational changes (53, 54). However, presumably due to its floppy nature in the absence of peptide (58), the structure of the empty MHC II has not yet been solved by x-ray crystallography. Other experimental techniques have been employed to show that the empty MHC II assumes at least two distinct forms: a peptide-receptive and a peptide-averse form (29, 30, 50, 51, 52). The receptive form mainly exists immediately after peptide dissociation and turns into the averse form within minutes. Given enough time, however, the averse form can isomerize back to the receptive form (50, 51). The structural mechanisms underlying the conversion from receptive to averse are little understood. One simulation study suggested that partial unfolding of the α1 helix gives rise to a helical segment that binds the P1 pocket of the groove in a peptidelike fashion (56). However, this effect was abrogated when the protonation state of the starting structure was adjusted (49, 57). These studies suggested an involvement of the β1 rather than the α1 helix in the narrowing of the binding groove. In particular, they have shown that the region around the β1-1 kink is highly dynamic (49, 57). Interestingly, the β1-1 kink is part of an epitope for two monoclonal antibodies that selectively bind the empty and not the peptide-loaded MHC II (53, 55). Additionally, MD simulations on an empty MHC I complex have suggested that the helix, which is the equivalent of the β1-helix in MHC II, is responsible for the closing and opening of the binding groove (59). In our simulations, we have observed a similar role of the β1 helix in binding-groove plasticity. Only in test cases where ΩMZ(2) (the β1-1 kink) was active, was a significant narrowing of the binding groove seen (Fig. 2, A and B). Previous observations in the literature regarding conformational heterogeneity of residues β53–68 around the β1-1 kink have been made (49, 53, 55, 56, 57), which are concordant with our own results suggesting that flexibility in the β1 helix provided by the β1-1 kink leads to a collapse of the binding groove. Case 2: structural effects of 5-hmC on the DDD Here we investigated the effect of 5hmC on local basepair arrangement in the DDD: a simple model system that has recently attracted interest due to a new crystal structure with added hydroxymethyl epigenetic marks on cytosines A9 and B9 (33). Two hydroxymethyl configurations were found in this structure. One points toward the backbone phosphate oxygen 5hmC/OP2; the other forms a weak hydrogen bond with the 3′-adjacent G/O6. For the purpose of this case study, we focused on the latter, as it was estimated to be the most prevalent configuration in the crystal (33). A schematic of the system is shown in Fig. 3 A. Next, we apply the four main steps of our protocol to investigate the effects of 5hmC on this structure. Step I: define a hypothesis Lercher et al. (33) observed that the 5hmC hydroxyl formed a noncanonical hydrogen bond with the 3′-G/O6. This oxygen is already part of a canonical (Watson-Crick) hydrogen bond with the C on the opposing strand. No structural differences between the DDD with and without the epigenetic mark were observed, suggesting that any effects that 5hmC might have on the surrounding basepairs cannot be seen in a static structure. We investigate the hypothesis that the hydroxyl-group on 5hmC subtly interferes with the 3′-adjacent G-C basepair. Step II: translate hypothesis into natural moves Given our hypothesis, we defined two sets of cNMs. The first set contained the two torsion angles around bonds C5-C5M (Ωϕ(1)) and C5M-O5 (Ωϕ(2)) in the 5hm epigenetic mark. This gave us control over the orientation of the hydroxyl group during simulation. The second set of cNMs described the collective movement of 5hmC and the 3′-adjacent G, when the MZ between them (∗ΩMZ(1)) was deactivated. This customized natural move was meant to simulate the stabilizing effect caused by a noncanonical intrastrand hydrogen bond between the two neighboring nucleotides. Fig. 3 B shows the cNMs. Note that the depiction of molten zone ∗ΩMZ(1) is an abstraction, as some of the detail was omitted for simplicity. The effect of the cNMs on the distribution of hydrogen-bond distances between the 5hmC hydroxyl and the 3′-G/O6 oxygen is shown in Fig. S2. Step III: generate test cases Given the cNMs that we defined above, we get a decomposition vector D of length 3 (see Materials and Methods). The first two elements refer to rotational freedom along the two torsion angles Ωϕ(1) and Ωϕ(2) in the hydroxyl group of 5hmC, and the third refers to ∗ΩMZ(1) that consists of the backbone atoms between 5hmC and the 3′-adjacent G. Similar to the protein example, each element in D can either be on or off (1/0), i.e., the relative arrangement of G and 5hmC in the case of ∗ΩMZ(1) and the sampling of torsion angles included in Ωϕ(1) and Ωϕ(2) can either be activated or deactivated. Thus, for a decomposition vector DDNA:{g(Ωϕ(1)),g(Ωϕ(2)),f(ΩMZ(1)∗)} of length 3, we get the following 23=8 possible test cases: 000T, 001T, 010T, 100T, 011T, 101T, 110T, and 111T. Note that we only considered test cases where both of the torsion angles were either active or inactive, as we were only interested in a fully flexible or fixed epigenetic mark for this study. Therefore, we omitted test cases 010T, 100T, 011T, and 101T. The remaining test cases included 000T, 001T, 110T, and 111T. Test case 110T was also ignored as it is very similar to test case 000T due to the deactivated molten zone restraining the orientation of the two neighboring bases. Thus, the set of test cases we included in our study were 000T,001T, and 111T. Step IV: conformational sampling and evaluation We ran four sets of simulations of the DDD: the three test cases 111T, 001T, 000T (Fig. 4 A), and a simulation without the epigenetic mark that served as a control. Fig. 4 B shows the distributions of parameters shear, stretch, and propeller, which changed progressively as we applied the different test cases. Note that we only show the distributions for the basepairs around one of the epigenetic marks, but the effect was seen on both ends. Interestingly, the shear was most affected in the GC basepair 3′-adjacent to x, while the stretch and propeller were mostly changed in the 5′-adjacent basepair. No large differences between the modified 111T system and the unmodified control were observed. However, once the orientation of epigenetic mark was fixed (test case 001T), a subtle shift in the distribution was detected. The effect was further increased when the relative movement between 5hmC and the 3′-adjacent G was deactivated (test case 000T). Changes were also observed in the basepair parameters stagger, buckle, and opening, but the effects were less systematic and did not correspond to the increasing epigenetic signal encoded in our test cases (all basepair parameter values are shown in Fig. S3). We did not investigate changes in the base stack parameters (Fig. S4), as we expected that the noncanonical epigenetic (intrastrand) hydrogen bond formation, which we enforce by customized natural moves, could directly impose particular base stacking. However, we were more interested to study distributions over DNA basepair parameters, which were less directly affected by hydrogen-bond formation between adjacent (on the same strand) bases. Biological discussion A sequence of enzymatic reactions drives a cycle of epigenetic cytosine modifications including 5mC, 5hmC, 5-formylcytosine, and 5-carboxylcytosine (60, 61). 5mC has been shown to increase dsDNA stability, which is consistent with its role in gene expression at CpG islands (31). 5hmC, sometimes referred to as the sixth base of the mammalian genome, can partly reverse the 5mC stabilizing given the right sequence context (32) and a study investigating a 27-bp oligonucleotide has observed that 5hmC increases DNA flexibility in MD simulations (62). Several DNA structures with 5hmC epigenetic marks have been solved to date, but no significant structural effects on the DNA helical parameters have been found (33, 63, 64). This is in contrast to a structure of a DNA dodecamer comprising three 5-formyl CpG sites that showed how 5-formylcytosine causes large structural changes that lead to helical underwinding (65). To demonstrate how cNMs can be used to study the effects of epigenetic marks, we chose a recent high-resolution structure of the DDD comprising a 5hmC epigenetic modification. When performing traditional natural-move Monte Carlo, we found that the presence of a single 5hmC epigenetic mark in DDD causes only minimal change in some of the helical parameters of the 3′-adjacent basepair. These results agree with the general view that a single 5hmC epigenetic mark has a limited structural effect on the surrounding helical parameters, which makes it difficult to identify experimentally (33, 63, 64). The results are also in concurrence with Lercher et al. (33), who found that their crystal structures with and without 5hmC were nearly identical with a root-mean-square deviation of 0.35 Å and a 0.8 Å widening of the major groove at the site of modification. However, using cNMs constrained the 5hmC hydroxyl group in the experimentally determined configuration and thereby increased noncanonical intrastrand hydrogen bonding during simulation and we were able to amplify some of the changes caused by the presence of 5hm. The effect was further increased when we deactivated the relative movement between 5hmC and the 3′-G, thereby effectively emulating the stabilizing effect of an intrastrand hydrogen bond. Thus, using cNMs, we were able to detect and amplify subtle structural effects on DNA helical parameters caused by a single epigenetic mark in the DDD. Discussion In this article, we describe a protocol for the testing of hypotheses regarding the functional motions in biological systems. It is based on the natural-move Monte Carlo method that allows for the sampling of conformations given a structural decomposition defined by the researcher. The use of both cNMMC and NMMC assume the decomposition of the molecular system into segments and molten zones. The implementation methodology (19, 20) of NMMC follows a segment-centric approach; if adjacent segments move with respect to each other, their translational and orientational updates are independent; otherwise, a larger segment including the adjacent segments is defined. In cNMMC, end-users may consider each molten zone as active/inactive or 1/0 so that adjacent segments may move independently or synchronously. Using this MZ-centric approach, each set of DOFs (each test case) is associated with a binary string so that test cases can be easily organized and annotated in a systematic and high-throughput manner. In this way, cNMMC reduces the technical barrier to the use of the NMMC approach (19, 20, 23, 28) to study the anatomy of necessary and sufficient sets of DOFs responsible for molecular function. The efficient chain closure algorithm (19) allows the user to introduce arbitrary DOFs into a system without substantially compromising computational run-time. We used this customization capability as the basis for a protocol for the investigation of structural mechanisms. The protocol allows for an investigative strategy using a range of simulations with distinct sets of customized natural moves to test hypotheses concerning the functional motions in biological systems. In molecular biology, a classical approach to testing hypotheses regarding the function of a certain gene is to interfere with its expression and see what happens to the organism. Similarly, in experimental structural biology, residues can be mutated or removed to identify functional regions such as protein-binding or enzyme-active sites (66). However, to our knowledge, this concept of reverse engineering has not been used for investigation of functional motions in simulations. For the first time, to our knowledge, our cNMs protocol enables the testing of hypotheses regarding the functional motions of a biological system by allowing the user to enhance or restrict the movement of certain structural regions. Hypotheses may be derived from biological intuition or computational and/or experimental methods such as ED-CG (15, 67), principal component analysis (PCA) (68), elastic network models (69), normal mode analysis (NMA) (18), and nuclear magnetic resonance (NMR) (70). ED-CG provides information on essential motion by PCA of MD simulations (15) or an elastic network model of a single atomic structure (67). Similarly, low-frequency modes calculated by NMA are often used to approximate collective functional motions in biomolecules (18). In most cases, all the important modes are contained in the normal mode basis set. Thus, NMA provides valuable information on the collective motion of biological systems that may guide the design of natural moves. However, it is often unclear which modes are functionally relevant, as the normal mode basis set contains a range of possible candidates (71). cNMMC simulations may be used to identify functionally relevant modes with different sets of natural moves that represent unique low-frequency modes. Furthermore, NMA or PCA can be costly, as the computational complexity associated with these methods is O(N3) (worst case), where N is the number of atoms in the system. This is due to solving the underlying eigenvalue problem associated with the Hessian matrix. Advanced solvers might produce better scaling, e.g., O(N3) with c < 3; but even by using these advanced methods, the computational cost associated with NMA or PCA will dominate O(N), which is the time complexity of modern algorithms (72, 73) for calculating most statistical (22) or empirical (42) force fields. On the contrary, the worst-case time complexity of NMMC is strictly O(N), because we developed a chain closure algorithm (22) that has linear complexity, O(Nd), in terms of the number of DOFs; here, Nd was used to solve the chain closure (the inverse kinematic) problem. Because Nd < N, the application of natural moves, unlike the calculation of NMA or PCA, will never dominate the computational cost of molecular simulations. This is the main quantifiable advantage of using cNMMC instead of NMA or PCA. A less quantifiable but still notable advantage of cNMMC compared to NMA (or PCA) is that it allows the use of highly unconventional experimentally inferred DOFs such as the hand-shaking motion of adjacent subunits in a chaperonin (23). These experimentally derived moves are not necessarily associated with or dependent on a single conformation (NMA) or conformational ensembles (PCA). They can be simply defined without any limitation to test any experimental observation or intuition. Therefore, NMMC not just supports moves derived using PCA or NMA, but any type of moves (e.g., move any part of the system and the rest will deform to follow the change). The scope of NMMC also differs from the scope of NMA or PCA. NMA (PCA) takes a minimum energy conformation (conformational ensemble) as its input and outputs collective motions or deformations of the molecular system. On the contrary, NMMC takes any collective motion (including but not limited to the ones derived from PCA or NMA) as input and provides distributions as output by exploring the relevant conformational space orders of magnitude more efficiently (22) than conventional methods such as Cartesian Monte Carlo or MD. In this capacity, NMMC has linear O(N) scaling, so it is perfectly fitted to high-throughput testing of customized natural moves. NMA-based Monte Carlo would require the successive recalculation of normal modes in concert with the changing molecular conformation and the computational cost would scale as O(N3) (worst case). Thus, cNMMC should be considered as a complementary approach to NMA or PCA. For example, NMA or PCA can be used in the construction of natural moves and cost-efficient NMMC can explore the conformational space. cNMMC can also be used to test the validity of low-frequency normal mode-based natural moves while exploring the conformational space distant from the minimum energy conformation used to generate the normal mode. Similarly to the above discussion on how cNMMC differs from NMA or PCA, we would like to highlight the differences of cNMMC over MD or Monte Carlo methods with imposed constraints. The latter two methods enable the user to impose constraints on certain DOFs. In contrast, the use of cNMMC primarily facilitates conformational change along a set of user-defined or experimentally inferred DOFs (referred to as natural moves); other DOFs are treated as subordinate (but not constrained) to fully facilitate the exploration of the conformational space along natural moves. This is a very different strategy from Cartesian (or generalized) coordinate-based exploration of the conformational space with constraints, regardless of whether the exploration algorithm is MD- or Monte Carlo-based. Due to the benefits of natural moves, where chain breakage is followed by closure, any part of the molecular assembly can be moved and the necessary subordinate or dependent DOFs will be rearranged to maintain the integrity of the system. This strategy provides users with the opportunity to focus on the essential moves or molecular deformations rather than the less important DOFs. By the straightforward definition of customized natural moves, cNMMC can facilitate the robust compilation of experimentally inferred (23) molecular motion into a molecular simulation protocol. This advantage is particularly relevant for computational structural biology, given the complexity and diversity of biomolecular architectures. Focusing on the natural moves as opposed to the corresponding constraints can provide a more intuitive way to describe, classify, and ultimately understand the mechanisms underpinning functionally relevant motions. The characterization of dynamics in biological molecules is one of the grand challenges of computational structural biology and biophysics, and can only be tackled by the tight collaboration of computational and experimental scientists. To tighten this partnership, the use of cNMMC can catalyze more active engagement of experimental biophysical scientists, who often have extensive experience working on a given biological system in conducting these types of molecular simulations. The more quantifiable advantage of cNMMC compared to MDs (or Monte Carlo) with imposed constraints is the large speed gain from reducing the number of essential DOFs. In a cNMMC protocol the investigation is commonly restricted to a few DOFs (e.g., 6+; orientational and translational parameters of a structural segment plus a few internal dihedral and bond angle DOFs), whereas it is less intuitive for a general user, who might not be specialized in molecular simulations, to automate the procedure for imposing constraints on the remaining DOFs. In addition, the use of dependent DOFs (see the Supporting Material), which significantly facilitates exploration along desired motions, is another unique feature of the presented technology compared to constrained MDs or Monte Carlo methods. With the advantage of being able to define moves liberally and sample conformations along these moves very efficiently, we managed to address applications (20, 21, 23, 25, 44) that were not feasible before NMMC. For example, the latest application (21) demonstrated that we could speed up simulations by orders of magnitude compared to MDs, while still being able to reproduce experimental observables. With cNMMC, these computational experiments will become more accessible to a wider scientific community including experimental laboratories. As described above with NMA and PCA, cNMMC is best used as a complementary method to constrained MDs (or Monte Carlo), which could refine our understanding of systems with DOFs that cNMMC predicts to be relevant. As well as other computational methods, experimental information on collective motions derived from NMR data (70) may also be used to guide the design of cNMs. A range of methods already exist that use NMR data to complement MD simulations (4). Additionally, preexisting expert knowledge is central to generating new ideas. The cNMMC protocol presented here is a first step to bridging the gap between the biological intuition of scientists and molecular simulations by allowing the introduction of arbitrary DOFs for the investigation of conformational changes and mechanisms. In summary, we describe a strategy for the systematic use of customized natural moves to test hypotheses regarding functional motions and have demonstrated the protocol’s ability to provide biological insight into a protein and a DNA system. Conclusion We demonstrated the use of a computationally cheap protocol that uses customized natural moves to investigate the nature of structural changes in a protein and an epigenetically modified DNA system. For each system we generated hypotheses derived from observations in the literature as well as our own preliminary simulation results and performed simulations on a set of different customized natural moves. We showed that this enables the systematic testing of DOFs, which allows for the investigation of causal relationships regarding functional motions in biological systems. Author Contributions S.D., B.K., C.M.D., and P.M. conceived and designed the experiments; S.D. performed the experiments; S.D., C.M.D., and P.M. analyzed the data; B.K. and P.M. contributed reagents/materials/analysis tools; S.D. and P.M. wrote the article; and B.K. and C.M.D. revised the article critically. Supporting Citations Reference (74) appears in the Supporting Material. Supporting Material Document S1. Supporting Materials and Methods and Figs. S1–S4 Document S2. Article plus Supporting Material Acknowledgments We acknowledge the use of the 1st Intel Xeon Phi Access Programme in the Hartree Centre of the UK Science & Technology Facilities Council and the use of the Advanced Research Computing facility at the University of Oxford in carrying out this work. This work was supported by the Engineering and Physical Sciences Research Council [EP/G03706X/1]. Supporting Materials and Methods and four figures are available at http://www.biophysj.org/biophysj/supplemental/S0006-3495(16)30474-X. Figure 1 Decomposing the MHC II binding groove into natural moves. (A) Cartoon representation of the MHC II binding groove (peptide not shown). The three positions α1-1, β1-1, and β1-2 where we defined molten zones ΩMZ(1), ΩMZ(2), and ΩMZ(3) are highlighted by arrows. Helix α1 is shown in green; helix β1 in blue. The HLA-DM binding site is shown in yellow (globular domain contacts not shown). The residues that form the epitope for antibodies specific for the empty binding groove are shown in dark blue. The two-headed arrow indicates where the binding-groove width was measured for analysis (distance between centers of mass of residues α60–65 and β65–70). (B) The initial decomposition resulting from the choice of MZs is shown schematically. Helices α1 and β1 are shown as two green (A1,A2) and three blue rectangles (B1–B3). Each rectangle represents a helical segment that is linked to adjacent segments by molten zones. Each molten zone can be selectively switched on or off (1/0). (C) Example showing test case 010T. The resulting segments are outlined by dotted lines. The six DOFs (three translations and three rotations) for each segment are shown on the top right. To see this figure in color, go online. Figure 2 Distributions of the binding-groove width and surface area generated during simulation. (A and B) The left column shows test cases in which molten zone ΩMZ(2) was activated (–1–T). Note the bimodal width and area distributions, which show that the MHC II binding groove takes on a wide and a narrow binding-groove conformation during simulation. (C and D) The right column shows test cases where the molten zone ΩMZ(2) in the β1-1 kink was deactivated (–0–T). Note that the binding-groove area remains stable for these test cases. To see this figure in color, go online. Figure 3 Defining cNMs for 5-hydroxymethylcytosine in the DDD. (A) Schematic showing the DDD with two added 5-hydroxymethyl (5hm) epigenetic marks. The red lines represent the 5hm epigenetic marks and the thick black horizontal lines represent the bases that are directly affected by the cNMs. (B) The gray line connecting the two nucleotides represents an abstracted backbone chain that may undergo chain breaks during NMMC moves. The dotted rectangles show the collective motion of two neighboring nucleotides when the interjacent molten zone ∗ΩMZ is deactivated or activated. The red lines show the epigenetic mark with the arrows highlighting the torsion angles around C5-C5M and C5M-O, the sampling of which may be deactivated or activated, depending on the test case. (C) Test cases 111T, 001T, and 000T are shown. The dotted lines show individual or collective DOFs depending on the state of the interjacent MZ (active/inactive). The arrows on the epigenetic marks represent rotations around the two torsion angles of 5hm that may be active or inactive. Note that only one of the two epigenetic marks is shown. However, both modifications are treated equivalently in each case. To see this figure in color, go online. Figure 4 the effect of 5-hydroxymethylcytosine on the DDD is amplified by cNMs. (A) The DDD is depicted with the backbone in orange and bases in gray. The two 5hmC modifications are colored based on atom type (O, red; C, green; and H, white). The sets of DOFs chosen for the 5hmC modification are shown on the right. Curved arrows indicate free torsional sampling, while the black crosses indicate fixed χ-torsion on rotations around corresponding bonds (χ1, C5-C5M; χ2, C5M-OH). 111T: Full sampling of all torsion angles; 001T: Fixed torsion angles in 5hmC; 000T: Fixed torsion angles in 5hmC and relative orientation between 5hmC and G. (B) Distributions of the shear, stretch, and propeller are shown for the three different test cases. Each column compares simulations without modification (control) in green against test cases 111T, 001T, and 000T in orange. The shear is shown for basepair 10 and the stretch and propeller are shown for basepair 8. To see this figure in color, go online. Table 1 All Possible Test Cases that Result from the Initial Decomposition Test Case Segments Number of Segments 111T {SA1,SA2,SB1,SB2,SB3} 5 110T {SA1,SA2,SB1,SB2+B3} 4 101T {SA1,SA2,SB1+B2,SB3} 4 011T {SA1+A2,SB1,SB2,SB3} 4 100T {SA1,SA2,SB1+B2+B3} 3 010T {SA1+A2,SB1,SB2+B3} 3 001T {SA1+A2,SB1+B2,SB3} 3 000T {SA1+A2,SB1+B2+B3} 2 The set of segments is shown for each test case. ==== Refs References 1 Henzler-Wildman K. Kern D. Dynamic personalities of proteins Nature 450 2007 964 972 18075575 2 Moraitakis G. Purkiss A.G. Goodfellow J.M. Simulated dynamics and biological macromolecules Rep. Prog. Phys. 66 2003 383 406 3 van Gunsteren W.F. Dolenc J. Mark A.E. Molecular simulation as an aid to experimentalists Curr. Opin. Struct. Biol. 18 2008 149 153 18280138 4 Esteban-Martín S. Bryn Fenwick R. Salvatella X. Synergistic use of NMR and MD simulations to study the structural heterogeneity of proteins Wiley Interdiscip. Rev. Comput. Mol. Sci. 2 2012 466 478 5 Karplus M. Lavery R. Significance of molecular dynamics simulations for life sciences Isr. J. Chem. 54 2014 1042 1051 6 McCammon J.A. Gelin B.R. Karplus M. Dynamics of folded proteins Nature 267 1977 585 590 301613 7 Dror R.O. Dirks R.M. Shaw D.E. Biomolecular simulation: a computational microscope for molecular biology Annu. Rev. Biophys. 41 2012 429 452 22577825 8 Freddolino P.L. Arkhipov A.S. Schulten K. Molecular dynamics simulations of the complete satellite tobacco mosaic virus Structure. 14 2006 437 449 16531228 9 Beberg, A. L., D. L. Ensign, …, V. S. Pande. 2009. Folding@Home: lessons from eight years of volunteer distributed computing. In 2009 IEEE International Symposium on Parallel & Distributed Processing. Institute of Electrical and Electronics Engineers, Piscataway, NJ. 1–8. 10 Lindorff-Larsen K. Piana S. Shaw D.E. How fast-folding proteins fold Science 334 2011 517 520 22034434 11 Orozco M. A theoretical view of protein dynamics Chem. Soc. Rev. 43 2014 5051 5066 24709805 12 Bernardi R.C. Melo M.C.R. Schulten K. Enhanced sampling techniques in molecular dynamics simulations of biological systems Biochim. Biophys. Acta 1850 2015 872 877 25450171 13 Duan M. Fan J. Huo S. Evaluation of dimensionality-reduction methods from peptide folding-unfolding simulations J. Chem. Theory Comput. 9 2013 2490 2497 23772182 14 Saunders M.G. Voth G.A. Coarse-graining methods for computational biology Annu. Rev. Biophys. 42 2013 73 93 23451897 15 Zhang Z. Lu L. Voth G.A. A systematic methodology for defining coarse-grained sites in large biomolecules Biophys. J. 95 2008 5073 5083 18757560 16 Bahar I. Lezon T.R. Eyal E. Global dynamics of proteins: bridging between structure and function Annu. Rev. Biophys 39 2010 23 42 20192781 17 Perilla J.R. Woolf T.B. Towards the prediction of order parameters from molecular dynamics simulations in proteins J. Chem. Phys. 136 2012 164101 22559464 18 Bahar I. Rader A.J. Coarse-grained normal mode analysis in structural biology Curr. Opin. Struct. Biol. 15 2005 586 592 16143512 19 Minary P. Levitt M. Conformational optimization with natural degrees of freedom: a novel stochastic chain closure algorithm J. Comput. Biol. 17 2010 993 1010 20726792 20 Sim A.Y.L. Levitt M. Minary P. Modeling and design by hierarchical natural moves Proc. Natl. Acad. Sci. USA 109 2012 2890 2895 22308445 21 Knapp B. Demharter S. Minary P. Exploring peptide/MHC detachment processes using hierarchical natural move Monte Carlo Bioinformatics 32 2016 181 186 26395770 22 Minary P. Levitt M. Probing protein fold space with a simplified model J. Mol. Biol. 375 2008 920 933 18054792 23 Zhang J. Minary P. Levitt M. Multiscale natural moves refine macromolecules using single-particle electron microscopy projection images Proc. Natl. Acad. Sci. USA 109 2012 9845 9850 22665770 24 Sim A.Y.L. Minary P. Levitt M. Modeling nucleic acids Curr. Opin. Struct. Biol. 22 2012 273 278 22538125 25 Moraga I. Wernig G. Garcia K.C. Tuning cytokine receptor signaling by re-orienting dimer geometry with surrogate ligands Cell 160 2015 1196 1208 25728669 26 Cresswell P. Assembly, transport, and function of MHC class II molecules Annu. Rev. Immunol. 12 1994 259 293 8011283 27 Jones E.Y. Fugger L. Siebold C. MHC class II proteins and disease: a structural perspective Nat. Rev. Immunol. 6 2006 271 282 16557259 28 Knapp B. Demharter S. Deane C.M. Current status and future challenges in T-cell receptor/peptide/MHC molecular dynamics simulations Brief. Bioinform. 16 2015 1035 1044 25725219 29 Sadegh-Nasseri S. McConnell H.M. A kinetic intermediate in the reaction of an antigenic peptide and I-Ek Nature 337 1989 274 276 2536141 30 Natarajan S.K. Assadi M. Sadegh-Nasseri S. Stable peptide binding to MHC class II molecule is rapid and is determined by a receptive conformation shaped by prior association with low affinity peptides J. Immunol. 162 1999 4030 4036 10201925 31 Münzel M. Globisch D. Carell T. 5-Hydroxymethylcytosine, the sixth base of the genome Angew. Chem. Int. Ed. Engl. 50 2011 6460 6468 21688365 32 Thalhammer A. Hansen A.S. Schofield C.J. Hydroxylation of methylated CpG dinucleotides reverses stabilisation of DNA duplexes by cytosine 5-methylation Chem. Comm 47 2011 5325 5327 21451870 33 Lercher L. McDonough M.A. Schofield C.J. Structural insights into how 5-hydroxymethylation influences transcription factor binding Chem. Comm 50 2014 1794 1796 24287551 34 Knapp B. Dunbar J. Deane C.M. Large scale characterization of the LC13 TCR and HLA-B8 structural landscape in reaction to 172 altered peptide ligands: a molecular dynamics simulation study PLoS One 10 2014 e1003748 35 Minary P. Methodologies for Optimization and Sampling in Computational Studies (MOSAICS), Vers 3.9 2007 Department of Computer Science, University of Oxford Oxford, UK http://www.cs.ox.ac.uk/mosaics 36 Hunter J.D. Matplotlib: A 2D graphics environment Comput. Sci. Eng. 9 2007 90 95 37 McKinney, W. 2011. pandas: a Foundational Python Library for Data Analysis and Statistics. Presented at PyHPC2011, November 18, 2011. http://pandas.pydata.org/talks.html. 38 Pos W. Sethi D.K. Wucherpfennig K.W. Crystal structure of the HLA-DM-HLA-DR1 complex defines mechanisms for rapid peptide selection Cell 151 2012 1557 1568 23260142 39 Michaud-Agrawal N. Denning E.J. Beckstein O. MDAnalysis: a toolkit for the analysis of molecular dynamics simulations J. Comput. Chem. 32 2011 2319 2327 21500218 40 Hischenhuber B. Havlicek H. Knapp B. Differential geometric analysis of alterations in MH α -helices J. Comput. Chem. 34 2013 1862 1879 23703160 41 DeLano W.L. The PyMOL Molecular Graphics System, Ver. 1.7.4 2014 Schrödinger LLC http://www.schrodinger.com/ 42 Pérez A. Marchán I. Orozco M. Refinement of the AMBER force field for nucleic acids: improving the description of α /γ conformers Biophys. J. 92 2007 3817 3829 17351000 43 Rohs R. Etchebest C. Lavery R. Unraveling proteins: a molecular mechanics study Biophys. J. 76 1999 2760 2768 10233091 44 Minary P. Levitt M. Training-free atomistic prediction of nucleosome occupancy Proc. Natl. Acad. Sci. USA 111 2014 6293 6298 24733939 45 Lu X.J. Olson W.K. 3DNA: a software package for the analysis, rebuilding and visualization of three-dimensional nucleic acid structures Nucleic Acids Res. 31 2003 5108 5121 12930962 46 van de Rijn M. Bernabeu C. Terhorst C. Recognition of HLA-A2 by cytotoxic T lymphocytes after DNA transfer into human and murine cells Science 226 1984 1083 1085 6333726 47 Schreier M.H. Iscove N.N. von Boehmer H. Clones of killer and helper T cells: growth requirements, specificity and retention of function in long-term culture Immunol. Rev. 51 1980 315 336 7000675 48 Vyas J.M. van der Veen A.G. Ploegh H.L. The known unknowns of antigen processing and presentation Nat. Rev. Immunol. 8 2008 607 618 18641646 49 Yaneva R. Springer S. Zacharias M. Flexibility of the MHC class II peptide binding cleft in the bound, partially filled, and empty states: a molecular dynamics simulation study Biopolymers 91 2009 14 27 18767126 50 Rabinowitz J.D. Vrljic M. McConnell H.M. Formation of a highly peptide-receptive state of class II MHC Immunity 9 1998 699 709 9846491 51 Joshi R.V. Zarutskie J.A. Stern L.J. A three-step kinetic mechanism for peptide binding to MHC class II proteins Biochemistry 39 2000 3751 3762 10736175 52 Kasson P.M. Rabinowitz J.D. McConnell H.M. Kinetics of peptide binding to the class II MHC protein I-Ek Biochemistry 39 2000 1048 1058 10653650 53 Zarutskie J.A. Sato A.K. Stern L.J. A conformational change in the human major histocompatibility complex protein HLA-DR1 induced by peptide binding Biochemistry 38 1999 5878 5887 10231540 54 Carven G.J. Stern L.J. Probing the ligand-induced conformational change in HLA-DR1 by selective chemical modification and mass spectrometric mapping Biochemistry 44 2005 13625 13637 16229453 55 Carven G.J. Chitta S. Stern L.J. Monoclonal antibodies specific for the empty conformation of HLA-DR1 reveal aspects of the conformational change associated with peptide binding J. Biol. Chem 279 2004 16561 16570 14757758 56 Painter C.A. Cruz A. Zavala-Ruiz Z. Model for the peptide-free conformation of class II MHC proteins PLoS One 3 2008 e2403 18545669 57 Rupp B. Günther S. Kühne R. Characterization of structural features controlling the receptiveness of empty class II MHC molecules PLoS One 6 2011 e18662 21533180 58 Sadegh-Nasseri S. Germain R. A role for peptide in determining MHC class II structure. (Letter) Nature 353 1991 167 170 1653903 59 Zacharias M. Springer S. Conformational flexibility of the MHC class I α 1-α 2 domain in peptide bound and free states: a molecular dynamics simulation study Biophys. J. 87 2004 2203 2214 15454423 60 Goll M.G. Bestor T.H. Eukaryotic cytosine methyltransferases Annu. Rev. Biochem. 74 2005 481 514 15952895 61 Ito S. Shen L. Zhang Y. Tet proteins can convert 5-methylcytosine to 5-formylcytosine and 5-carboxylcytosine Science 333 2011 1300 1303 21778364 62 Wanunu M. Cohen-Karni D. Drndic M. Discrimination of methylcytosine from hydroxymethylcytosine in DNA molecules J. Am. Chem. Soc. 133 2011 486 492 21155562 63 Renciuk D. Blacque O. Spingler B. Crystal structures of B-DNA dodecamer containing the epigenetic modifications 5-hydroxymethylcytosine or 5-methylcytosine Nucleic Acids Res. 41 2013 9891 9900 23963698 64 Szulik M.W. Pallan P.S. Stone M.P. Differential stabilities and sequence-dependent base pair opening dynamics of Watson-Crick base pairs with 5-hydroxymethylcytosine, 5-formylcytosine, or 5-carboxylcytosine Biochemistry 54 2015 1294 1305 25632825 65 Raiber E.-A. Murat P. Balasubramanian S. 5-Formylcytosine alters the structure of the DNA double helix Nat. Struct. Mol. Biol. 22 2015 44 49 25504322 66 Morrison K.L. Weiss G.A. Combinatorial alanine-scanning Curr. Opin. Chem. Biol. 5 2001 302 307 11479122 67 Zhang Z. Pfaendtner J. Voth G.A. Defining coarse-grained representations of large biomolecules and biomolecular complexes from elastic network models Biophys. J. 97 2009 2327 2337 19843465 68 Balsera M.A. Wriggers W. Schulten K. Principal component analysis and long time protein dynamics J. Phys. Chem. 100 1996 2567 2572 69 Atilgan A.R. Durell S.R. Bahar I. Anisotropy of fluctuation dynamics of proteins with an elastic network model Biophys. J. 80 2001 505 515 11159421 70 Ravera E. Salmon L. Luchinat C. Insights into domain-domain motions in proteins and RNA from solution NMR Acc. Chem. Res. 47 2014 3118 3126 25148413 71 Ma J. Usefulness and limitations of normal mode analysis in modeling dynamics of biomolecular complexes Structure 13 2005 373 380 15766538 72 Mattson W. Rice B.M. Near-neighbor calculations using a modified cell-linked list method Comput. Phys. Commun. 119 1999 135 148 73 Yao Z. Wang J.-S. Cheng M. Improved neighbor list algorithm in molecular simulations using cell decomposition and data sorting method Comput. Phys. Commun. 161 2004 27 35 74 Metropolis N. Rosenbluth A.W. Teller E. Equation of state calculations by fast computing machines J. Chem. Phys. 21 1953 1087

PMC005xxxxxx/PMC5002071.txt

==== Front Biophys JBiophys. JBiophysical Journal0006-34951542-0086The Biophysical Society S0006-3495(16)30540-910.1016/j.bpj.2016.07.009Channels and TransportersExploring the Dynamics of the TWIK-1 Channel Oakes Victoria 1Furini Simone 2Pryde David 3Domene Carmen carmen.domene@kcl.ac.uk14∗1 Department of Chemistry, King’s College London, London, United Kingdom2 Department of Medical Biotechnologies, University of Siena, Siena, Italy3 Worldwide Medicinal Chemistry, Pfizer Neuroscience and Pain Research Unit, Cambridge, United Kingdom4 Chemistry Research Laboratory, Mansfield Road, University of Oxford, Oxford, United Kingdom∗ Corresponding author carmen.domene@kcl.ac.uk23 8 2016 23 8 2016 111 4 775 784 30 12 2015 11 7 2016 © 2016 Biophysical Society.2016Biophysical SocietyThis is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).Potassium channels in the two-pore domain family (K2P) have various structural attributes that differ from those of other K+ channels, including a dimeric assembly constituted of nonidentical domains and an expansive extracellular cap. Crystallization of the prototypical K2P channel, TWIK-1, finally revealed the structure of these characteristics in atomic detail, allowing computational studies to be undertaken. In this study, we performed molecular-dynamics simulations for a cumulative time of ∼1 μs to discern the mechanism of ion transport throughout TWIK-1. We observed the free passage of ions beneath the extracellular cap and identified multiple high-occupancy sites in close proximity to charged residues on the protein surface. Despite the overall topological similarity of the x-ray structure of the selectivity filter to other K+ channels, the structure diverges significantly in molecular-dynamics simulations as a consequence of nonconserved residues in both pore domains contributing to the selectivity filter (T118 and L228). The behavior of such residues has been linked to channel inactivation and the phenomenon of dynamic selectivity, where TWIK-1 displays robust Na+ inward flux in response to subphysiological K+ concentrations. Editor: Emad Tajkhorshid. ==== Body Introduction The two-pore domain K+ family (K2P) forms a structurally and functionally distinct class of K+ channels. These channels are responsible for background leak K+ currents that stabilize the negative resting potential of the cell, and also play roles in ion homeostasis, hormone secretion, cell development, and excitability (1). K2P channels can be modulated by a vast array of regulatory stimuli, such as pH (2), temperature (3), mechanical stress (4), the presence of polyunsaturated fatty acids (5), and volatile anesthetics (6). The expression of these channels in the heart and brain has also led to increased exploration of their therapeutic potential for the treatment of various neuronal and cardiac disorders (5, 7). K2P channels assemble as a dimer of dimers, with each subunit containing four transmembrane helices (H1–H4) and two pore loops (P1 and P2). The recent crystallization of multiple K2P channels (TWIK-1 (8), TRAAK (9, 10), and TREK-2 (11)) has greatly advanced our understanding of this unique architecture, exhibiting various conserved features throughout the family. For example, an extracellular cap (EC) between H1 and P1 is present, extending 35 Å above the transmembrane helices, with the apex of each subunit connected by a disulphide bond. An unrestricted pathway at the cytoplasmic entrance is also observed throughout, revealing apertures of comparable dimension to other K+ channels that are considered open. This is in line with experimental studies suggesting that structural changes at or near the selectivity filter (SF) form the predominant gating mechanism across the K2P family, which shares its canonical structure with other K+ channel families. With regard to the electrophysiological properties of the K2P channels, in asymmetrical K+ concentrations, almost all K2P subfamilies (TREK, TALK, TASK, THIK, and TRESK) conform to the typical properties of leak K+ currents, demonstrating outward (or open) rectification. The functional properties of the TWIK family have remained elusive due to low levels of activity recorded in physiological K+ gradients (12). This phenomenon was originally attributed to the sumoylation of a lysine residue in the C-terminal domain, which could be inactivated by a single-point mutation (13). However, this was later disproved as the primary mechanism of TWIK-1 silencing (14). The mutation of consecutive isoleucine residues in the C-terminal domain was found to induce strong expression of TWIK-1 in the cell membrane (14); however, meager currents were still recorded relative to other K2P channels in the same conditions (15, 16). Such observations led to the proposal that the prohibition of ionic current is an inherent property of TWIK-1 (17). Furthermore, reports have suggested that TWIK-1 does not exhibit such open rectification. Various mutant channels were found to both increase the conductance and shift the rectification properties to those of other K2P channels, indicating a convoluted regulation mechanism. In particular, TWIK-1 was the first channel to display variable selectivity in response to external stimuli such as lowered K+ concentrations (18), increased NH4+ and Rb+ concentrations (19), and acidification (20). Hypokalemia (when extracellular K+ concentrations are lower than 3.5 mM) is observed in up to 20% of hospitalized patients and has been associated with an increased risk of sudden cardiac arrest (21). The adjusted selectivity of TWIK-1 in these conditions indicates the presence of an inward Na+ current in response to the Na+ concentration gradient, and this property is known to depolarize cardiomyocytes and potentially contribute to cardiac arrhythmia (7, 22). As a consequence, TWIK-1 has emerged as a putative drug target for antiarrhythmic drugs (23). In-depth exploration of the molecular determinants of conduction, selectivity, and gating in TWIK-1 will likely contribute to the development of targeted therapies. Previous molecular-dynamics (MD) studies of TWIK-1 identified a hydrophobic cuff in the inner pore that is responsible for a cyclical dewetting process and consequently an unfavorable barrier to conduction (24). This effect was suggested to be influenced by the presence of lipid molecules in proximal fenestrations (25), and was supported by studies involving gain-of-function mutagenesis of such residues to hydrophilic components (24). However, the behavior of the SF and its implications for the mechanism of conduction, selectivity, and gating have not yet been examined. Therefore, we conducted an MD study utilizing the crystal structure of TWIK-1 to obtain insight into these phenomena. Materials and Methods System setup The crystal structure of TWIK-1 was retrieved from the Protein Data Bank (PDB: 3UKM) at a resolution of 3.4 Å (residues 19–288) (8). Five potassium ions were resolved in the crystal structure in the internal SF sites (S1–S4) and S0, suggesting that the structure is representative of an open, conductive state. Three of these ions were kept in positions S0, S2, and S4, and those in positions S1 and S3 were converted to water molecules to represent one of the low-energy conformations identified in previous studies (26). Crystallographic waters were kept. Missing loops were modeled using Modloop (27) and combined with the crystal structure. N- and C-termini were acetylated and methylated, respectively. Residues C69 of opposing subunits were linked by a disulfide bond. Default protonation states were used for ionizable residues, supported by PropKa calculations (28). SOLVATE1.0 was used to solvate the protein and fill cavities present in the structure. A preequilibrated lipid bilayer of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) molecules was used. The protein was aligned to the bilayer normal and inserted into the membrane. All lipid molecules within 1.2 Å of protein atoms were removed. The combined system was then solvated to produce a rectangular water box of dimensions 93 × 93 × 118 Å. Potassium and chloride ions were added to neutralize the system to a biologically relevant ion concentration (150 mM) using the Autoionize Plugin of VMD (29). All water molecules deemed as overlapping (distance < 1.2 Å) with the protein, lipids, and ions were removed, resulting in a system size of ∼90,000 atoms. Trajectories of 200 ns were produced and denoted as HSE (δ position in H122 protonated, D230 unprotonated), HSEP (δ position in H122 protonated, D230 protonated), HSD (ε position in H122 protonated, D230 unprotonated), HSDP (ε position in H122 protonated, D230 protonated), HSP (both δ and ε positions in H122 protonated, D230 unprotonated), and HSPP (both δ and ε positions in H122 protonated, D230 protonated). MD simulations NAMD2.9 was employed to calculate trajectories (30). We used the CHARMM36 force field for the protein, CHARMM36 for lipids (31), the TIP3P model for water (32), and the CHARMM NBFIX parameters for ions (33, 34). The particle mesh Ewald method was used for the treatment of periodic electrostatic interactions, with an upper threshold of 1 Å for grid spacing (35). Electrostatic and van der Waals forces were calculated every time step. A cutoff distance of 12 Å was used for van der Waals forces. A switching distance of 10 Å was chosen to smoothly truncate the nonbonded interactions. Only atoms in a Verlet pair list with a cutoff distance of 13.5 Å (reassigned every 20 steps) were considered (36). The SETTLE algorithm was used to constrain all bonds involving hydrogen atoms, to allow the use of a 2 fs time step throughout the simulation (37). The Nose-Hoover-Langevin piston method was employed to control the pressure with a 200 fs period, 50 fs damping constant, and a desired value of 1 atmosphere (38, 39). The system was coupled to a Langevin thermostat to sustain a temperature of 310 K throughout, to maintain the model membrane above its gel transition temperature. Equilibration protocol The systems were subjected to 1000 steps of minimization and equilibrated for a total of 3.5 ns. The duration of each equilibration step was 500 ps with a gradual reduction of restraints throughout: 1) protein atoms, ions in the SF, lipid headgroups, and water molecules within protein cavities restrained; 2) protein atoms, ions in the SF, and water molecules within protein cavities restrained; 3) protein atoms and ions in the SF restrained; 4) protein backbone atoms, SF atoms, and ions in the SF restrained; 5) SF atoms and ions in the SF restrained; 6) SF backbone atoms and ions in the SF restrained; and 7) SF ions restrained only. Results and Discussion K2P channels exhibit a unique architecture (illustrated in Fig. 1 A) formed from the assembly of two identical subunits (denoted A and B). Each subunit consists of two nonidentical pore domains, with the former including an expansive EC. Despite the substantial structural variations, TWIK-1 displays the archetypal conductive SF structure, which is highly conserved in numerous human and bacterial K+-channel crystal structure analogs (Fig. 2 A). Backbone carbonyls from each subunit and pore domain within the subunits assemble in a cage-like structure to form four adjacent binding sites (S1–S4) that are capable of binding dehydrated K+ ions. Additional sites are capable of binding partially hydrated species at the intracellular (SC) and extracellular (S0) exits. The sequence is divergent from the signature sequence TXGYG, where X represents any hydrophobic amino acid; P1 is TTGYG and P2 is TIGLG. On closer inspection, a degree of asymmetry is observed in the pore; the distance between the T118 (P1) carbonyls is 4.3 Å, compared with 4.8 Å between I226 (P2) atoms. Furthermore, the SF is known to be influenced by external pH, with ionizable residues present at the top of the SF in both P1 and P2. H122, in particular, has been established as the putative proton sensor in TWIK-1, as well as in the K2P channels TASK-1 and TASK-3, that responds to changes in extracellular pH (17). How these features influence the dynamics of the SF and hence ion permeation on an atomistic level is currently unknown. To address this issue, we performed multiple independent MD simulations to gain insight into the properties of the TWIK-1 SF, as outlined in Materials and Methods. In all systems, the root mean-square deviations of the transmembrane domain (Fig. S1 in the Supporting Material) and the EC (Fig. S2) show initial jumps associated with the release of restraints present in the initial equilibration procedure, and remain under 2.5 Å and 3.5 Å, respectively, throughout, suggesting that the channel is stable and representative of the determined structure. It is well established in the literature that cytoplasmic gating does not modulate activity in K2P channels. In agreement with this, the cytoplasmic gate does not undergo any significant constriction throughout our simulations, and the slide helix remains parallel to the membrane normal. Ion transport to the SF K2P channels possess an extracellular domain that is distinct from other K+ channel families, extending ∼35 Å above the transmembrane domain. The geometry of this region imposes obvious steric constraints on the diffusion pathway to the central pore (Fig. 1 A). In addition, the residual composition is predominantly negatively charged, providing a sink for surrounding cations. To gain insight into the transport of ions in this region, we tracked the x and y positions of ions entering the region between the SF and the turret above, excluding occupancy of the S0–S4 sites. The bounding box was defined by z coordinates of the Cα atoms of G80 in the EC and residue G121 at the mouth of the SF, with the x-y coordinates restricted by the region occupied by the transmembrane helices. Density plots of ion distribution, displayed in Fig. 1, B–F, show that bidirectional diffusion of ions occurs through side portals of the protruding domain. Negligible diffusion is seen in Fig. 1 G, corresponding to the HSPP simulation in which both H122 and D230 are protonated. The density plots in Fig. 1 pinpoint three unique regions of increased ion density. First, the region above the SF, denoted 1 in Fig. 1 E, is consistently a high-occupancy region, with the exception of the HSPP simulation. The charged nature of the SF and above extracellular helices, in addition to contacts with surrounding residues, all contribute to binding in this region, with subtle differences dependent on the filter conformation. A central site can be occupied between the carbonyls of G121 and G229 in both subunits, at the upper bounds of the S0 site. Additional instances are also observed in which ions can be found in off-axis sites interacting with H122 (protonated in the δ or ε position), G121, G229, D230 (unprotonated), S86 (EC), or N87 (EC) residues. A site was identified in the HSDP simulation (labeled 2 in Fig. 1 F) occupying a region directly behind the P1 SF sequence, with ions predominantly interacting with N101, H122, N242, and E235. An additional site in close proximity to this (denoted 3 in Fig. 1 G) was also found in the HSPP simulation. This site is present between the P1 SF sequence and P-loop helix, and is defined by direct interactions with E207 (H3-P2 loop), V232 (P2-H4 loop), E235 (P2-H4 loop), and extracellular water molecules. The close proximity of these sites to the SF may have implications for the conformation of the SF and hence ion permeation. Ion-binding sites in the SF To probe the behavior of ions within the SF, we tracked the positions of the four ions in contact with the SF for the longest period of time (Fig. 2, B–G) and analyzed their behavior. No complete permeation events were observed within any trajectory. However, individual ion movements provide insight into the stability of the K+ binding sites and the behavior of ions within them. A comparison of the conduction profiles of all of the simulations reveals that the S2–S4 K+ configuration, as well as the structure observed in the crystallographic data, shows variable stability in the TWIK-1 SF and is sustained for ∼5–200 ns. This is in spite of the absence of a concentration gradient or an applied voltage. With the exception of HSPP, the ion occupying the S4 site is generally stagnant for the timescale of our simulations in the absence of additional ions, due to the consistent coordination to backbone carbonyls and side-chain hydroxyl groups of T117 and T225, resulting in a coordination number of 7 or 8 throughout. In contrast, the ion that originally occupies S2 is subject to reduced coordination at both the upper and lower bounds of the site, and abstraction from this site is observed in all simulations except HSD. T118 in P1 shows increased conformational freedom, resulting in prolonged periods where ion contacts cannot be formed and hence elevation of the ion in the site, which consequently reduces the coordinating ability of I226 in the equivalent position in P2. Additionally, structural changes originating from the top of the P2 SF sequence also lead to reduced coordination of G227, further contributing to the instability of the occupying ion. In the case of HSD, conformational changes at the top of the P1 domain of the SF result in a constriction at the S0 site and the removal of multiple carbonyls from the S1 site, rendering the S2 site the most favorable for an ion in this region of the SF. Qualitatively, these observations are in agreement with the crystallographic data, which demonstrate ion density in S1 to S4 sites. In the HSE, HSDP, and HSPP simulations, the S1–S4 conformation is sustained for ∼100–200 ns. Due to the increased conformational freedom of the upper region of the filter in both the P1 and P2 domains, the ion in S1 is subject to coordination by carbonyls from two G119 residues. The remaining contributions come from Y120 and G227 residues and additional water molecules depending on the conformation of the SF, which generally exhibits a full coordination shell despite divergence from the canonical structure. Furthermore, transient constrictions in the S0 region of the P1 domain, similar to those observed in the HSD simulation, likely contribute to ion maintenance in this site, and expulsion to the extracellular solution is only permitted in the absence of these constrictions. Structure of the SF Experimental evidence indicates that the K+ channel SF can adopt multiple conformations in addition to the stereotypical conductive conformation (40). Certain voltage-gated K+ channels have shown Na+ conductance in the absence of internal K+ ions, suggesting that the SF can adopt different open conformations with shifting selectivity (41, 42, 43, 44, 45, 46). In the case of TWIK-1, inward rectification of Na+ and outward rectification of K+ have been measured in low extracellular K+ concentrations and acidic conditions, even though it exhibits the attributes of a highly selective channel in normal physiological conditions (18), albeit with low conductance properties. These results indicate that TWIK-1 has at least three distinct SF conformations: one conductive, one inactivated, and one with altered selectivity properties. Therefore, we focused our attention on characterizing the conformations of the TWIK-1 SF in response to the range of ionic configurations we identified. The crystal structure is representative of a typical conductive conformation that consists of four contiguous binding sites and is capable of providing a full coordination shell to dehydrated K+ ions (47). This structure is observed initially in the MD trajectories (Fig. 3 A) but diverges rapidly in all simulations to a number of conformations that contain defective coordination sites yet maintain the general framework of the SF and can be considered partially conductive. This is due to an amalgamation of distinctive P1 and P2 conformations (Fig. 3 B). The disparity compared with the typical structure in P1 is localized in the region of T118, which exhibits rotational freedom, causing lateral expansion of the S2 and S3 sites. Crystallization of the noninactivating E71A KcsA mutant revealed remarkable similarities to the structure of P1 in this region (48). In the P2 domain, the canonical structure is lost in the upper region, with residue D230 exhibiting a dynamic behavior, both in protonated and unprotonated forms. A consequence of this higher mobility is a reduced ion coordination by protein atoms in binding sites S0–S2. The functional state of this conformation cannot be attained in our simulations, but as all sites remain viable to accommodate K+ ions, it is likely that conduction, even if less favorable, can still occur in this state. This is observed in all simulations except HSD, where a constriction in the S0 site is observed in the P1 domain (Fig. 3 C), occluding water and ions from entering the SF from this angle for the remainder of the simulation. The predominance of such states throughout the simulation may have implications for the low conductance properties of TWIK-1 relative to other K2P channels. The vacancy of both the central S2 and S3 sites induces conformations in which the S2 site is physically occluded by backbone rearrangements of SF residues (Fig. 3 D). In the HSE and HSP simulations, this occurs by movement of T118 and G119 in P1, whereas in the HSEP and HSPP simulations, this state is observed in I226 and G227 in P2. In these configurations, ions and water molecules are excluded from the S2 site for the remainder of the simulation. The S4 and S3 sites are maintained, with diffuse S0 and S1 sites containing multiple molecules as in the previous conformation. The constriction of the filter shares similar attributes with the crystal structure of KcsA in low K+ concentrations, where the SF is blocked by a constriction involving V76 and G77 residues, and V76 is oriented away from its optimal position (49). This was confirmed to be nonconductive (50, 51, 52) and suggested to be representative of a C-type inactivated state (53). It must be noted that this state is stable on a millisecond timescale, which is unattainable during our simulations; therefore, it is possible that this conformation is an intermediary state that can block permeation transiently, but does not represent the typical inactivated state. Finally, in the HSPP simulation, a novel, to our knowledge, structure of the SF was observed in which all binding sites were depleted of ions but the SF remained open (Fig. 3 E). Previous computational studies suggested the existence of a pathway behind the SF (54) for water transport in collapsed K+ channels (55). However, these results demonstrate an atomistic representation of the K+ SF that can stably occupy water in the absence of ions and potentially allow water permeation directly through the SF. These observations provide insight into the unique behavior of the SF of TWIK-1 as a consequence of a single sequence difference in each domain. Structural changes in P1 Compelling experimental evidence suggests that the hydrogen (H)-bond network behind the SF plays an integral role in determining the structure of the SF, influenced by the presence of structural water molecules and the orientation of local residues (56). Therefore, we analyzed the atomic interactions that stabilized the observed conformations. An examination of T118 is of particular importance due to its confirmed role in the variable selectivity of TWIK-1 in response to low extracellular K+ concentrations (18). To understand the interrelation between the behavior of this residue and the ionic configurations, we characterized its conformation throughout our simulations using the Ψ (backbone) and X (side chain) dihedral angles (Fig. 4, A and B). Overall, the Ψ backbone conformations could be prorated into seven clusters, with Ψ values centered at approximately −95°, −45°, 20°, 65°, 110°, 140°, and 170°. These clusters were denoted I–VII in ascending order. In each cluster, the side-chain dihedral angle X could occupy three rotameric states, g+, g−, and t, corresponding to average X angles of 36°, −41°, and −176°, respectively (57). These conformations will be referred to by these classifications henceforth, and the structures of the most highly populated conformations are shown in Fig. 4 B. Conformation II (g−) is representative of the crystal structure with the carbonyls poised for ion binding, and therefore can be considered to be conductive. This conformation is one of the least populated conformations in our simulations, displaying a complex H-bond network as a means of stabilization for a fully conductive filter. T118 and G119 act as H-bond donors to the T113 carbonyl on the pore helix, and the T118 side chain is capable of H-bonding with both S116 and S222 side chains. The remaining filter residues interact with up to two structured water molecules, which in turn contact the side chains of T113, H122, and T123. In the additional T118 conformations we identified, the backbone carbonyl deviates from this orientation; hence, the extent to which T118 can bind permeant ions in S2 and S3 is compromised. Conformation IV (g−) represents a flipped state, with the carbonyl perpendicular to the pore axis. Further rotation of T118 results in a blocked conformation, V (g−). Backbone atoms in this region physically occlude the permeation axis, resulting in complete removal of the S2 site and blockage of the SF. Such conformations have been established as part of the normal functioning of K+ channels via MD simulations, depending on the ion occupation of the SF. The H-bond network is highly conserved with regard to conformation II (g−), with minimal reorientations involving the T113 carbonyl and water molecules that now interact with the T118 backbone. In conformation III (t), the carbonyl is rotated ∼75° relative to the crystal structure, displaying an H-bond between the side-chain hydroxyl and backbone carbonyl of T118, and conserved H-bonding properties elsewhere. As hydrophobic residues usually occupy this position, this conformation appears to be unique to TWIK-1 and, to the best of our knowledge, has not been characterized in previous computational studies. A prerequisite of the H-bond network described in these conformations is that H122 must be in a so-called down state, where the histidine ring is embedded in the small cavity behind the SF. It must be noted, therefore, that H122 can also occupy an up state, with the ring displaying enhanced flexibility above the filter. In this case, extracellular water molecules penetrate the region previously occupied by the H122 side chain, and interact with selectivity filter residues to stabilize similar conformations. An additional conformation is also observed in which T118 is in the region of the Ramachandran plot corresponding to a β-sheet, with the T118 side chain occupying the pore and H-bonding with surrounding water molecules. In this case, H122 is primarily in the up state, and is predominant throughout our simulations. In both conformations III (t) and VII (t), coordinating atoms in the S2 and S3 sites are removed, providing a direct link between the behavior of T118 and the structure of the TWIK-1 SF. The S2 site has been identified as the most selective site in K+ channels; therefore, these conformations may represent a potentially nonselective state, as shown experimentally in reduced K+ concentrations and extracellular pH (18, 20). The importance of the so-called pH sensor, H122, for the different conformations of P1 suggests a further association between the latter stimulus and the dynamic selectivity phenomenon. Structural changes in P2 The P2 sequence in TWIK-1 is also not conserved with respect to other K+ channels, since the representative TXGYG sequence is constituted of TIGLG in this channel. The sampled conformations of this domain display a variable H-bond network involving water molecules and pore helix (Y217 and I221), SF (I226, G227, and G229), and loop (D230 and Y231) residues. Favorable hydrophobic interactions between the bulky side chains of I226 and L228 can also be identified. The lower region of the P2 filter domain (S2–S4) is highly conserved with respect to other K+ channels and thus exhibits paramount stability relative to P1, demonstrating only conductive and constricted conformations. With regard to the upper region of the filter, the preservation of the S0–S2 sites is dependent on the maintenance of interactions between Y217 and D230 (Fig. 5). Detachment of these residues (in both protonation states) propagates structural changes throughout the SF, notably rotation of L228 and G229, resulting in numerous structural variants in which the upper sites are removed and incapable of ion binding. Most commonly, the behavior of this region is largely determined by dynamic interactions with surrounding water molecules and cations, as well as transient interactions with K131. Comparison with other K+ channels Investigations of SF dynamics have focused largely on the prokaryotic K+ channel, KcsA. The crystal structure of KcsA exhibits a complex network of H-bonds involving contributions from the SF (G77, Y78, and G79), adjacent pore helices (W67 and E71), the succeeding extracellular loop (D80), and an adjacent water molecule, stabilizing the conductive conformation of the filter (49). In the nonconductive case, a reorganized H-bond arrangement was observed that was devoid of direct E71-Y78 and E71-G77 interactions, which were replaced by three structured water molecules behind the SF, the presence of which is thought to control the recovery from slow inactivation (58). A multitude of structural, functional, and computational investigations have since indicated that the W67, E71, D80 triad is crucial in determining the degree to which C-type inactivation occurs (48, 59, 60). Crystal structures of the E71A mutant provided an atomistic description of a noninactivating SF. The nonflipped state showed remarkable similarities to the conductive state of KcsA in spite of low K+ concentrations. Remarkably, the W67-D80 interaction was maintained. In contrast, W67 and D80 occupy alternative conformational states in the flipped state, with the latter extending above the SF, resulting in a novel, to our knowledge, SF framework in which the S2 and S3 sites are merged due to rotations of V76. Residues that occupy equivalent positions in TWIK-1 play an integral role in the conformations we observed, despite their low conservation. In P1, T113 and H122 occupy the positions of E71 and D80 in KcsA, undergoing interactions with each other and SF residues mediated by water molecules. This network, in addition to specific T118 interactions, typically serves to maintain conductive and nonconductive states with the SF, and it also stabilizes partially conductive conformations. These open states epitomize the flipped E71A mutant, with T118 and H122 mimicking the flexible nature of V76 and D80, respectively. In P2, the W67-D80 interaction is conserved in Y217 and D230 in the conductive conformation, but is primarily detached in the remaining conformations, with swelled S0 and S1 sites. This is consistent with the position of L228, where KcsA contains Y78, suggesting that the presence of this residue may destabilize this region. To assess whether the dynamics we observed in the SF are unique to TWIK-1, we aligned multiple sequences constituting the SF and the surrounding environment (Fig. 6), and compared interactions that are known to stabilize the SF. In Kv1.2, which is representative of the voltage-gated K+ channel family, residues D379 and W366 are integral and display exact conservation with D80 and W67 in KcsA, respectively. In Kir2.2, which is representative of the inward-rectifying K+ channel family, a salt-bridge between E139 (conserved with E71) and R149 is responsible for maintenance of the conductive SF structure. This suggests that TWIK-1 cannot be likened to such channel families. With regard to other K2P channels, TRAAK and TREK-2 demonstrate highly conserved sequences in the pore region but diverge from TWIK-1 with regard to the T118, H122, and L228 positions, which contain I, N, and F residues, respectively. This is consistent with our proposal that the variability of the SF structure originates from these residues, and supports experimental evidence indicating that TWIK-1 is also atypical within the K2P family with regard to its conduction properties. Conclusions Throughout our computational study, we gained insights into the highly dynamic behavior of the TWIK-1 SF. Multiple K+ ions can occupy alternate sites in the SF, bound in a typical cage-like manner to surrounding carbonyl atoms; however, we find the behavior of SF is disparate from that previously recorded. The divergence of the SF in comparison with other K+ channels is exemplified at the S2 site, with diverse behaviors from both nonconserved P domains in the SF converging here. We identified multiple conformations in P1 that exhibited lateral expansion of the S2 and S3 sites, primarily depending on the conformational state of T118. Introduction of this residue into other K2P channels induced the dynamic-selectivity phenomenon observed in subphysiological conditions in native TWIK-1 channels; therefore, it is possible that such conformations may have varied selectivity properties. The dependence on the surrounding H-bond network, in particular the behavior of H122, raises the possibility that external stimuli, such as K+ concentration and pH, can be translated to structural changes in the SF. In addition, we observed a widening at the extracellular mouth of the P2 domain, removing the upper sites in the SF. The protonation state of the residues of the selectivity was extensively tested. With the exception of a single simulation (with H122 protonated in the δ position and the unprotonated state of D230), ions were always observed to leave the binding site S2. Upon exclusion of ions from S2, we observed additional conformational states that physically blocked the SF, similar to what was observed in KcsA in low K+ concentrations. Finally, in a simulation in which both H122 and D230 were in protonated forms and both ions were lost from the filter, the SF remained in an open conformation and all sites were filled with water molecules. TWIK-1 is the first channel, to our knowledge, to display such noninactivating characteristics in a computational simulation. Overall, these results shed light on the conduction properties of TWIK-1, which exhibits low levels of activity in physiological conditions, yet a robust inward Na+ current in subphysiological K+ concentrations and upon acidification. The connection between such properties and the paradoxical depolarization of cardiomyocytes has potential implications for the development of channel-based therapies for associated cardiac disorders. Author Contributions V.O. performed the modeling and analyzed the data. V.O., S.F., and C.D. interpreted the data and wrote the manuscript. D.P. critically revised the manuscript. Supporting Material Document S1. Figs. S1–S3 Document S2. Article plus Supporting Material Acknowledgments We thank ARCHER, the UK National Supercomputing Service (http://www.archer.ac.uk); the Hartree Centre; and the National Service for Computational Chemistry Software for providing computational resources. This work was supported by the Biotechnology and Biological Sciences Research Council and Pfizer Neusentis. Three figures are available at http://www.biophysj.org/biophysj/supplemental/S0006-3495(16)30540-9. Figure 1 Extracellular ion transport pathways in TWIK-1. (A) Side view (upper panel) and top view (bottom panel) of the region between the EC (G80) and the SF (G121), where the position of K+ ions (orange spheres) was tracked. (B–G) Color maps showing the position of extracellular ions entering the filter region in simulations HSE, HSD, HSP, HSEP, HSDP, and HSPP, respectively. Ion positions were measured as the x and y coordinates of the center of mass and then discretized into bins of 0.5 Å. To see this figure in color, go online. Figure 2 (A) Crystal structure of the P1 and P2 domains of the TWIK-1 SF, with the initial ion configuration used in all simulations. Definitions of the binding sites are provided on the left. The SF residues T117–G121 and T225–G229 are displayed in licorice representation, with van der Waals spheres representing ions and water molecules. Oxygen, nitrogen, carbon, sodium, and potassium atoms are shown in red, blue, cyan, yellow, and orange, respectively. (B–G) Ion trajectories in simulations HSE (B), HSD (C), HSP (D), HSEP (E), HSDP (F), and HSPP (G). The black traces correspond to the center of mass of the oxygen atoms of the SF residues that contribute to the binding sites. The blue, red, yellow, and pink trajectories correspond to the trajectories of individual K+ ions. Representative snapshots of each ion configuration can be found in Fig. S3. To see this figure in color, go online. Figure 3 Snapshots of the identified SF conformations. The SF residues T117–G121 and T225–G229 are displayed in licorice representation, with van der Waals spheres representing ions and water molecules, and oxygen, nitrogen, carbon, sodium, and potassium atoms shown in red, blue, cyan, yellow, and orange, respectively. (A) HSD, 10 ns. (B) HSDP, 5 ns. (C) HSD, 200 ns. (D) HSEP, 200 ns. (E) HSPP, 200 ns. To see this figure in color, go online. Figure 4 (A) Heat map of Ψ and X angles of T118 throughout all simulations. (B) Snapshots of the full P1 SF in the most populated T118 conformations (from left to right): II (g−)/conductive, IV (g−)/flipped, V (g−)/inactivated, III (t)/open conformation (I), and VII (t)/open conformation with H122 in the up state. The upper and lower panels represent the front and side views of the SF, respectively. SF residues T117–H122 and H-bond partner residues S113, T123, and S222 are shown in cyan and orange, respectively, with red spheres representing water molecules. To see this figure in color, go online. Figure 5 Snapshots of the observed P2 conformations in licorice representation, using the coloring scheme defined in Fig. 2 for P2 residues, and the carbon atoms of additional H-bonding residues shown in orange. (A) Conductive conformation. (B) Conformation with S2–S4 sites maintained, with disruption in the upper sites. To see this figure in color, go online. Figure 6 Sequence alignment of the SF region with multiple K+ channels. P1 and P2 denote the nonidentical domains observed in K2P channels. To see this figure in color, go online. ==== Refs References 1 Enyedi P. Czirják G. Molecular background of leak K+ currents: two-pore domain potassium channels Physiol. Rev. 90 2010 559 605 20393194 2 Sandoz G. Douguet D. Lesage F. Extracellular acidification exerts opposite actions on TREK1 and TREK2 potassium channels via a single conserved histidine residue Proc. Natl. Acad. Sci. USA 106 2009 14628 14633 19667202 3 Kang D. Choe C. Kim D. Thermosensitivity of the two-pore domain K+ channels TREK-2 and TRAAK J. Physiol. 564 2005 103 116 15677687 4 Maingret F. Fosset M. Honoré E. TRAAK is a mammalian neuronal mechano-gated K+ channel J. Biol. Chem. 274 1999 1381 1387 9880510 5 Lesage F. Maingret F. Lazdunski M. Cloning and expression of human TRAAK, a polyunsaturated fatty acids-activated and mechano-sensitive K+ channel FEBS Lett. 471 2000 137 140 10767409 6 Patel A.J. Honoré E. Lazdunski M. Inhalational anesthetics activate two-pore-domain background K+ channels Nat. Neurosci. 2 1999 422 426 10321245 7 Es-Salah-Lamoureux Z. Steele D.F. Fedida D. Research into the therapeutic roles of two-pore-domain potassium channels Trends Pharmacol. Sci. 31 2010 587 595 20951446 8 Miller A.N. Long S.B. Crystal structure of the human two-pore domain potassium channel K2P1 Science 335 2012 432 436 22282804 9 Brohawn S.G. del Mármol J. MacKinnon R. Crystal structure of the human K2P TRAAK, a lipid- and mechano-sensitive K+ ion channel Science 335 2012 436 441 22282805 10 Brohawn S.G. Campbell E.B. MacKinnon R. Physical mechanism for gating and mechanosensitivity of the human TRAAK K+ channel Nature 516 2014 126 130 25471887 11 Dong Y.Y. Pike A.C.W. Carpenter E.P. K2P channel gating mechanisms revealed by structures of TREK-2 and a complex with Prozac Science 347 2015 1256 1259 25766236 12 Lesage F. Guillemare E. Barhanin J. TWIK-1, a ubiquitous human weakly inward rectifying K+ channel with a novel structure EMBO J. 15 1996 1004 1011 8605869 13 Rajan S. Plant L.D. Goldstein S.A. Sumoylation silences the plasma membrane leak K+ channel K2P1 Cell 121 2005 37 47 15820677 14 Feliciangeli S. Tardy M.P. Lesage F. Potassium channel silencing by constitutive endocytosis and intracellular sequestration J. Biol. Chem. 285 2010 4798 4805 19959478 15 Goldstein S.A. Price L.A. Pausch M.H. ORK1, a potassium-selective leak channel with two pore domains cloned from Drosophila melanogaster by expression in Saccharomyces cerevisiae Proc. Natl. Acad. Sci. USA 93 1996 13256 13261 8917578 16 Ben-Abu Y. Zhou Y. Yifrach O. Inverse coupling in leak and voltage-activated K+ channel gates underlies distinct roles in electrical signaling Nat. Struct. Mol. Biol. 16 2009 71 79 19098918 17 Chatelain F.C. Bichet D. Lesage F. TWIK1, a unique background channel with variable ion selectivity Proc. Natl. Acad. Sci. USA 109 2012 5499 5504 22431633 18 Ma L. Zhang X. Chen H. TWIK-1 two-pore domain potassium channels change ion selectivity and conduct inward leak sodium currents in hypokalemia Sci. Signal. 4 2011 ra37 21653227 19 Ma L. Xie Y.P. Chen H. Silent TWIK-1 potassium channels conduct monovalent cation currents Biophys. J. 102 2012 L34 L36 22768960 20 Ma L. Zhang X. Chen H. Acid-sensitive TWIK and TASK two-pore domain potassium channels change ion selectivity and become permeable to sodium in extracellular acidification J. Biol. Chem. 287 2012 37145 37153 22948150 21 Paice B.J. Paterson K.R. Lawson D.H. Record linkage study of hypokalaemia in hospitalized patients Postgrad. Med. J. 62 1986 187 191 3714603 22 Schaefer T.J. Wolford R.W. Disorders of potassium Emerg. Med. Clin. North Am. 23 2005 723 747., viii–ix. (viii–ix.) 15982543 23 Schmidt C. Wiedmann F. Thomas D. Class I antiarrhythmic drugs inhibit human cardiac two-pore-domain K+ (K2 2 p) channels Eur. J. Pharmacol. 721 2013 237 248 24070813 24 Aryal P. Abd-Wahab F. Tucker S.J. A hydrophobic barrier deep within the inner pore of the TWIK-1 K2P potassium channel Nat. Commun. 5 2014 4377 25001086 25 Aryal P. Abd-Wahab F. Tucker S.J. Influence of lipids on the hydrophobic barrier within the pore of the TWIK-1 K2P channel Channels (Austin) 9 2015 44 49 25487004 26 Aqvist J. Luzhkov V. Ion permeation mechanism of the potassium channel Nature 404 2000 881 884 10786795 27 Fiser A. Sali A. ModLoop: automated modeling of loops in protein structures Bioinformatics 19 2003 2500 2501 14668246 28 Li H. Robertson A.D. Jensen J.H. Very fast empirical prediction and rationalization of protein pKa values Proteins 61 2005 704 721 16231289 29 Humphrey W. Dalke A. Schulten K. VMD: visual molecular dynamics J. Mol. Graph. 14 1996 33 38., 27–28 8744570 30 Phillips J.C. Braun R. Schulten K. Scalable molecular dynamics with NAMD J. Comput. Chem. 26 2005 1781 1802 16222654 31 Klauda J.B. Venable R.M. Pastor R.W. Update of the CHARMM all-atom additive force field for lipids: validation on six lipid types J. Phys. Chem. B 114 2010 7830 7843 20496934 32 Jorgensen W.L. Chandrasekhar J. Klein M.L. Comparison of simple potential functions for simulating liquid water J. Chem. Phys. 79 1983 926 935 33 Beglov D. Roux B. Finite representation of an infinite bulk system: solvent boundary potential for computer simulations J. Chem. Phys. 100 1994 9050 9063 34 Luo Y. Roux B. Simulation of osmotic pressure in concentrated aqueous salt solutions J. Phys. Chem. Lett. 1 2010 183 189 35 Darden T. York D. Pedersen L. Particle mesh Ewald: an N⋅log(N) method for Ewald sums in large systems J. Chem. Phys. 98 1993 10089 10092 36 Verlet L. Computer “experiments” on classical fluids. I. Thermodynamical properties of Lennard-Jones molecules Phys. Rev. 159 1967 98 37 Miyamoto S. Kollman P.A. SETTLE: an analytical version of the SHAKE and RATTLE algorithm for rigid water models J. Comput. Chem. 13 1992 952 962 38 Feller S.E. Zhang Y. Brooks B.R. Constant pressure molecular dynamics simulation: the Langevin piston method J. Chem. Phys. 103 1995 4613 4621 39 Martyna G.J. Tobias D.J. Klein M.L. Constant pressure molecular dynamics algorithms J. Chem. Phys. 101 1994 4177 4189 40 Hoshi T. Armstrong C.M. C-type inactivation of voltage-gated K+ channels: pore constriction or dilation? J. Gen. Physiol. 141 2013 151 160 23319730 41 Kiss L. LoTurco J. Korn S.J. Contribution of the selectivity filter to inactivation in potassium channels Biophys. J. 76 1999 253 263 9876139 42 Korn S.J. Ikeda S.R. Permeation selectivity by competition in a delayed rectifier potassium channel Science 269 1995 410 412 7618108 43 Wang Z. Zhang X. Fedida D. Regulation of transient Na+ conductance by intra- and extracellular K+ in the human delayed rectifier K+ channel Kv1.5 J. Physiol. 523 2000 575 591 10718739 44 Starkus J.G. Kuschel L. Heinemann S.H. Ion conduction through C-type inactivated Shaker channels J. Gen. Physiol. 110 1997 539 550 9348326 45 Starkus J.G. Kuschel L. Heinemann S.H. Macroscopic Na+ currents in the “Nonconducting” Shaker potassium channel mutant W434F J. Gen. Physiol. 112 1998 85 93 9649585 46 Gang H. Zhang S. Na+ permeation and block of hERG potassium channels J. Gen. Physiol. 128 2006 55 71 16769794 47 Doyle D.A. Morais Cabral J. MacKinnon R. The structure of the potassium channel: molecular basis of K+ conduction and selectivity Science 280 1998 69 77 9525859 48 Cordero-Morales J.F. Cuello L.G. Perozo E. Molecular determinants of gating at the potassium-channel selectivity filter Nat. Struct. Mol. Biol. 13 2006 311 318 16532009 49 Zhou Y. Morais-Cabral J.H. MacKinnon R. Chemistry of ion coordination and hydration revealed by a K+ channel-Fab complex at 2.0 A resolution Nature 414 2001 43 48 11689936 50 Domene C. Furini S. Dynamics, energetics, and selectivity of the low-K+ KcsA channel structure J. Mol. Biol. 389 2009 637 645 19393663 51 Domene C. Klein M.L. Parrinello M. Conformational changes and gating at the selectivity filter of potassium channels J. Am. Chem. Soc. 130 2008 9474 9480 18588293 52 Cuello L.G. Jogini V. Perozo E. Structural basis for the coupling between activation and inactivation gates in K+ channels Nature 466 2010 272 275 20613845 53 Cuello L.G. Jogini V. Perozo E. Structural mechanism of C-type inactivation in K+ channels Nature 466 2010 203 208 20613835 54 Furini S. Beckstein O. Domene C. Permeation of water through the KcsA K+ channel Proteins 74 2009 437 448 18636474 55 Hoomann T. Jahnke N. Pohl P. Filter gate closure inhibits ion but not water transport through potassium channels Proc. Natl. Acad. Sci. USA 110 2013 10842 10847 23754382 56 Raghuraman H. Islam S.M. Perozo E. Dynamics transitions at the outer vestibule of the KcsA potassium channel during gating Proc. Natl. Acad. Sci. USA 111 2014 1831 1836 24429344 57 van der Kamp M.W. Schaeffer R.D. Daggett V. Dynameomics: a comprehensive database of protein dynamics Structure 18 2010 423 435 20399180 58 Ostmeyer J. Chakrapani S. Roux B. Recovery from slow inactivation in K+ channels is controlled by water molecules Nature 501 2013 121 124 23892782 59 Cordero-Morales J.F. Jogini V. Perozo E. Molecular driving forces determining potassium channel slow inactivation Nat. Struct. Mol. Biol. 14 2007 1062 1069 17922012 60 Cordero-Morales J.F. Jogini V. Perozo E. A multipoint hydrogen-bond network underlying KcsA C-type inactivation Biophys. J. 100 2011 2387 2393 21575572

PMC005xxxxxx/PMC5002091.txt

==== Front Arthritis Res TherArthritis Res. TherArthritis Research & Therapy1478-63541478-6362BioMed Central London 108910.1186/s13075-016-1089-8Research ArticleA2A adenosine receptor upregulation correlates with disease activity in patients with systemic lupus erythematosus http://orcid.org/0000-0003-2416-8134Bortoluzzi Alessandra +390532239651brtlsn1@unife.it 1Vincenzi Fabrizio vncfrz@unife.it 2Govoni Marcello gvl@unife.it 1Padovan Melissa pdvmss@unife.it 1Ravani Annalisa rvnnls@unife.it 2Borea Pier Andrea andrea.borea@unife.it 2Varani Katia vrk@unife.it 21 Department of Medical Science, Section of Rheumatology, University of Ferrara, Ferrara and Azienda Ospedaliero-Universitaria Sant’Anna di Ferrara, Ferrara Cona (Ferrara) Via Aldo Moro 8 44124 Cona, Ferrara, Italy 2 Department of Medical Sciences, Institute of Pharmacology, University of Ferrara, Via Fossato di Mortara 17-19, Ferrara, Italy 26 8 2016 26 8 2016 2016 18 1 19221 6 2016 3 8 2016 © The Author(s). 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Background Adenosine is a purine nucleoside implicated in the regulation of the innate and adaptive immune systems, acting through its interaction with four cell surface receptors: A1, A2A, A2B, and A3. There is intense interest in understanding how adenosine functions in health and during disease, but surprisingly little is known about the actual role of adenosine-mediated mechanisms in systemic lupus erythematosus (SLE). With this background, the aim of the present study was to test the hypothesis that dysregulation of A1, A2A, A2B, and A3 adenosine receptors (ARs) in lymphocytes of patients with SLE may be involved in the pathogenesis of the disease and to examine the correlations between the status of the ARs and the clinical parameters of SLE. Methods ARs were analyzed by performing saturation-binding assays, as well as messenger RNA and Western blot analysis, with lymphocytes of patients with SLE in comparison with healthy subjects. We tested the effect of A2AAR agonists in the nuclear factor kB (NF-kB) pathway and on the release of interferon (IFN)-α; tumor necrosis factor (TNF)-α; and interleukin (IL)-2, IL-6, IL-1β, and IL-10. Results In lymphocytes obtained from 80 patients with SLE, A2AARs were upregulated compared with those of 80 age-matched healthy control subjects, while A1, A2B, and A3 ARs were unchanged. A2AAR density was inversely correlated with Systemic Lupus Erythematosus Disease Activity Index 2000 score disease activity through time evaluated according to disease course patterns, serositis, hypocomplementemia, and anti-double-stranded DNA positivity. A2AAR activation inhibited the NF-kB activation pathway and diminished inflammatory cytokines (IFN-α, TNF-α, IL-2, IL-6, IL-1β), but it potentiated the release of anti-inflammatory IL-10. Conclusions These data suggest the involvement of A2AARs in the complex pathogenetic network of SLE, acting as a modulator of the inflammatory process. It could represent a compensatory pathway to better counteract disease activity. A2AAR activation significantly reduced the release of proinflammatory cytokines while enhancing those with anti-inflammatory activity, suggesting a potential translational use of A2AAR agonists in SLE pharmacological treatment. Keywords Systemic lupus erythematosusA2A adenosine receptorDisease activityAny financial support was not given for the work reported on in the manuscript (Internal founding).noneissue-copyright-statement© The Author(s) 2016 ==== Body Background Systemic lupus erythematosus (SLE) is the prototypic multisystem autoimmune disorder with a broad spectrum of clinical presentations encompassing almost all organs and tissues [1]. The irreversible break in immunological tolerance is manifested by immune responses against endogenous nuclear antigens and the subsequent formation of autoantibodies and immune complexes (ICs). SLE has classically been considered an autoimmune disease with a predominant adaptive immune system component because T and B cells have been considered the most important pathogenetic players [2]. During the early inflammatory phase, plasmacytoid dendritic cells (DCs) are able to internalize nucleic acids containing interferogenic ICs that reach the endosomes and stimulate Toll-like receptor 7 or 9, leading to interferon (IFN)-α gene transcription [3–5]. IFN-α contributes to the maturation of myeloid DCs that can activate autoreactive T cells through antigen presentation and costimulation. This favors the development of T helper 1 cells responsible for the high-level production of proinflammatory cytokines [6–8] and enhances B-cell maturation and differentiation, antibody production, and IC formation. In SLE, the IC- and IFN-α-secreting monocytes modulate interleukin (IL)-10 function [8]. The capability of IL-10 to suppress production of inflammatory cytokines such as tumor necrosis factor (TNF)-α and IL-6, implicated in promoting autoimmunity and tissue inflammation in SLE, is attenuated [8]. Growing evidence emphasizes that the purine nucleoside adenosine plays an active role as a local regulator of inflammation in different pathologies. Adenosine is a ubiquitous nucleoside involved in various physiological and pathological functions by stimulating the G protein-coupled A1, A2A, A2B, and A3 adenosine receptors (ARs) [9–12]. The role of ARs is well known in physiological conditions and in a variety of pathologies, including inflammatory damage, neurodegenerative disorders, and cancer [13–15]. In particular, A2AAR stimulation mediates inhibition of TNF-α, IL-1β, IL-2, IL-6, and IFN-α [16–18] and increases the production of the anti-inflammatory cytokine IL-10 [19]. With this background, the aim of the present study was to explore the arrangement and functionality of ARs in SLE and to evaluate their relationship with clinical phenotype and disease activity. Methods Patients and study design Patients with SLE regularly attending our lupus clinic and satisfying the 1997 revised American College of Rheumatology criteria [20] were consecutively recruited from the Rheumatology Unit, Sant’Anna Hospital, University of Ferrara, Ferrara, Italy. We recorded clinical, demographic, and serological data, as well as data regarding therapy, including corticosteroids (measured as prednisone equivalent), antimalarials, and immunosuppressants. Disease activity routinely assessed using the Systemic Lupus Erythematosus Disease Activity Index 2000 (SLEDAI-2 K) [21] and cumulative damage assessed with the Systemic Lupus International Collaborating Clinics Index were extracted by retrieving information from clinical records and a dedicated database. Moreover, disease activity and progress through time were considered according to four different patterns defined using the SLEDAI-2 K, excluding serological descriptors (hypocomplementemia and anti-double-stranded DNA [anti-dsDNA] antibodies) to focus on clinical activity: chronic active disease (CAD), relapsing-remitting disease, clinical quiescent disease (CQD), and minimal disease activity [22]. Seroimmunologic tests included complement components C3 and C4 dosages, antinuclear antibody (ANA), anti-dsDNA, anti-Sjögren’s-syndrome-related antigen A (anti-SSA), anti-Sjögren’s-syndrome-related antigen B (anti-SSB), anti-Smith (anti-Sm), antiribonucleoprotein (anti-RNP), anticardiolipin (aCL), anti-β2-glycoprotein I (anti-β2-GPI), and lupus anticoagulant (LA). C3 and C4 (in grams per liter) were measured by nephelometry, and hypocomplementemia was defined by local laboratory reference values (e.g., C3 < 0.8 g/L and C4 < 0.11 g/L detected on at least two separate occasions). ANA were detected by indirect immunofluorescence using HEp-2 cells as a substrate; positivity was defined as a titer ≥1:160. Anti-dsDNA were detected by indirect immunofluorescence using Crithidia luciliae with a cutoff titer of 1:40; positivity was certified if confirmed in two separate measurements. Anti-SSA, anti-SSB, anti-Sm, and anti-RNP were detected by using an immunoblotting technique. aCL and anti-β2-GPI were measured by enzyme-linked immunosorbent assay (ELISA) [23]. LA was measured in accordance with the recommendation of the Scientific and Standardization Committee of the International Society of Thrombosis and Hemostasis. Positivity for antiphospholipid antibodies (aPL) and LA was defined if confirmed in two separate measurements performed 12 weeks apart [24]. Healthy volunteers (n = 80) matched for age and sex ratio from the Ferrara University Hospital Blood Bank served as a control group. Sample collection and human lymphocyte preparation Lymphocytes were isolated and prepared as previously described from the peripheral blood of control subjects and patients with SLE [25–27]. Leukocytes were separated from erythrocytes with a solution of 6 % dextran T500 (Sigma-Aldrich, St. Louis, MO, USA), suspended in Krebs-Ringer phosphate buffer, and layered onto 10 ml of Ficoll-Hypaque density gradient (GE Healthcare Life Sciences, Little Chalfont, UK). After centrifugation, mononuclear cells were washed in 0.02 M phosphate-buffered saline (PBS) at pH 7.2 and containing 5 mM MgCl2 and 0.15 mM CaCl2. They were then decanted into a culture flask and placed in a humidified incubator (5 % CO2) for 2 h at 37 °C. This procedure, aimed at removing monocytes that adhered to the culture flasks, resulted in a purified lymphocyte preparation containing at least 99 % small lymphocytes identified by morphological criteria. To obtain membrane suspensions, cell fractions were centrifuged in a hypotonic buffer at 20,000 × g for 10 minutes. The resulting pellet was incubated in Tris-HCl 50 mM buffer, pH 7.4, with 2 IU/ml adenosine deaminase (Sigma-Aldrich) for 30 minutes at 37 °C. After centrifugation at 40,000 × g for 10 minutes, the final pellet was used for radioligand binding assays. The protein concentration was determined by using a Bio-Rad Laboratories (Hercules, CA, USA) method with bovine serum albumin as the reference standard [25]. Real-time reverse transcriptase-polymerase chain reaction experiments Total cytoplasmic RNA was obtained from human lymphocytes by using the acid guanidinium thiocyanate phenol method. Quantitative real-time reverse transcriptase-polymerase chain reaction (RT-PCR) assay [25–28] of A1, A2A, A2B, and A3 ARs messenger RNAs (mRNAs) was performed using gene-specific fluorescently labeled TaqMan MGB Probe (minor groove binder) in an ABI Prism 7700 Sequence Detection System (Applied Biosystems, Foster City, CA, USA). Real-time RT-PCR for A1, A2A, A2B, and A3 ARs was carried out with the Assays-on-Demand TM Gene expression Products NM_000674, NM_000675, NM_000676, and NM_000677 (Applied Biosystems), respectively. For the real-time RT-PCR of the reference gene, the endogenous control human β-actin was used, and the probe was fluorescently labeled with VIC™ dye (Applied Biosystems). Western blot analysis Human lymphocytes were washed with ice-cold PBS and lysed in radioimmunoprecipitation assay buffer (Sigma-Aldrich) containing protease inhibitors and 1 mM sodium orthovanadate. Proteins were eluted in Laemmli buffer, resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and transferred to polyvinylidene fluoride membranes. Next, the membranes were incubated with specific antibodies for ARs (Alpha Diagnostic International, San Antonio, TX, USA), followed by washing and incubation with HRP-conjugated secondary antibodies. After a stripping step, the blots were reprobed with anti-β-actin antibody (clone EPR1123Y; EMD Millipore, Billerica, MA, USA). Saturation binding to ARs Because AR mRNA and protein expression experiments in patients with SLE have shown an upregulation of A2AARs compared with control subjects, we carried out experiments to examine saturation binding to this receptor subtype. For these assays, different concentrations of 3H-ZM 241385 (0.01–30 nM) as a radioligand, and cell membranes (60 μg of protein per assay) were incubated for 60 minutes at 4 °C [27]. The radioligand 3H-4-(2-(7-amino-2-(2-furyl)-[1, 2, 4]triazolo[2,3-a] [1, 3, 5] triazin-5-ylamino)ethyl)phenol (specific activity 27 Ci/mmol) was purchased from BIOTREND Chemikalien (Cologne, Germany). Nonspecific binding was determined in the presence of 1 μM 3H-ZM 241385. Bound and free radioactivity were separated by filtering the assay mixture through Whatman GF/B glass fiber filters (GE Healthcare Life Sciences) by using a Brandel cell harvester (Brandel, Gaithersburg, MD, USA) [28]. The filter-bound radioactivity was counted in a 2810TR liquid scintillation counter (PerkinElmer, Waltham, MA, USA). Pro- and anti-inflammatory cytokines release in cultured lymphocytes Isolated lymphocytes from healthy subjects or patients with SLE were suspended at a density of 106 cells/ml in RPMI 1640 medium supplemented with 2 % fetal bovine serum (EuroClone, Pero, Italy) and seeded into 24-well plates. Lymphocytes were incubated for 24 h in the absence or in the presence of an A2AAR agonist, CGS-21680 (2-p-(2-carboxyethyl)phenethylamino-5′-N-ethylcarboxamidoadenosine; 100 nM and 1 μM). In some experiments, cells were treated with a selective A2AAR antagonist, SCH 442416 (2-(2-furanyl)-7-[3-(4-methoxyphenyl)propyl]-7H-pyrazolo[4,3-e] [1, 2, 4] triazolo[1,5-c]pyrimidin-5-amine; 1 μM), 15 minutes before the agonist CGS-21680 to verify the specific involvement of these receptors in cytokine release. CGS-21680 was obtained from Sigma-Aldrich and SCH 442416 was purchased from Tocris Bioscience (Bristol, UK). At the end of incubation, the cell suspension was collected and centrifuged at 1000 × g for 10 minutes at 4 °C. IFN-α, TNF-α, IL-2, IL-6, IL-1β, and IL-10 levels were determined using a specific quantitative sandwich ELISA kit (R&D Systems, Minneapolis, MN, USA) according to the manufacturer’s instructions [25]. Nuclear factor kB activation in human cultured lymphocytes Nuclear extracts from human cultured lymphocytes of the examined patients were obtained by using a nuclear extract kit (Active Motif, Carlsbad, CA, USA) following the manufacturer’s instructions. Nuclear factor (NF)-kB subunit p65 activation was evaluated in lymphocyte nuclear extracts by using the TransAM NF-kB kit (Active Motif). The primary antibody against NF-kB recognized an epitope in the subunits that is accessible only when it is activated and bound to its DNA target. The reaction was developed with streptavidin-HRP, and optical density was read by spectrophotometry at 450-nm wavelength [27]. Data and statistical analysis Using dissociation equilibrium constants for saturation binding, affinity or Kd values, and the maximum densities of specific binding sites, Bmax values were calculated for a system of one- or two-binding-site populations by nonlinear curve fitting [28]. All experimental data are reported as mean ± SEM of independent experiments as indicated in the figure legends. Statistical analysis of the data was performed by using Student’s t test or one-way analysis of variance (ANOVA) followed by Dunnett’s test. The analysis was carried out using the GraphPad Prism 5.0 statistical software package (GraphPad Software, La Jolla, CA, USA), and differences were considered statistically significant with a p value less than 0.01. Results Clinical characteristics A total of 80 patients with SLE (71 women and 9 men) with a mean ± SD age of 44 ± 11.9 years, disease duration of 139 ± 100 months, and SLEDAI-2 K score of 4 ± 4.3 were studied. In addition, 80 healthy subjects matched for age and sex ratio were enrolled. Demographic, clinical, and pharmacological treatments of the study subjects are reported in Table 1.Table 1 Clinical and demographic features of the study subjects, as well as pharmacological treatments in patients with systemic lupus erythematosus Patients with SLE (n = 80), n (%) Clinical parameters  Female/male 71/9  Age, years, mean ± SD 44 ± 11.9  Disease duration, monthsa 139 ± 100  SLEDAI-2 K score, mean ± SD 4 ± 4.3  SDI, mean ± SD 0.8 ± 1.2 Disease activity patterns  CQD 46 (57.5 %)  CAD 18 (22.5 %)  MDA 14 (17.5 %)  RRD 2 (2.5 %) Serological parameters  aPL (aCL, β2-GPI, and/or LA) 39 (48.75 %)  ENA 48 (60 %)  Hypocomplementemia 53 (62.2 %)  Anti-dsDNA antibody, ongoing/previous 41 (51.2 %)/17 (21.1 %) Treatments  Corticosteroids, 2.5 up to 12.5 mg/day 67 (83.7 %)  Hydroxychloroquine, 200 mg/day 48 (60 %) Ongoing immunosuppressant therapy 25 (31.2 %)  Mycophenolate mofetil 11 (13.7 %)  Cyclosporine A 4 (5 %)  Azathioprine 5 (6.2 %)  Methotrexate, 10–15 mg/week 3 (3.7 %)  Thalidomide 1 (1.2 %)  IVIg 1 (1.2 %)  PEX 1 (1.2 %) Anticoagulants 11 (13.7) Antiaggregant 33 (41.25) Abbreviations: SLEDAI-2 K Systemic Lupus Erythematosus Disease Activity Index 2000, SDI Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index, CQD clinical quiescent disease, MDA minimal disease activity, CAD chronic active disease, RRD relapsing-remitting disease, aPL antiphospholipid antibodies, aCL anticardiolipin antibodies, LA lupus anticoagulant, β 2 -GPI β2-glycoprotein I antibodies, ENA extractable nuclear antigen antibodies, PEX plasma exchange, IVIg intravenous immunoglobulin, anti-dsDNA anti-double-stranded DNA aDisease duration at the time of sample collection A2AAR mRNA and protein expression are upregulated in lymphocytes of patients with SLE AR mRNA and protein expression were evaluated in lymphocytes of patients with SLE in comparison with those of healthy subjects by means of quantitative RT-PCR assay and Western blot analysis, respectively. Figure 1a reports the relative A1, A2A, A2B, and A3 AR mRNA levels determined by RT-PCR in human lymphocytes of healthy subjects and patients with SLE. Among these receptors, only A2AAR mRNA expression was significantly increased in patients with SLE respect to control subjects. Western blot and densitometric analysis indicated a significant increase in A2AAR protein expression in lymphocytes of patients with SLE compared with those of healthy subjects, while no differences were found in A1, A2B, or A3 ARs (Fig. 1b, c).Fig. 1 Messenger RNA (mRNA) and protein expression of adenosine receptors (ARs) in human lymphocytes from patients with systemic lupus erythematosus (SLE) and healthy subjects. a Relative AR mRNA levels were determined by real-time reverse transcriptase-polymerase chain reaction. Experiments were performed in duplicate with lymphocytes obtained from individual patients with SLE (n = 80) and healthy subjects (n = 80), and data are shown as mean ± SEM. b Western blot analysis showing immunoblot signals of ARs in one patient with SLE and one healthy subject, representative of blots obtained with lymphocytes from 80 patients with SLE and 80 healthy control subjects. β-actin was used as a loading control. c Densitometric analysis of AR expression in human lymphocytes from patients with SLE (n = 80) and healthy subjects (n = 80) indicated as a ratio of β-actin (loading control). Data are expressed as the mean ± SEM of densitometric analysis results obtained from the indicated number of subjects. * p < 0.01 versus control group by one-way analysis of variance with Dunnett’s test Alteration of A2AAR affinity and density in lymphocytes of patients with SLE Figure 2a and b show the saturation curves and Scatchard plots of [3H]-ZM 241385 in human lymphocytes, confirming the upregulation of A2AARs in patients with SLE compared with healthy subjects. The affinity of the radioligand [3H]-ZM 241385 for A2AAR (expressed as Kd, nanomoles) was decreased in lymphocytes from patients with SLE compared with that of the control group. Interestingly, the A2AAR density, expressed as a Bmax value, significantly increased in patients with SLE compared with healthy subjects, reaching a 2.3-fold increment (Fig. 2a, b).Fig. 2 Adenosine 2A receptors (A2AARs) are upregulated in lymphocytes from patients with systemic lupus erythematosus (SLE). a Saturation curves and b Scatchard plot showing the binding of 3H-ZM 241385 to A2AARs in lymphocyte membranes derived from 80 healthy control subjects (solid circles) and 80 patients with SLE (solid squares) are also shown. Saturation binding experiments were performed as described in the Methods section. Data in the saturation curves are expressed as the mean ± SEM of results pooled from one experiment performed in duplicate for the indicated number of subjects. c Linear regression analysis between Systemic Lupus Erythematosus Disease Activity Index 2000 (SLEDAI-2 K) score and maximum number of A2AARs (Bmax) in lymphocytes (n = 80, r = 0.68, p < 0.0001 by using Pearson’s or Spearman’s correlation (Pearson’s r = −0.68, Spearman’s r = −0.75) Clinical correlations An inverse correlation was found between A2AAR density (expressed as Bmax values in femtomoles per milligram of protein), SLEDAI-2 K measured at the time of blood sampling (Fig. 2c), and disease activity through time evaluated according to the course patterns (CQD versus CAD; p < 0.0001). In addition, A2AAR density inversely correlated with serositis (p = 0.0043), hypocomplementemia (p = 0.0005), and anti-dsDNA positivity (Table 2). Regarding treatments, modulation of A2AAR density was found among corticosteroid users (p = 0.0078). With regard to A2AAR affinity, only one significant correlation was found with anti-dsDNA-positive patients (p = 0.008) (Table 2).Table 2 Clinical, serological, and pharmacological treatments in patients with systemic lupus erythematosus: correlation with adenosine A2A receptor affinity and density Patients (n) K d (nM) p Value Bmax (fmol/mg of protein) p Value Disease activity patterns  CQD/CAD 46/18 1.96 ± 1.19/1.77 ± 0.89 NS 141.6 ± 48.9/76 ± 28.6 < 0.0001 Organ involvement  Renal, yes/no 13/67 1.33 ± 0.59/1.94 ± 1.15 0.06 114 ± 45.05/120.8 ± 52.3 NS  Neuropsychiatric, yes/no 20/60 1.73 ± 0.81/1.88 ± 1.18 NS 103.85 ± 48.2/125 ± 51.2 NS  Articular, yes/no 46/34 1.79 ± 1.22/1.9 ± 0.9 NS 110.85 ± 49.3/131.8 ± 51.5 0.06  Cutaneous, yes/no 25/55 1.46 ± 0.71/1.89 ± 1.2 NS 105.60 ± 43.9/123.8 ± 53.7 NS  Hematological, yes/no 41/39 1.70 ± 0.89/1.98 ± 1.28 NS 116.85 ± 47.8/123.6 ± 54.5 NS  Serositis, yes/no 21/59 1.60 ± 1.3/1.9 ± 0.9 NS 92.86 ± 38.7/129.2 ± 51.6 0.0043 Serological parameters  aCL, yes/no 22/58 1.91 ± 1.01/1.81 ± 1.14 NS 134.04 ± 52.55/114.27 ± 47.3 NS  Anti-β2-GPI, yes/no 9/71 1.59 ± 0.46/1.87 ± 1.15 NS 116.11 ± 53.15/120.2 ± 51.1 NS  LA, yes/no 28/52 1.92 ± 1.20/1.8 ± 1.05 NS 109.03 ± 51.60/125.5 ± 50.2 NS  ENA, yes/no 58/22 1.92 ± 1.15/1.77 ± 1.07 NS 114,5 ± 47.34/123.1 ± 53.52 NS  Hypocomplementemia, yes/no 53/27 1.57 ± 0.78/2.15 ± 1.32 NS 97.88 ± 39.25/145.08 ± 51.78 < 0.0001  Anti-dsDNA, yes/no 41/39 1.65 ± 1.01/2.31 ± 1.18 0.008 106.26 ± 48.42/153.04 ± 41.76 < 0.0001 Treatments  Corticosteroids, 2.5 up to 12.5 mg/day, yes/no 67/13 1.83 ± 1.17/1.88 ± 0.63 NS 113.12 ± 50.30/153.69 ± 41.47 0.0078  Hydroxychloroquine, 200 mg/day, yes/no 48/32 1.93 ± 1.07/1.69 ± 1.13 NS 126.89 ± 51.09/108.93 ± 49.72 NS  Immunosuppressants or induction therapy, yes/no 25/55 1.98 ± 1.19/1.54 ± 0.8 NS 124.31 ± 48.04/110.15 ± 56.46 NS  Anticoagulants, yes/no 11/69 2.01 ± 1.47/1.81 ± 1.04 NS 128.91 ± 64.01/128.91 ± 64 NS  Antiaggregants, yes/no 33/47 2.03 ± 1.24/1.7 ± 0.97 NS 123.39 ± 50.61117.13 ± 51.66 NS Abbreviations: aCL anticardiolipin antibodies, anti-β 2 -GPI anti-β2-glycoprotein I, CAD chronic active disease, CQD clinical quiescent disease, dsDNA double-stranded DNA, ENA extractable nuclear antigen antibodies, LA lupus anticoagulant, NS not significant Analysis was carried out using unpaired t tests. A2AAR activation reduces proinflammatory cytokine production from lymphocytes To investigate the potential anti-inflammatory role of A2AAR stimulation in SLE, we evaluated the effect of CGS-21680 on the release of some of the most relevant proinflammatory cytokines involved in the pathogenesis of SLE, such as IFN-α, TNF-α, IL-6, IL-1β, and IL-2. In cultured lymphocytes of patients with SLE, a marked release of IFN-α was observed following the incubation of the cells with 0.1 mg/ml lipopolysaccharide (LPS) for 24 h (Fig. 3a). Interestingly, CGS-21680 at concentrations of 100 nM and 1 μM was able to inhibit the LPS-induced IFN-α release in lymphocytes of both patients with SLE and healthy subjects. However, the effect of CGS-21680 was significantly greater in lymphocytes obtained from patients with SLE than in those of healthy subjects (p < 0.0001), most likely due to the upregulation of A2AARs (see Table 3). The inhibitory effect of the A2AAR agonist was counteracted by the selective antagonist SCH 442416 (1 μM), demonstrating the specific A2AAR-mediated response (Fig. 3a). Similar results were obtained when we evaluated the capability of CGS-21680 to inhibit the release of TNF-α induced by LPS (Fig. 3b). Again, the inhibitory effect of the A2AAR agonist was more evident in lymphocytes of patients with SLE than in those of healthy subjects (p < 0.0001) (see Table 3). The anti-inflammatory effect of A2AAR activation induced by CGS-21680 was also confirmed when we analyzed the production of other proinflammatory ILs, such as IL-6 (Fig. 3c), IL-1β (Fig. 3d), and IL-2 (Fig. 4a). As reported in Table 3, a greater inhibitory effect was obtained in lymphocytes of patients with SLE. Moreover, the use of the selective A2AAR antagonist SCH 442416 (1 μM) demonstrated that the effect was mediated by A2AARs.Fig. 3 Adenosine 2A receptor (A2AAR) stimulation inhibits proinflammatory cytokine release. The effect of a well-known A2AAR agonist (CGS-21680; 100 nM and 1 μM) and antagonist (SCH 442416; 1 μM) on (a) interferon (IFN)-α, (b) tumor necrosis factor (TNF)-α, (c) interleukin (IL)-6, and (d) IL-1β release in cultured lymphocytes of patients with systemic lupus erythematosus (SLE) (n = 20) and healthy subjects (n = 20) that were stimulated by 0.1 mg/ml lipopolysaccharide (LPS) as determined by enzyme-linked immunosorbent assay. Data are expressed as the mean ± SEM of three independent experiments performed in triplicate * p < 0.01 versus LPS-treated cells by one-way analysis of variance with Dunnett’s test Table 3 Effect of CGS 21680 in lymphocytes from healthy subjects (n = 20) or SLE patients (n = 20) on different inflammatory mediators Healthy subjects SLE patients Cellular stimulation CGS 21680 (1μM) stimulation % of reduction/fold of increase Cellular stimulation CGS 21680 (1μM) stimulation % of reduction/fold of increase IFN-α 67 ± 4a 37 ± 3a 44.39 ± 1.21d 74 ± 4a 32 ± 2a 56.89 ± 1.52d,* TNF-α 225 ± 12a 102 ± 10a 54.66 ± 1.14d 227 ± 10a 79 ± 5a 64.95 ± 2.33d,* IL-6 305 ± 15a 167 ± 12a 45.38 ± 1.18d 322 ± 13a 105 ± 6a 67.71 ± 1.25d,* IL-1β 294 ± 14a 198 ± 13a 32.64 ± 1.02d 310 ± 11a 140 ± 7a 54.20 ± 2.35d,* IL-2 36 ± 3b 21 ± 2b 41.59 ± 1.08d 42 ± 3b 19 ± 2b 54.59 ± 1.76d,* IL-10 2832 ± 112a 4802 ± 189a 1.70 ± 0.02e 2973 ± 104a 5329 ± 199a 1.79 ± 0.02e,** NF-kB 245 ± 14c 161 ± 13c 34.07 ± 1.08d 266 ± 10c 150 ± 6c 43.69 ± 1.47d,* The data are expressed as mean ± SEM; *p < 0.0001 vs healthy subjects; ** p < 0.01 vs healthy subjects a Cellular stimulation with LPS; values in pg/ml b Cellular stimulation with PMA + ionomycin; values in pg/ml c Cellular stimulation with LPS; values in % of controls d Percentage of reduction e Fold of increase Fig. 4 CGS-21680 effect on interleukin (IL)-2 and IL-10 release and on nuclear factor (NF)-kB activation. Effect of CGS-21680 (100 nM and 1 μM) and SCH 442416 (1 μM) in cultured lymphocytes of patients with systemic lupus erythematosus (SLE) (n = 20) and healthy subjects (n = 20) stimulated by lipopolysaccharide (LPS) (0.1 mg/ml) or phorbol 12-myristate 13-acetate (PMA) (2 ng/ml) and ionomycin (0.2 μM) on (a) IL-2 release, (b) IL-10 release, and (c) NF-kB activation. Data are expressed as the mean ± SEM of three independent experiments performed in triplicate * p < 0.01 versus LPS-treated cells by one-way analysis of variance with Dunnett’s test CGS-21680 increases production of anti-inflammatory cytokine IL-10 in lymphocytes The incubation of lymphocytes with the A2AAR agonist CGS-21680 (1 μM) augmented basal IL-10 release in lymphocytes from healthy subjects and from patients with SLE (Fig. 4b). A more pronounced effect of CGS-21680 was obtained when cells were stimulated with LPS (0.1 mg/ml), although LPS alone did not alter IL-10 production. In the presence of LPS, the effect of CGS-21680 was significantly greater (p < 0.01) in lymphocytes of patients with SLE than in those of healthy subjects (see Table 3). A2AAR activation inhibits LPS-induced NF-kB activation in lymphocytes Many of the anti-inflammatory effects of A2AAR stimulation are mediated by the inhibition of NF-kB activation [25]. To verify if CGS-21680 was able to inhibit NF-kB in lymphocytes from patients with SLE in comparison with those of healthy subjects, the activation of NF-kB p65 subunits following LPS treatment was investigated. As shown in Fig. 4c, the A2AAR agonist determined a marked reduction of LPS-stimulated NF-kB p65 subunit activation in nuclear extract from lymphocytes obtained from patients with SLE and healthy subjects, with a significantly greater effect in the former (see Table 3). The inhibitory effect of CGS-21680 was completely counteracted by the selective A2AAR antagonist SCH 442416 (1 μM). Discussion The primary aim of this study was to investigate the role of ARs in SLE pathogenesis and to assess potential relationships between these receptors and clinical data. Within the complexity of the pathogenic mechanisms of lupus, innate immune responses play a significant role contributing either to tissue injury via release of inflammatory cytokines or to the aberrant hyperactivation of T and B cells, qualified as the most important players leading to autoreactive autoantibody production and resultant end-organ injury [3, 5, 29–31]. The role of the adenosinergic system is attractive for its multifunctionality in this wide spectrum of inflammation-related processes [10, 13, 14] and for its potential engagement in SLE. AR mRNA and protein analysis supported higher A2AAR expression in lymphocytes from patients with SLE with than in those of control subjects, while no changes in A1, A2B, or A3 ARs were found, suggesting a specific A2AAR alteration. Moreover, saturation binding experiments confirmed the upregulation of A2AARs in lymphocytes of patients with SLE. Notably, the highest levels of A2AAR density were tightly correlated with the lowest levels of clinical—namely, clinimetric—indexes and serological parameters (anti-DNA, C3 and C4) of disease activity, suggesting that the endogenous activation of these receptors could lead to mitigation of the disease. This aspect is further supported by the finding of an inverse correlation between CAD progression of the disease and receptor density, suggesting that the mutual modulation of A2AAR expression identifies well a persistent and stable regulation of the inflammatory status. The hypothesis that A2AAR upregulation could represent a compensatory mechanism to better counteract the inflammatory background in SLE is supported by a preclinical study in an MRL/lpr mouse model of lupus nephritis [32] in which the A2AAR mRNA expression in the kidneys of MRL/lpr mice was significantly increased compared with that in control mice. In this study, the treatment with an A2AAR agonist ameliorated the severity of nephritis and renal vasculitis and reduced leukocytic infiltration. Because IFN-α plays a central role in SLE pathogenesis [5], we investigated the anti-inflammatory effect of A2AAR activation on this cytokine in cultured lymphocytes. The effects of the IFN signature on lupus lymphocytes have been studied mainly in the regulatory T-cell subpopulation, where the action of IFN-α diminished their activity [33], while in B cells it stimulated antibody production [34]. We demonstrated that the A2AAR agonist CGS-21680 inhibited IFN-α release in cultured lymphocytes with a greater effect in patients with SLE than in healthy subjects. This observation further supports the competence of A2AAR signaling, as suggested by a previous study [35], to promote peripheral tolerance by generation of regulatory T cells. The reduction of inflammatory response by A2AAR activation was also confirmed when we studied the release of typical proinflammatory cytokines such as TNF-α, IL-6, IL-1β, and IL-2. Furthermore, we found that CGS-21680 mediated a significant increase of anti-inflammatory IL-10, which is an important immunoregulator that supports T-cell differentiation and suppresses proinflammatory cytokines [36]. It is well known that activation of NF-kB pathways leads to enhanced B-cell survival and T-cell activation and maturation [37]. Moreover, NF-kB positively regulates gene-encoding cytokines and other inflammatory factors, suggesting that this transcription factor could be one of the master regulators of inflammatory responses. Thus, the inhibition of NF-kB by CGS-21680 could explain the reduction of LPS-stimulated proinflammatory cytokines in cultured lymphocytes of patients with SLE. Conclusions Taken together, our data demonstrate the presence of A2AAR upregulation in patients with SLE and a significant inverse correlation of A2AAR density with SLEDAI-2 K score and CAD. The anti-inflammatory response of A2AARs opens up a new perspective on the translational role of the A2AAR agonists in the pharmacological treatment of SLE, highlighting their therapeutic potential in the management of this disorder. Abbreviations aCLAnticardiolipin antibodies ANAAntinuclear antibody ANOVAAnalysis of variance anti-β2-GPIAnti-β2-glycoprotein I anti-RNPAntiribonucloprotein anti-SSAAnti-Sjögren’s-syndrome-related antigen A anti-SSBAnti-Sjögren’s-syndrome-related antigen B anti-Smanti-Smith aPLAntiphospholipid antibodies ARAdenosine receptor C3 and C4Complement components C3 and C4 CADChronic active disease CQDClinical quiescent disease DCDendritic cell dsDNADouble-stranded DNA ELISAEnzyme-linked immunosorbent assay ENAExtractable nuclear antigen antibodies ICImmune complex IFN-αInterferon-α ILInterleukin IVIgIntravenous immunoglobulin LALupus anticoagulant LPSLipopolysaccharide MDAMinimal disease activity mRNAMessenger RNA NF-kBNuclear factor kB NSnot significant PBSPhosphate-buffered saline PEXPlasma exchange PMAphorbol 12-myristate 13-acetate RRDRelapsing-remitting disease RT-PCRReverse transcriptase-polymerase chain reaction SDISystemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index SLESystemic lupus erythematosus SLEDAI-2 KSystemic Lupus Erythematosus Disease Activity Index 2000 TNF-αtumor necrosis factor-α Acknowledgements The authors are grateful to the study participants. Funding Internal funding. Availability of data and materials For access to study data, please contact the corresponding author. Authors’ contributions AB designed and performed the study, analyzed and interpreted data, and drafted the manuscript. KV, MG, and PAB performed the study, interpreted data, and revised the manuscript. MP interpreted and discussed the data and critically revised the manuscript. FV and AR analyzed and interpreted data, and critically revised the manuscript. All authors read and approved the final manuscript. Authors’ information Not applicable. Competing interests The authors declare that they have no competing interests. Consent for publication The authors received informed consent from the participants for data publication in an anonymized form. Ethics approval and consent to participate The study was approved by the local ethics committee of the University Hospital of Ferrara (Italy), and written informed consent was obtained from each participant in accordance with the principles outlined in the Declaration of Helsinki. ==== Refs References 1. Rahman A Isenberg DA Systemic lupus erythematosus N Engl J Med. 2008 358 929 39 10.1056/NEJMra071297 18305268 2. Gualtierotti R Biggioggero M Penatti AE Meroni PL Updating on the pathogenesis of systemic lupus erythematosus Autoimmun Rev. 2010 10 3 7 10.1016/j.autrev.2010.09.007 20863908 3. Celhar T Magalhães R Fairhurst AM TLR7 and TLR9 in SLE: when sensing self goes wrong Immunol Res. 2012 53 58 77 10.1007/s12026-012-8270-1 22434514 4. Batteux F Palmer P Daëron M Weill B Lebon P FCγRII (CD32)-dependent induction of interferon-α by serum from patients with lupus erythematosus Eur Cytokine Netw. 1999 10 509 14 10586117 5. Obermoser G Pascual V The interferon-α signature of systemic lupus erythematosus Lupus. 2010 19 1012 9 10.1177/0961203310371161 20693194 6. Iwata Y Furuichi K Kaneko S Wada T The role of cytokine in the lupus nephritis J Biomed Biotechnol. 2011 2011 594809 10.1155/2011/594809 22028590 7. Malek TR The biology of interleukin-2 Annu Rev Immunol. 2008 26 453 79 10.1146/annurev.immunol.26.021607.090357 18062768 8. Yuan W DiMartino SJ Redecha PB Ivashkiv LB Salmon JE Systemic lupus erythematosus monocytes are less responsive to interleukin-10 in the presence of immune complexes Arthritis Rheum. 2011 63 212 8 10.1002/art.30083 20954190 9. Burnstock G Introductory overview of purinergic signalling Front Biosci Elite Ed 2011 3 896 900 10.2741/e298 21622101 10. Fredholm BB Chern Y Franco R Sitkovsky M Aspects of the general biology of adenosine A2A signaling Prog Neurobiol. 2007 83 263 76 10.1016/j.pneurobio.2007.07.005 17804147 11. Fredholm BB IJzerman AP Jacobson KA Linden J Müller CE International Union of Basic and Clinical Pharmacology. LXXXI. Nomenclature and classification of adenosine receptors—an update Pharmacol Rev 2011 63 1 34 10.1124/pr.110.003285 21303899 12. Burnstock G Purine and pyrimidine receptors Cell Mol Life Sci. 2007 64 1471 83 10.1007/s00018-007-6497-0 17375261 13. Gessi S Merighi S Fazzi D Stefanelli A Varani K Borea PA Adenosine receptor targeting in health and disease Expert Opin Investig Drugs. 2011 20 1591 609 10.1517/13543784.2011.627853 22017198 14. Antonioli L Csóka B Fornai M Colucci R Kókai E Blandizzi C Adenosine and inflammation: what’s new on the horizon? Drug Discov Today. 2014 19 1051 68 10.1016/j.drudis.2014.02.010 24607729 15. Di Virgilio F Vuerich M Purinergic signaling in the immune system Auton Neurosci Basic Clin. 2015 191 117 23 10.1016/j.autneu.2015.04.011 16. Haskó G Kuhel DG Chen JF Schwarzschild MA Deitch EA Mabley JG Adenosine inhibits IL-12 and TNF-α production via adenosine A2a receptor-dependent and independent mechanisms FASEB J. 2000 14 2065 74 10.1096/fj.99-0508com 11023991 17. Schnurr M Toy T Shin A Hartmann G Rothenfusser S Soellner J Role of adenosine receptors in regulating chemotaxis and cytokine production of plasmacytoid dendritic cells Blood. 2004 103 1391 7 10.1182/blood-2003-06-1959 14551144 18. Lappas CM Rieger JM Linden J A2A adenosine receptor induction inhibits IFN-γ production in murine CD4+ T cells J Immunol. 2005 174 1073 80 10.4049/jimmunol.174.2.1073 15634932 19. Moore CC Martin EN Lee GH Obrig T Linden J Scheld WM An A2A adenosine receptor agonist, ATL313, reduces inflammation and improves survival in murine sepsis models BMC Infect Dis. 2008 8 141 10.1186/1471-2334-8-141 18937852 20. Hochberg MC Updating the American College of Rheumatology revised criteria for the classification of systemic lupus erythematosus Arthritis Rheum. 1997 40 1725 10.1002/art.1780400928 9324032 21. Gladman DD Ibañez D Urowitz MB Systemic Lupus Erythematosus Disease Activity Index 2000 J Rheumatol. 2002 29 288 91 11838846 22. Zen M Bassi N Nalotto L Canova M Bettio S Gatto M Disease activity patterns in a monocentric cohort of SLE patients: a seven-year follow-up study Clin Exp Rheumatol. 2012 30 856 63 22765883 23. Harris EN Gharavi AE Patel SP Hughes GR Evaluation of the anti-cardiolipin antibody test: report of an international workshop held 4 April 1986 Clin Exp Immunol. 1987 68 215 22 3652514 24. Horbach DA van Oort E Donders RC Derksen RH de Groot PG Lupus anticoagulant is the strongest risk factor for both venous and arterial thrombosis in patients with systemic lupus erythematosus: comparison between different assays for the detection of antiphospholipid antibodies Thromb Haemost. 1996 76 916 24 8972011 25. Varani K Padovan M Vincenzi F Targa M Trotta F Govoni M A2A and A3 adenosine receptor expression in rheumatoid arthritis: upregulation, inverse correlation with disease activity score and suppression of inflammatory cytokine and metalloproteinase release Arthritis Res Ther. 2011 13 R197 10.1186/ar3527 22146575 26. Varani K Caramori G Vincenzi F Adcock I Casolari P Leung E Alteration of adenosine receptors in patients with chronic obstructive pulmonary disease Am J Respir Crit Care Med. 2006 173 398 406 10.1164/rccm.200506-869OC 16322645 27. Varani K Massara A Vincenzi F Tosi A Padovan M Trotta F Normalization of A2A and A3 adenosine receptor up-regulation in rheumatoid arthritis patients by treatment with anti-tumor necrosis factor α but not methotrexate Arthritis Rheum. 2009 60 2880 91 10.1002/art.24794 19790066 28. Varani K Vincenzi F Tosi A Gessi S Casetta I Granieri G A2A adenosine receptor overexpression and functionality, as well as TNF-α levels, correlate with motor symptoms in Parkinson’s disease FASEB J. 2010 24 587 98 10.1096/fj.09-141044 19776336 29. Manderson AP Botto M Walport MJ The role of complement in the development of systemic lupus erythematosus Annu Rev Immunol. 2004 22 431 56 10.1146/annurev.immunol.22.012703.104549 15032584 30. Shlomchik MJ Craft JE Mamula MJ From T to B and back again: positive feedback in systemic autoimmune disease Nat Rev Immunol. 2001 1 147 53 10.1038/35100573 11905822 31. Robson MG Walport MJ Pathogenesis of systemic lupus erythematosus (SLE) Clin Exp Allergy. 2001 31 678 85 10.1046/j.1365-2222.2001.01147.x 11422126 32. Zhang L Yang N Wang S Huang B Li F Tan H Adenosine 2A receptor is protective against renal injury in MRL/lpr mice Lupus. 2011 20 667 77 10.1177/0961203310393262 21183557 33. Yan B Ye S Chen G Kuang M Shen N Chen S Dysfunctional CD4+ , CD25+ regulatory T cells in untreated active systemic lupus erythematosus secondary to interferon-α-producing antigen-presenting cells Arthritis Rheum. 2008 58 801 12 10.1002/art.23268 18311820 34. Oka H Hirohata S Regulation of human B cell responsiveness by interferon-α: interferon-α-mediated suppression of B cell function is reversed through direct interactions between monocytes and B cells Cell Immunol. 1993 146 238 48 10.1006/cimm.1993.1023 8174169 35. Zarek PE Huang CT Lutz ER Kowalski J Horton MR Linden J A2A receptor signaling promotes peripheral tolerance by inducing T-cell anergy and the generation of adaptive regulatory T cells Blood. 2008 111 251 9 10.1182/blood-2007-03-081646 17909080 36. Horwitz DA Gray JD Behrendsen SC Kubin M Rengaraju M Ohtsuka K Decreased production of interleukin-12 and other Th1-type cytokines in patients with recent-onset systemic lupus erythematosus Arthritis Rheum. 1998 41 838 44 10.1002/1529-0131(199805)41:5<838::AID-ART10>3.0.CO;2-S 9588735 37. Chuang HC Lan JL Chen DY Yang CY Chen YM Li JP The kinase GLK controls autoimmunity and NF-kB signaling by activating the kinase PKC-θ in T cells Nat Immunol. 2011 12 1113 8 10.1038/ni.2121 21983831

PMC005xxxxxx/PMC5002092.txt

==== Front BMC Public HealthBMC Public HealthBMC Public Health1471-2458BioMed Central London 358110.1186/s12889-016-3581-0Research ArticlePre-vaccine era cervical human papillomavirus infection among screening population of women in west Austria Borena Wegene +43.512.9003.71737wegene.borena@i-med.ac.at 1Grünberger Margarethe 1Widschwendter Andreas Andreas.Widschwendter@i-med.ac.at 2Kraxner Karl Heinz heinz.kraxner@mynet.at 3Marth Elisabeth christian.marth@i-med.ac.at 4Mayr Peter office@dr-petermayr-gyn.at 5Meier Joerg joerg.meier@medway.at 6Ruth Norman dr.ruth@snw.at 7Guerrero Aida Tort Aida.Guerrero@i-med.ac.at 1Marth Christian christian.marth@i-med.ac.at 2Holm-von Laer Dorothee Dorothee.von-Laer@i-med.ac.at 11 Division of Virology, Medical University of Innsbruck, Schoepfstrasse 41, 6020 Innsbruck, Austria 2 Department of Gynaecology and Obstetrics, Medical University of Innsbruck, Innsbruck, Austria 3 Landeck, Austria 4 Innsbruck, Austria 5 Kufstein, Austria 6 Imst, Austria 7 Woergl, Austria 26 8 2016 26 8 2016 2016 16 1 8891 4 2016 24 8 2016 © The Author(s). 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Background In order to evaluate the newly implemented gender-neutral HPV vaccination program, knowledge on the pre-vaccine prevalence of HPV infection is of paramount importance. Data on HPV infection among the women with no known previous cytological abnormalities are inexistent in Austria. This study presents data on the prevalence and distribution of HPV genotypes among women with no known cytological abnormalities in west Austria. Methods Women between 18 and 65 years of age attending annual cervical cancer screening examinations were included in the study. Data on socio-demographic and reproductive factors were collected using structured questionnaires. Corresponding cervical swab samples were tested for the presence of HPV DNA and were genotyped. Questionnaire data and HPV status were linked with the corresponding cytological findings. Results A total of 542 women were included in the study. The mean age of the study participants was 35.9 (SD = 11.5). The prevalence of HPV infection was 20.5 %. HPV 16 (6.5 %), HPV 33 (3.3 %) and HPV 31 (3.0 %) were the dominant genotypes detected. Multivariate analysis showed that women younger than 30 years of age, smokers, women with a higher number of lifetime sexual partners and those living in the eastern districts of the study region were at significantly higher risk of HPV infection. Conclusions With this study we present the first data on the prevalence of cervical HPV genotypes among a screening population in Austria. The results not only fill the missing information on HPV infection in this group of women in the country, they also provide baseline data for a future evaluation of the impact of the Austrian gender-neutral HPV immunization program. Moreover, our finding of higher HPV prevalence in the eastern compared to the western district of the study region may – at least partly – explain the east–west gradient in the standardized incidence rate of cervical cancer in the region. Electronic supplementary material The online version of this article (doi:10.1186/s12889-016-3581-0) contains supplementary material, which is available to authorized users. Keywords HPV epidemiologyWomens’ healthCervical cancerHPV vaccineTiroler Wissenschaftsfonds (to W.B.)GZ:UNI-0404/1446).Borena Wegene MFF Tirol262Borena Wegene issue-copyright-statement© The Author(s) 2016 ==== Body Background Genital human papillomavirus (HPV) infection is the most common sexually transmitted viral infection [1]. It is mostly transient resulting in no cervical lesions or leading to low-grade lesions that often regress spontaneously. However, a non-negligible proportion of the infections persists and results in subsequent pre-cancerous and cancerous lesions [2, 3]. For the prevention of cervical and other genital neoplasia, three vaccines – Gardasil® (quadrivalent vaccine against HPV types 6, 11, 16 and 18), Gardasil 9® (nonavalent vaccine against HPV types 6, 11, 16, 18, 31, 33, 45, 52 and 58) and Cervarix® (bivalent vaccine against HPV types 16 and 18) – have been approved for use [4–7]. Developed nations like Australia and several countries in Europe and North America have integrated HPV immunisation in their vaccination programs soon after the approval. Accordingly, in these countries significant reductions have been observed in many high-risk HPV (hrHPV) infections as well as in HPV-associated warts and pre-cancerous lesion [8–10]. In Austria these vaccines have been recommended and available for use since their approval [11, 12]. However, it is only since the beginning of 2014 that the country integrated HPV vaccine in the free of charge immunization program [12]. Starting from the autumn of 2014, 4th grade elementary school girls and boys receive two doses of Gardasil (0–6 months) as part of school-based immunization program. In order to assess the long-term impact of this program, pre-vaccine data on the prevalence of HPV infection and genotype distribution are of paramount importance [11]. Data on the prevalence of HPV infection is scarce in Austria. Until the end of 2013 the only data available were those on HPV prevalence in cervical cancer samples [13, 14]. L Rössler et al. recently published data on the prevalence of HPV among Austrian women with high- grade intraepithelial lesions [15]. However, to date, there are no data on the epidemiology of HPV infection among women without any known cytological abnormalities (Additional file 1: Figure S1). One of the early parameters in evaluating the success of HPV immunisation programs is the comparison of HPV prevalence in the pre- and post-HPV-vaccine era. The main aim of this study is to present data on the distribution of HPV genotypes among screening population of women in west Austria. With this study, we also aim to address the role of HPV infection in an interesting difference in the cervical cancer incidence observed between eastern (standardized incidence rate (SIR) in the range of 1.2 to 1.5) and western (SIR in the range of 0.6 to 0.9) districts of the study region [16]. To our knowledge this is the first report on the epidemiology of cervical HPV infection among screening population of women in Austria. Methods Study population The study population included sexually active women aged 18–65 years undergoing routine gynaecological screening at five randomly selected gynaecological clinics in west Austria (Tyrol) between March 2013 and February 2015. Screening for cervical cancer in Austria is opportunistic in nature but is provided free-of-charge at a yearly basis. Women who have consented to participate in the study filled out questionnaires on socio-demographic, reproductive and sexual characteristics. The questionnaires were then sent back to our institute (Division of Virology of the Medical University of Innsbruck) with the corresponding cervical samples. For the ascertainment of persistent HPV infection, those women who tested positive for one or more of the known hrHPV types were invited for a follow up HPV testing approximately a year after the baseline. The samples One of the two swabs taken from the patient was sent to our institute, the other one was sent to a pathology laboratory (Dr. P. O.) for cytological (PAP) examination. Cervical swabs were taken with Abbott Cervi-Collect specimen collection kit. DNA extraction (Biomerieux EasyMag 2.0) took place within 2 days of arrival of samples at our institute. The extracted DNA samples were stored at −20 °C until HPV testing. Detection and Genotyping of HPV For the detection and genotyping of HPV, we used a two-step testing procedure. The first test is a real-time PCR that identifies the presence or absence of HPV DNA (HPV RealQuality, AB Analitica, Padova, Italy). The second step involves genotyping of all positive samples using a reverse line blot hybridization system (AmpliQuality HPV-TYPE EXPRESS, AB Analitica, Padova, Italy). This genotyping kit was able to identify 40 different HPV types, including all high-risk, probable high-risk and low-risk types. Membrane stripes coated with genotype specific DNA probes are used to identify the different genotypes through hybridization of denatured PCR product. Multiple infections were easily detected. The analysis includes dUTP/UNG system for the prevention of contamination due to carry-over. Cytology Cytological examinations were interpreted according to Munich Nomenclature for the Cytological Diagnosis of Cervical Pap smears [17]. In our study cytological findings were classified as abnormal (PAP III+) if the pathologist reported PAP III (unclear finding corresponding in the Bethesda classification system to ASCUS), PAP IIID (cells of mild or moderate dysplasia corresponding in the Bethesda classification system to LSIL or HSIL respectively), PAP IVA (cells of severe dysplasia or carcinoma in situ corresponding in the Bethesda classification system to HSIL or AIS, respectively), PAP IVB (cells of invasive carcinoma not safely excluded, corresponding in the Bethesda classification system to HSIL or AIS with features suspicious for invasion, respectively) or PAP V (cells of invasive cervical carcinoma). Data analysis HPV prevalence was computed in crude and age standardized manner. Chi square or Fischer’s exact test (two-tailed) were used for analysis of distribution of several hrHPV genotypes. Logistic regression model was used to compute odds ratios for HPV positivity across several variables including smoking status, educational level, marital status, number of life time sexual partners and age at first sexual contact with corresponding 95 % confidence intervals. P-values <0.05 were considered significant. Prevalence ratio of HPV infection was also computed across Tyrolean districts. Statistical analyses were performed in SPSS (Version 20.0. Armonk, NY: IBM Corp.). Results Baseline characteristics Baseline characteristics of the study participants are presented in Table 1. A total of 542 women were included in this study. The mean age of the study participants was 35.9 (SD = 11.5) years. One third of the women were current smokers and 27 % were overweight. Mean age at first sexual contact was 16.8 years (SD = 2.2). A quarter of the study participants reported to have had more than five life-time sexual partners. Two hundred thirteen (39.3 %) participants in this study were from eastern Tyrolean districts (Kufstein, Kitzbühel) whereas 205 (37.8 %) represent two western districts (Imst and Landeck). Just over a fifth of the participants were from the city of Innsbruck (IBK) or its suburbs (IL). Additional file 2: Figure S2 presents a map of the studied region.Table 1 Baseline characteristics of study participants Total (n = 542) Age, years  Mean (SD) 35.9 (11.5)  Median 34.5 Participants per district, n (%)  Oberland 205 (37.8)  Unterland 213 (39.3)  IBK 68 (12.5)  IL and others 55 (10.1) BMI, kg/m2  Mean (SD) 23.4 (4.0) Smoking status, n (%)  Current smoker 175 (32.3)  Non-smokers 364 (67.1) Life time sexual Partners, n (%)   ≤ 5 340 (62.7)  6–10 95 (17.5)   ≥ 11 41 (7.6) HPV vaccinated, n (%)  At least one vaccine 20 (3.7) Reasons for not being vaccinated, n (%)  Did not know 201 (37.1)  Expensive 42 (7.8)  Fear of side effect 78 (14.4)  Other reasons 155 (28.6)  No response 86 (15.9) If vaccine available for free, n (%)  Will get vaccinated 179 (33)  Will not get vaccinated 72 (13.3)  Do not know 233 (43)  No response 58 (10.7) SD standard deviation, BMI body mass index Oberland western districts (Imst, Landeck) Unterland Eastern districts (Kufstein, Kitzbuehel), IBK Innsbruck, IL suburbs of Innsbruck HPV immunization Immunisation coverage among the studied population was very low (3.7 %). The attitude of the study participants towards HPV vaccine is presented in Table 1. Not knowing about the vaccine and/or not being informed about it was the most frequent reason given by the study participants for not being vaccinated. When asked if they would get vaccinated provided the vaccine was available for free, the majority of the study participants showed undecidedness and a minority (13 %) declined clearly. HPV prevalence Figure 1 shows the distribution of all detected genotypes among HPV positive women. A total of 111 (20.5 %) participants were positive for one or more of HPV genotypes. Among the high risk types HPV 16 is the most commonly detected genotype followed by HPV 33 and HPV 31. Out of the low risk genotypes HPV 53 and HPV 54 are the most commonly detected types followed by HPVs 82 and 90. Only three participants (2.7 % of HPV positive participants) were positive for HPV 6 and none had HPV 11.Fig. 1 Distribution (%) of all detected HPV genotypes among screening population of women in west Austria Fifty women (45.5 %) had HPV infection with multiple genotypes whereas the slight majority presented with single-genotype infections (Additional file 3: Figure S3). The majority of the participants (91.8 %) are positive for one of the genotypes classified as definitive, probable or possible carcinogenic HPV types (IARC classification) (Additional file 4: Figure S4) whereas only nine of HPV positive participants (8.2 %) had infections with only low risk genotypes [18]. A total of 73 women (65.7 % of HPV positive women) are positive for one or more of the high-risk genotypes that are covered in the nonavalent HPV vaccine (HPV-types 16, 18, 31, 33, 45, 52 and 58). Table 2 characterises the prevalence of any HPV and selected high-risk types by socio-domographic and behavioural data. HPV infection was significantly higher among individuals younger than 30 years of age. Smoking, being single and/or divorced, having several lifetime sexual partners and starting sexual contact at a younger age were significantly associated with higher cervical HPV DNA detection rate. On the contrary higher body mass index (BMI) was not associated with increased risk of HPV infection showing rather a protective effect with a borderline statistical significance. Analysis across Tyrolean districts showed that HPV prevalence was higher among women from the eastern districts compared to those from the west. The significant association disappeared when analysis was adjusted for age, indicating that the observed difference may be due to differences in the age distribution of the study participants in these two regions. A further analysis was done among women above the age of 30 years (n = 342) as the clinical and diagnostic relevance of HPV infection is higher in women beyond the age of 30 (Table 3). In this sub-population we found no significant difference in the age distribution across the districts (P = 0.53). Multivariable analysis of HPV prevalence in this age group showed still significantly higher HPV prevalence in the eastern compared to the western districts.Table 2 Prevalence of HPV Infection by demographic and behavioural characteristics among non-vaccinated women (n = 522)a Variables Variable categories Any HPV type HrHPVb (n = 111) (n = 73) OR (95 % CI) OR (95 % CI) Age <30 years 1 1 ≥30 years 0.38 (0.25–0.60)* 0.34 (0.21–0.57)* Marital status Married/Partnershaft 1 1 Single/divorced 1.87 (1.17–2.99)*,e 1.16 (0.66–2.10) Educational status Basic schooling 1 1 High school and more 0.81 (0.53–1.25) 0.83 (0.50–1.37) Districts Oberlandc 1 1 Unterlandd 1.76 (1.06–2.92)* 1.57 (0.88–2.79) Smoking non-smoker 1 1 smoker 4.10 (2.6–6.41)* 3.16 (1.89–5.26)* BMI <25 1 1 ≥25 0.55 (0.32–0.95)* 0.64 (0.34–1.19) Age at first sexual contact <16 years 1 1 ≥16 years 0.44 (0.27–0.72)* 0.61 (0.35–1.08) LSP <6 1 1 ≥6 2.40 (1.39–3.62)* 1.95 (1.13–3.36)* BMI body mass index, LSP number of life time sexual partners, OR odds ratio, CI confidence interval *p-value = statistically significant aonly non-vaccinated women bHPVs 16, 18, 31, 33, 45, 52 58 (all the HrHPV types included in nonavalen HPV vaccine) cwestern districts (Imst, Landeck) dEastern districts (Kufstein, Kitzbuehel) esignificant for study participants above 30 years of age Table 3 Multivariable adjusted prevalence of HPV Infection by demographic and behavioural characteristics among women above and below 30 years of agea Women ≥30 years Women < 30 years Variables Variable categories Any HPV type HrHPVb Any HPV type HrHPVb (n = 49) (n = 31) (n = 55) (n = 41) OR (95 % CI) OR (95 % CI) OR (95 % CI)a OR (95 % CI) Marital status Married/Partnershaft 1 1 1 1 Single/divorced 1.89 (0.87–4.14) 1.28 (0.49–3.32) 1.29 (0.58–2.85) 0.82 (0.35–1.93) Educational status Basic schooling 1 1 1 1 High school and more 0.96 (0.47–1.97) 1.60 (0.67–3.82) 0.52 (0.24–1.17) 0.39 (0.17–0.94)* Districts Oberlandc 1 1 1 1 Unterlandd 2.43 (1.06–5.60)* 2.66 (0.98–7.24)e 0.79 (0.32–1.96) 0.98 (0.38–2.50) Smoking non-smoker 1 1 1 1 smoker 2.64 (1.30–5.37)* 2.09 (0.90–4.84) 3.34 (1.53–7.29)* 3.07 (1.33–7.09)* BMI <25 1 1 1 1 ≥25 0.39 (0.16–0.96)* 0.38 (0.13–1.17) 1.30 (0.51–3.30) 1.98 (0.84–4.67) Age at first sexual contact <16 years 1 1 1 1 ≥16 years 1.06 (0.37–3.00) 2.43 (0.51–11.5) 0.88 (0.40–1.94) 1.18 (0.50–2.77) LSP <6 1 1 1 1 ≥6 2.23 (1.06–4.68)* 2.11 (0.88–5.06) 0.86 (0.39–1.88) 0.98 (0.43–2.71) BMI body mass index, LSP number of life time sexual partners OR odds ratio, CI confidence interval, LSP lifetime sexual partner *p-value = statistically significant aanalysis using only non-vaccinated women, analysis adjusted for age, age at first sexual contact, LSP and smoking bHPVs 16, 18, 31, 33, 45, 52 58 (all the HrHPV types included in nonavalen HPV vaccine) cwestern districts (Imst, Landeck) dEastern districts (Kufstein, Kitzbuehel) eborderline significant The peak age of overall HPV infection, multiple infections and infection with the commonest genotypes lies between 21 and 25 years. This is followed by two lower peaks at 36 to 40 years and at 50 to 55 years. Figure 2 shows the trends in HPV prevalence across age groups. The overall trends in HPV prevalence peaks were statistically significant for overall HPV positivity (P = 0.001) and for common high risk type infections (P = 0.035).Fig. 2 Trends in age specific HPV prevalence among screening population of women in west Austria The compliance for HPV re-testing a year after the initial test was low. A total of only 48 (47.1 %) women initially positive for any hrHPV genotype were re-tested 1 year after the baseline examination. Twenty five (52.1 %) of the infections were persistent. No statistically significant association was observed between HPV persistence and any of the sociodemographic factors. Cytology Data on cytological examination were available for a total of 384 women (Table 4). Twenty one (3.7 %) women were diagnosed with PAPIII+, which were significantly associated with overall, high risk type and multiple HPV infection (P < 0.0001). Of the high-risk types, HPV 16 and 18 were significantly associated with adverse cytological outcomes. Twenty four percent of HPV 16 positive women have accompanying cytological findings of PAP III+. The proportion of smokers is significantly higher among those with PAP III+ [OR = 4.15 (95 % CI: 1.63–10.57)]. A total of 48 high-risk positive patients were re-tested once again 9–15 months after the baseline examination. Twenty-three (52 %) women had persistent infection, the majority of them being HPV 16. Persistent HPV infection was higher among those with abnormal PAP finding, however not statistically significant [OR = 1.56 (95 % CI: 0.41–5.84)].Table 4 HPV prevalence and distribution of high risk and low risk genotypes across cytological findings PAP positive PAP negative (PAP III+) (n = 21) (PAP I-II) (n = 363) Age, years 32.6 (8.1) 35.9 (11.5) Mean (SD) Age at firts sexual contact, years 16.8 (2.8) 16.8 (2.4) Mean (SD) BMI, Mean (SD) 22.5 (3.3) 23.4 (4.0) Current smoker, n (%) 14 (66.7) 117 (32.5)* HPV positive, n (%) 19 (90.5) 75 (20.7)* Multiple HPV infection, n (%) 7 (33.3) 33 (9.1)* Nonavalent type HrHPV, n (%)a 15 (71.4) 48 (13.2)* HPV 16, n (%) 9 (45) 20 (5.8)* HPV 18, n (%) 2 (9.5) 4 (1.1)* HPV 31, n (%) 1 (4.8) 13 (3.6) HPV 33, n (%) 1 (4.8) 14 (3.9) SD standard deviation, HrHPV high risk HPV genotypes PAP III + abnormal PAPanicolaou (cervical smear cytology) including PAP III, PAP IIID, PAP IVA and IVB and PAP V PAPI-II normal PAPanicolaou (cervical smear cytology) *P = sig a16, 18, 31, 33, 45, 52 58 (all HrHPV types included in nonavalen HPV vaccine) Discussion HPV prevalence Prevalence of HPV DNA in a screening population of women aged 18 to 65 years in west Austria was 20.5 %, with the highest prevalence (33.9 %) seen among women aged 21 to 25 years. Although the overall prevalence was slightly higher than the crude HPV prevalence worldwide, the pattern of age specific HPV prevalence was shown to correspond to that seen in other regions [19–21]. Slight differences may be due to differences in the type of test used for the HPV detection or differences in the study population selected. Slightly over 60 % of all the infections are detected among women below 25 years of age, which may represent a high susceptibility of HPV infection right after sexual debut. Like in studies from other region, our study also showed other smaller peaks in higher age groups [22]. There is no clear reason as to why HPV prevalence peaks once again in a later age. One commonly accepted assumption is the fact that divorce and separation tends to put many women back into new-partner sexual relationships. Previous studies attribute the increased prevalence of sexually transmitted infections among older women to an increased divorce rate [23, 24]. Demographic data in Austria (Statistik Austria) report that the peak age for divorce is around 35–40 years [24] which corresponds to our finding of a second HPV peak at around this age range. The second HPV re-peak between 51 and 55 years of age may likewise be explained by an increased risk of new infections as well as by the accompanying age-related anatomical changes in the vulva and vaginal mucosa following menopause at this age. Another explanation for the HPV surge in an older age may be the hypothesis of a latent HPV infection persisting below detection limit and reactivating at some point due to declining immunity or other comorbidities commonly seen in advanced age. This hypothesis has been evidenced in some animal studies or in patients with recurrent respiratory papillomatosis [25–28]. As in several other studies, HPV 16 is clearly the most dominant genotype. It is detected in cervical samples of one out of 20 of the study population. Although HPV 18 is the second common genotype present in cervical cancer samples, it is obviously surpassed by several other genotypes among women with no known cytological abnormalities [29]. Genotypes 31 and 33 – for example - were shown to dominate cervical infection following HPV 16 both in our study and in a study from several European countries [29]. This pattern was also shown to persist among women with high-grade intraepithelial neoplasia [29]. Moreover, data from the WHO ICO information centre on HPV infection (last updated on 20th March 2015) reports that HPV 33 may be the second common genotype detected in cervical cancer samples next to HPV 16 in Austria [30]. The plausibility of this report, however, is questionable. Moreover, in our study, only one of the women with HPV 33 positive samples was diagnosed with abnormal PAP. Further epidemiological studies need to be undertaken in order to clarify or solidify this prevalence pattern among cervical cancer patients in Austria. HPV infection and lifestyle factors Our study showed a threefold higher risk of HPV infection among current smokers, which persisted even after adjusting for other behavioural factors like number of lifetime sexual partners and age at first sexual contact. This might indicate that smoking per se predisposes to HPV infection as is also shown by other studies [31–33]. Tobacco use was reported to have local and systemic immunosuppression effect increasing the probability of acquisition of an HPV infection following an exposure [33]. The exact biological mechanism, however, is not yet clear. This highlights the need to evaluate the exact role of smoking in the natural history of genital HPV infection. Contrary to smoking, high BMI was not associated with increased risk of HPV infection. The fact that HPV infection was significantly higher in non-overweight individuals in our study might be a finding by chance or it might indicate a possible association between physical appearance and sexual behaviour. HPV infection across Tyrolean districts Our finding of higher HPV prevalence in eastern districts of Tyrol compared to those in the west seems to parallel the regional differences in cervical cancer incidence and mortality observed by W Oberaigner et al. [16]. This association persisted even after eliminating the significant age difference between the two regions. Other confounding factors like number of life time sexual partners and age at first sexual contact were considered and adjusted for. Yet the significant difference in HPV prevalence between east and west was persistent, particularly among women above the age of 30 years. However, since questions regarding sexual behaviour may be sensitive, responders’ bias and residual confounding due to these factors can still not be convincingly excluded. Other host and /or pathogen associated factors differentially predisposing to HPV infection and persistence across these geographical regions may be interesting to investigate. HPV immunization Surveys from several countries that have introduced free-of-charge HPV vaccination program since the approval of the vaccines have demonstrated a high vaccine coverage rate of more than 70 % [8, 34]. Accordingly, the prevalence of hrHPV infection has reduced significantly in the years after vaccine introduction. The low immunization coverage (3.7 %) in the Austrian population may be, on one hand, a reflection of the reluctance of the country to provide the vaccine free of charge for the eligible age group coupled with the conservative view of the population towards vaccines particularly towards newly introduced vaccines. On the other hand, however, with the median age of 35 years, the majority of our study population was beyond the age of eligibility as the vaccines were introduced explaining the low coverage to some extent. Our study assessed the attitudes of the study population towards HPV vaccine. The fact that majority of the studied women showed undecidedness about getting vaccinated- which may indirectly reflect undecidedness about letting their children vaccinated - indicates the urgent demand for public health action towards raising awareness on HPV vaccine. Further surveys need to be undertaken which investigate the level of information and the possible barriers to HPV vaccine acceptance. Conclusion This study presents the first data on prevalence and genotype distribution of cervical HPV infections among women with no previously known cytological abnormalities in west Austria. Factors including smoking, number of life time sexual partners, age at first sexual contact and several socio-demographic factors, namely marital status and geographical region were identified as risk factors for age adjusted HPV prevalence. The finding of higher HPV prevalence in the eastern compared to the western district of the study region may - at least partly - explain the east–west gradient in the standardized incidence rate of cervical cancer in the region. The strikingly low HPV vaccine coverage may change with the recently introduced school-based vaccination program [35]. On the other hand, this data – with the extremely low vaccine coverage rate – will serve as baseline pre-vaccine information for further evaluation of the impact of the newly implemented gender-neutral HPV vaccine program in Austria. Additional files Additional file 1: Figure S1. Prevalence and distribution of ten most frequent HPV genotypes among women with no or low grade cervical lesions in Austria. (http://www.hpvcentre.net/statistics/reports/AUT_FS.pdf). Accessed March 2016. (TIF 48 kb) Additional file 2: Figure S2. Map of Tyrol, West Austria: Districts included in the study (circled) (Tyrolean regional government website. https://www.tirol.gv.at/fileadmin/themen/landesentwicklung/raumordnung/bilder/tiris/bezirksuebersicht_tirol_01.jpg. Accessed September 2015. (TIF 167 kb) Additional file 3: Figure S3. Distribution of single and multiple HPV genotypes among screening population of women in west Austria (n=542). (JPG 52 kb) Additional file 4: Figure S4. Distribution (%) of HPV genotypes (high risk types, HPV 6 and HPV 11) among screening population of women in west Austria. (TIF 221 kb) Abbreviations AISAdenocarcinoma in situ ASCUSAtypical squamous cells of undetermined significance BMIBody mass index CIConfidence interval DNADioxyribonucleic acid dUTP/UNGUridine-5’-triphosphate/uracil-N-glycosylase HPVHuman papillomavirus HrHPVHigh risk human papillomavirus HSILHigh grade suqamous epithelial lesion IARCInternational agency for research on cancer IBKInnsbruck ICOInstitut Catalana D‘Oncologia ILInnsbruckland LSILLow grade squamous epithelial lesion OROdds ratio PCRPolymerase chain reaction SDStandard deviation SIRStandardized incidence rate WHOWorld Health Organization Acknowledgements Authors thank: all medical technical assistants (MTAs) at the division of virology, Medical University of Innsbruck, for the technical assistance of HPV testing and genotyping. Authors also thank Tyrolean science fund (TWF) and MFF for the financial support. Funding Tyrolean science fund (Tiroler Wissenschaftsfonds: to W.B. GZ:UNI-0404/1446) and MFF Tirol (Projektvorschlag Nr. 262). Availability of data and materials For a spreadsheet of the raw data – excel or SPSS file containing information from the questionnaires and results of HPV tests – the corresponding author may be contacted (wegene.borena@i-med.ac.at). Authors' contributions All authors have contributed substantially to the conception, design and manuscript preparation. AW, KHK, EM, PM, JM and NR contributed to data acquisition. AG and MG assisted in HPV sequencing and data entry. CM und DVL contributed to the conception of the study and critically reviewed the manuscript. WB contributed to all stages of the work, carried out the main part of data analysis and manuscript preparation. All authors read and approved the final manuscript. Competing interests The authors declare that they have no competing interests. Consent for publication Not applicable. Ethics approval and consent to participate This study was approved by the Ethical Committee of the Medical University of Innsbruck (Ethics committee approval number: UN 4821). Each study participant has been informed about the study and has signed consent for participation. ==== Refs References 1. zur Hausen H Papillomaviruses causing cancer: evasion from host-cell control in early events in carcinogenesis J Natl Cancer Inst 2000 3 92(9) 690 8 10.1093/jnci/92.9.690 10793105 2. Scheurer ME Tortolero-Luna G Adler-Storthz K Human papillomavirus infection: biology, epidemiology, and prevention Int J Gynecol Cancer 2005 15 5 727 46 10.1111/j.1525-1438.2005.00246.x 16174218 3. Trottier H Franco EL The epidemiology of genital human papillomavirus infection Vaccine 2006 24 Suppl 1 S1 15 16406226 4. Garland SM Hernandez-Avila M Wheeler CM Quadrivalent vaccine against human papillomavirus to prevent anogenital diseases N Engl J Med 2007 356 19 1928 43 10.1056/NEJMoa061760 17494926 5. Joura EA Giuliano AR Iversen OE Bouchard C Mao C Mehlsen J Moreira ED Jr Ngan Y Broad spectrum HPV vaccine study. A 9-valent HPV vaccine against infection and intraepithelial neoplasia in women N Engl J Med 2015 372 8 711 23 10.1056/NEJMoa1405044 25693011 6. WHO Countries using HPV vaccine http://www.who.int/immunization/diseases/hpv/decision_implementation/en/. Accessed 2 Feb 2015 7. Paavonen J Naud P Salmerón J Efficacy of human papillomavirus (HPV)-16/18 AS04-adjuvanted vaccine against cervical infection and precancer caused by oncogenic HPV types (PATRICIA): final analysis of a double-blind, randomised study in young women Lancet 2009 374 9686 301 14 10.1016/S0140-6736(09)61248-4 19586656 8. Tabrizi SN Brotherton JM Kaldor JM Assessment of herd immunity and cross-protection after a human papillomavirus vaccination programme in Australia: a repeat cross-sectional study Lancet Infect Dis 2014 14 10 958 66 10.1016/S1473-3099(14)70841-2 25107680 9. Harrison C Britt H Garland S Decreased management of genital warts in young women in Australian general practice post introduction of national HPV vaccination program: results from a nationally representative cross-sectional general practice study PLoS One 2014 9 9 10.1371/journal.pone.0105967 25180698 10. Brotherton JM Fridman M May CL Early effect of the HPV vaccination programme on cervical abnormalities in Victoria, Australia: an ecological study Lancet 2011 377 9783 2085 92 10.1016/S0140-6736(11)60551-5 21684381 11. Bresse X Goergen C Prager B Universal vaccination with the quadrivalent HPV vaccine in Austria: impact on virus circulation, public health and cost-effectiveness analysis Expert Rev Pharmacoecon Outcomes Res 2014 14 2 269 81 10.1586/14737167.2014.881253 24450951 12. http://www.bmg.gv.at/home/Schwerpunkte/Gesundheitsfoerderung_Praevention/Impfen/HPV_Impfung. Accessed May 2015 13. Bachtiary B Obermair A Dreier B Impact of multiple HPV infection on response to treatment and survival in patients receiving radical radiotherapy for cervical cancer Int J Cancer 2002 102 3 237 43 10.1002/ijc.10708 12397642 14. Birner P Bachtiary B Dreier B Signal-amplified colorimetric in situ hybridization for assessment of human papillomavirus infection in cervical lesions Mod Pathol 2001 14 7 702 9 10.1038/modpathol.3880375 11455003 15. Rössler L Reich O Horvat R Human papillomavirus in high-grade cervical lesions: Austrian data of a European multicentre study Wien Klin Wochenschr 2013 125 19–20 591 9 10.1007/s00508-013-0403-6 24043265 16. Oberaigner W Mühlböck H Harrasser L Tumorregister Tirol, Bericht für das Diagnosejahr 2010 2013 Innsbruck IET-Bericht 17. Cirkel C Barop C Beyer DA Method comparison between Munich II and III nomenclature for Pap smear samples J Turk Ger Gynecol Assoc 2015 16 4 203 7 10.5152/jtgga.2015.0147 26692769 18. World Health Organization International Agency for Research on Cancer IARC Monographs on the Evaluationof Carcinogenic Risks to Humans 2007 19. WHO HPV information center. http://www.hpvcentre.net/dataquery.php. Accessed June 2015 20. Iftner T Eberle S Iftner A Prevalence of low-risk and high-risk types of human papillomavirus and other risk factors for HPV infection in Germany within different age groups in women up to 30 years of age: an epidemiological observational study J Med Virol 2010 82 11 1928 39 10.1002/jmv.21910 20872721 21. Dunne EF Unger ER Sternberg M Prevalence of HPV infection among females in the United States JAMA 2007 297 8 813 9 10.1001/jama.297.8.813 17327523 22. Franceschi S Rajkumar R Snijders PJ Papillomavirus infection in rural women in southern India Br J Cancer 2005 14; 92 3 601 6 15668709 23. Idso C Sexually transmitted infection prevention in newly single older women: a forgotten health promotion need Journal for Nurse Practitioners 2009 5 6 440 6 10.1016/j.nurpra.2009.02.015 24. http://www.statistik.at/web_de/statistiken/menschen_und_gesellschaft/bevoelkerung/ehescheidungen/019911.html. Accessed June 2015. 25. Gravitt PE The known unknowns of HPV natural history J Clin Invest 2011 121 12 4593 9 10.1172/JCI57149 22133884 26. Selvakumar R Schmitt A Iftner T Regression of papillomas induced by cottontail rabbit papillomavirus is associated with infiltration of CD8+ cells and persistence of viral DNA after regression J Virol 1997 71 7 5540 8 9188628 27. Abramson AL Nouri M Mullooly V Latent Human Papillomavirus infection is comparable in the larynx and trachea J Med Virol 2004 72 3 473 7 10.1002/jmv.20013 14748072 28. Chan PK Ho WC Wong MC Epidemiologic risk profile of infection with different groups of human papillomaviruses J Med Virol 2009 81 9 1635 44 10.1002/jmv.21575 19623668 29. Tjalma WA Fiander A Reich O HERACLES/SCALE Study Group Differences in human papillomavirus type distribution in high-grade cervical intraepithelial neoplasia and invasive cervical cancer in Europe Int J Cancer 2013 132 4 854 67 10.1002/ijc.27713 22752992 30. WHO ICO HPV Information center http://www.hpvcentre.net/parser.php?xml=M1_Cervical cancer by histology_HPV type distribution Top&iso=AUT&title=Module 1: HPV prevalence estimates - Invasive cervical cancer - Top 10 HPV oncogenic types (Country/Regions). Accessed Aug 2016. 31. Sarian LO Hammes LS Longatto-Filho A Increased risk of oncogenic human papillomavirus infections and incident high-grade cervical intraepithelial neoplasia among smokers: experience from the Latin American screening study Sex Transm Dis 2009 36 4 241 8 10.1097/OLQ.0b013e3181935a7d 19265732 32. Syrjänen K Shabalova I Petrovichev N Smoking is an independent risk factor for oncogenic human papillomavirus (HPV) infections but not for high-grade CIN Eur J Epidemiol 2007 22 10 723 35 10.1007/s10654-007-9180-8 17828436 33. Sopori M Effects of cigarette smoke on the immune system Nat Rev Immunol 2002 2 5 372 7 10.1038/nri803 12033743 34. Kavanagh K Pollock KG Potts A Introduction and sustained high coverage of the HPV bivalent vaccine leads to a reduction in prevalence of HPV 16/18 and closely related HPV types Br J Cancer 2014 110 11 2804 11 10.1038/bjc.2014.198 24736582 35. Austrian Ministry of Health. http://www.bmg.gv.at/cms/home/attachments/3/8/4/CH1426/CMS1389279419763/pressetext_hpv.pdf. Accessed October 2015.

PMC005xxxxxx/PMC5002093.txt

==== Front BMC Musculoskelet DisordBMC Musculoskelet DisordBMC Musculoskeletal Disorders1471-2474BioMed Central London 122410.1186/s12891-016-1224-3Research ArticleEffects of different surgical techniques on mid-distal humeral shaft vascularity: open reduction and internal fixation versus minimally invasive plate osteosynthesis Xue Zichao xzch_2230@163.com 1Jiang Chaolai jiangchaolai@126.com 1Hu Chuanzhen seak2007@qq.com 2Qin Hui qh197826@163.com 1Ding Haoliang 610766885@qq.com 1An Zhiquan anzhiquan@126.com 11 Department of Orthopaedic Surgery, Shanghai Jiao Tong University Affiliated Sixth People’s Hospital, No.600 Yishan Road, Shanghai, 200233 China 2 Department of Orthopaedic Surgery, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, No.197 Rui Jin Er Road, Shanghai, 200025 China 26 8 2016 26 8 2016 2016 17 1 37012 3 2016 18 8 2016 © The Author(s). 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Background Humeral shaft fractures are generally managed with the conventional posterior open reduction and internal fixation (ORIF) or minimally invasive plate osteosynthesis (MIPO). This study was aimed at comparing the outcomes of these surgical techniques in terms of the vascular integrity of the mid-distal humeral shaft. Methods Twelve upper limbs were harvested from 6 fresh cadavers. ORIF or MIPO was randomly performed on either side of each pair of limbs. The axillary artery was perfused with a latex-lead tetraoxide red solution to visualize the vascular structures. The vascular integrity of the humerus was examined by plain radiography and dissection. The periosteal filling achieved with each technique was scored and the scores compared. Results In each limb, one main nutrient artery entering the mid-distal humeral shaft anteromedially (83.3 %) or medially (16.7 %) was first identified. No case of injury to the main nutrient artery was noted for either surgical technique. Injuries to the accessory nutrient arteries entering the mid-distal humeral shaft from the posterior aspect were absent in the MIPO cases, but occurred in 52.9 % of the ORIF cases. In addition, MIPO was also superior to the open plate technique showed superior periosteal filling than. Conclusions Our results showed that the MIPO technique is superior to the ORIF in terms of preserving the vascular integrity of the mid-distal humeral shaft. Keywords Humeral fractureReductionInternal fixationBlood supplythe National Natural Science Foundation of China81171704An Zhiquan issue-copyright-statement© The Author(s) 2016 ==== Body Background Diaphyseal fractures of the humerus are common injuries of the upper arm, and in cases where surgical intervention is necessary, open reduction and internal plate fixation (ORIF) with conventional posterior plating osteosynthesis is considered the best approach [1–3]. However, this approach raises the risks of compromised blood supply and non-union of the fracture due to the associated damage to the soft tissues around the fracture site [2]. Further, ORIF can lead to extensive stripping of soft tissue, disruption of the periosteal blood supply, and iatrogenic radial nerve palsy. In recent times, a new technique of minimally invasive plate osteosynthesis (MIPO) has been gaining popularity in the treatment of mid-distal humeral shaft fractures [4–7]. In our previous study on 33 patients who underwent MIPO or ORIF, we found that MIPO afforded a lower incidence of iatrogenic radial nerve palsies and more rapid fracture union than ORIF [8]. Studies have also shown that in the case of femoral fractures, MIPO better preserves the vascular integrity of the femur than open reduction and plate osteosynthesis [9]. Considering these findings and the minimally invasive nature of MIPO, we speculated that this technique might also help minimize arterial damage in the case of humeral shaft fractures. In this study, we applied plates on the intact humeri of fresh cadavers by using either the ORIF or MIPO technique and compared the effects of these techniques on the vascular integrity of the humerus. Methods Twelve upper limbs were harvested from six fresh cadavers (4 male and 2 female) aged 54 to 87 years (mean age, 68.3 years), as available in department of anatomy of medicial school of Shanghai Jiao Tong University. All donors were natural deaths without any history of upper limb trauma arterial thrombosis, or any history of vascular sclerosis, hypertension or nicotine addiction. The limbs were harvested and operated upon within 48 h of death. Surgical procedures The right and left humeri of each cadaver were randomized to undergo either ORIF or MIPO. MIPO was commenced with a 3-cm-long proximal incision made medial to the insertion of the deltoid and lateral to the biceps. Then, the cortex of the anterior humeral shaft was exposed. Another 3-cm-long distal incision was made proximal to the flexion crease, along the lateral border of the biceps. The brachialis was bluntly split to expose the humeral shaft. A submuscular tunnel was prepared and a plate was submuscularly inserted from the distal incision, adjusted to adhere to the anterior aspect of the humeral shaft, and fixed with screws placed distally and proximally [10]. On the contralateral humerus, ORIF was performed by making a conventional posterior longitudinal incision through the triceps, followed by the placement of a plate and its fixation with distal and proximal screws. One author, an attending surgeon, performed all of the surgeries. Vascular perfusion Before the operation, the axillary artery and vein were catheterized and secured with two non-occlusive silk ties [11]. Next, the axillary artery was flushed with 300 mL of warm saline until no blood clots came out from the axillary vein. To ensure optimal results of perfusion, the remaining arterial branches were ligated. After completion of the surgery, the axillary artery was perfused with 150 mL of staining solution (latex: water: lead tetraoxide = 1: 1: 2 in volume) until the dye was extruded through the axillary vein. Thereafter, radiography of the limbs was performed to evaluate the status of the vascular structures. The limbs were refrigerated overnight to harden the dye. Vascular dissection The day after vascular perfusion, all the limbs were dissected anteriorly and posteriorly. An anterior median incision was taken such that it connected the two small incisions made previously. The biceps and the brachial muscle were dissected to expose the brachial artery (the plate was removed if MIPO had been performed). The previously made posterior incision was then reopened to expose the radial nerve and the arteria profunda brachii (the plate was removed if ORIF had been performed). Then, the vascular structures were traced to their origin to assess the integrity of the main nutrient arteries and the accessory nutrient arteries. Periosteal filling scores For evaluation of the degree of periosteal filling, the periosteal vessels were completely exposed and the filling conditions scored by a previously described method [9]. For the purpose of evaluation, the humeral shaft was divided into four zones: proximally fixed portion of the plate, bridging portion of the plate, distally fixed portion of the plate, and bone distal to the plate. Periosteal filling in each zone was separately scored into 4 grades depending on the degree of staining: 0, no staining; 1, mild staining; 2, moderate staining; and 3, marked staining. The difference between the periosteal filling scores of the right and left limbs of each cadaver were calculated to determine the difference between the two surgical techniques. A difference of 9–12 points was considered marked; 5–8 points, moderate; and 2–4, mild; a difference of 0–1 implied that the effects of both techniques were similar. Results Vascular integrity One orthopedic surgeon and one radiologist interpreted the x-ray by a standardized manner to reduce inter-observer variability to the greatest possible extent. Perfusion was successfully achieved in all the 12 limbs. The vascular structure was visualized by plain radiography, as shown in Fig. 1. In each limb, only one main nutrient artery was detected, and in all cases, it originated from the brachial artery and entered the mid-distal portion of the humeral shaft. The main nutrient arteries entered the humeral shaft on the anteromedial aspect in 10 limbs (83.3 %) and on the medial aspect in 2 limbs (16.7 %), and they did not give rise to any branches before entering the humeral shaft. The main nutrient arteries in all the specimens showed good integrity and filling (Fig. 2).Fig. 1 Radiograph images of vascular perfusion. a shows the main nutrient artery (denoted by arrow) of the humeral shaft, b shows the accessory nutrient arteries (denoted by arrow heads) of the humeral shaft Fig. 2 Dissection and observation of main nutrient artery. The integrity of the main nutrient arteries was not affected by either surgical technique. P, the proximal end; D, the distal end; BA, the brachial artery; B, the brachial muscle. The arrows denote the main nutrient arteries Accessory nutrient arteries were found to originate from the arteria profunda brachii (2–4 branches) and enter the humeral shaft posteriorly, through the radial groove (Fig. 3). Specimens in which MIPO was performed showed good preservation of the accessory nutrient arteries in terms of both structural integrity and blood flow (Fig. 3), unlike specimens that underwent ORIF, which showed disruption of both structural integrity and blood flood (Fig. 4) (Table 1).Fig. 3 Posterior dissection and observation of specimen underwent MIPO. Accessory nutrient arteries were preserved unaffected by MIPO. Accessory nutrient arteries originated from the deep brachial artery (2–4 branches) and entered the mid-distal portion of the humeral shaft posteriorly, through the radial groove. P, the proximal end; D, the distal end; Apb, the arteria profunda brachii; RN, radial nerve; Acm, arteria collateralis media. The arrow heads denote the accessory nutrient arteries Fig. 4 Posterior dissection and observation of specimen underwent ORIF. ORIF damaged accessory mutrient arteries and resulted in profound exudation of the perfusion solution. ORIF also led to very poor periosteal filling condition. P, the proximal end; D, the distal end; Apb, the arteria profunda brachii; RN, radial nerve Table 1 The number of damaged accessory nutrient arteries Cadaver number MIPO ORIF 1 0/4 3/4 2 0/3 2/3 3 0/3 2/3 4 0/1 0/1 5 0/4 2/4 6 0/2 0/2 Percentage 0 % 52.9 % MIPO minimally invasive plate osteosynthesis, ORIF open reduction and internal fixation Periosteal filling Periosteal filling was evaluated and graded by the method described previously (Fig. 5). As shown in Table 2, the filling and staining of the periosteum achieved with MIPO were better than those achieved with ORIF (Table 2). The differences in the periosteal filling scores obtained for the two surgical techniques were marked in 3 pairs, moderate in 2 pairs, and mild in 1 pair.Fig. 5 Representative pictures of each grade of periosteal filling. a-d showed periosteal filling scored as 0–3 respectively Table 2 Score of periosteal filling No MIPO ORIF Score (MIPO - ORIF) 1 12 2 10 2 12 3 9 3 11 4 7 4 11 3 8 5 10 6 4 6 11 2 9 Discussion MIPO has been used in the treatment of humeral shaft fractures and has shown some advantages over ORIF [8]. In this study, we sought to determine how these surgical techniques affect the vascular integrity of the mid-distal portion of the humeral shaft. In keeping with previous reports, our study showed that the humerus has one main nutrient artery and several accessory nutrient arteries [12]. Our findings indicated that MIPO was superior to ORIF in maintaining local vascular integrity and promoting periosteal filling at the fracture site. Neither ORIF nor MIPO affects the stability and functioning of the shoulder and elbow joints, thus allowing early postoperative mobilization and good joint function [13]. In addition, MIPO is also associated with lower risk for iatrogenic radial nerve palsies in comparison with ORIF [4, 8]. Adequate blood supply is essential for the bone union process after fractures [14]. Mid-distal humeral shaft fractures are generally associated with damage to the main nutrient artery of the humeral shaft [15, 16]. Therefore, the blood supply to the fracture site mainly relies on an extraosseous blood supply derived from surrounding soft tissues [17, 18]. However, in the conventional method of open reduction and internal fixation of fractures of the mid-distal humeral shaft, the stripping of the soft tissues and periosteum around the fracture site is unavoidable. This may compromise the poor blood supply to the distal fracture fragments, thereby increasing the risk for non-union. The findings of our study showed that MIPO caused less damage to the accessory nutrient arteries and their blood flow, unlike the case with ORIF, where they were frequently damaged and often necessitated ligation. And it has been confirmed that these accessory arteries is crucial to fracture healing and ligation of them will lead to adverse outcome [19]. Put together, these findings indicate that MIPO might be superior to ORIF in preserving the blood supply of the mid-distal portion of the humeral shaft. And because the fracture pattern is unpredictable, it is preferable to use a minimal invasive approach to preserve the remaining blood supply and minimize the iatrogenic disruption of the perfusion [20]. Some drawbacks of MIPO also need to be considered. Closed reduction required for MIPO is technically difficult; therefore, the surgeon performing the procedure should have received sufficient training and the surgery is prolonged. Further, frequent intraoperative fluorescent examination may be necessary to ensure proper reduction, thereby further extending the operation time. Moreover, angulation deformity is an inherent risk of closed reduction. This study has some limitations. The two surgical techniques were performed on intact bones, and therefore, the actual impact of these techniques on vascular integrity in the presence of shaft fractures could not be assessed in this study. Furthermore, as the study focused on the effect of different approaches, we did not unify the instruments, among which the screws could have an influence on the integrity of the accessory nutrient arteries. Conclusions Thus, our findings in this cadaveric study showed that the MIPO was superior to the ORIF in preserving the nutrient arteries and periosteal vasculature. This implies that MIPO may help maintain good vascularization of the fracture site, thereby promoting bone union in cases of humeral shaft fractures. Abbreviations MIPOMinimally invasive plate osteosynthesis ORIFOpen reduction and internal fixation None. Funding This study was supported by a grant from the National Natural Science Foundation of China (81171704). Availability of data and materials All the data supporting the conclusions of this article can be request from Zhiquan An (anzhiquan@126.com) as corresponding author. Authors’ contributions ZX, CJ, CH, HQ, HD carried out the studies, participated in the sequence alignment and ZX drafted the manuscript. ZA, ZX, CJ participated in the design of the study. ZA conceived of the study, and participated in its design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript. Competing interests The authors declare that they have no competing interests. Consent for publication Not applicable. Ethics approval and consent to participate The study protocol was approved by the Ethics Committee of Shanghai 6th People’s Hospital. None of the body donors were from a vulnerable population and all donors or next of kin provided written informed consent that was freely given. The research procedure was explained to the family or the next of kin and their written informed consents was obtained before the cadavers were collected. All information, including identity and note record, was kept in confidential. ==== Refs References 1. Bhandari M Devereaux PJ McKee MD Schemitsch EH Compression plating versus intramedullary nailing of humeral shaft fractures--a meta-analysis Acta Orthop 2006 77 2 279 84 10.1080/17453670610046037 16752291 2. Changulani M Jain UK Keswani T Comparison of the use of the humerus intramedullary nail and dynamic compression plate for the management of diaphyseal fractures of the humerus. A randomised controlled study Int Orthop 2007 31 3 391 5 10.1007/s00264-006-0200-1 16900354 3. Niall DM O'Mahony J McElwain JP Plating of humeral shaft fractures--has the pendulum swung back? Injury 2004 35 6 580 6 10.1016/j.injury.2003.10.021 15135277 4. Apivatthakakul T Arpornchayanon O Bavornratanavech S Minimally invasive plate osteosynthesis (MIPO) of the humeral shaft fracture. Is it possible? A cadaveric study and preliminary report Injury 2005 36 4 530 8 10.1016/j.injury.2004.05.036 15755436 5. Livani B Belangero WD Bridging plate osteosynthesis of humeral shaft fractures Injury 2004 35 6 587 95 10.1016/j.injury.2003.12.003 15135278 6. Ma HR Ma SL Yin TZ Coupled exofixator for comminuted fracture of humeral shaft Chin J Traumatol 2005 8 1 33 35 15676087 7. Yang KH Helical plate fixation for treatment of comminuted fractures of the proximal and middle one-third of the humerus Injury 2005 36 1 75 80 10.1016/j.injury.2004.03.023 15589919 8. An Z Zeng B He X Chen Q Hu S Plating osteosynthesis of mid-distal humeral shaft fractures: minimally invasive versus conventional open reduction technique Int Orthop 2010 34 1 131 5 10.1007/s00264-009-0753-x 19301000 9. Farouk O Krettek C Miclau T Schandelmaier P Guy P Tscherne H Minimally invasive plate osteosynthesis and vascularity: preliminary results of a cadaver injection study Injury 1997 28 Suppl 1 A7 12 10.1016/S0020-1383(97)90110-8 10897282 10. Zhiquan A Bingfang Z Yeming W Chi Z Peiyan H Minimally invasive plating osteosynthesis (MIPO) of middle and distal third humeral shaft fractures J Orthop Trauma 2007 21 9 628 33 10.1097/BOT.0b013e31815928c2 17921838 11. Anous MM A clean technique for cadaver injection studies Plast Reconstr Surg 1988 82 1 199 3380913 12. Laing PG The arterial supply of the adult humerus J Bone Joint Surg Am 1956 38-A 5 1105 16 13367086 13. Kobayashi M Watanabe Y Matsushita T Early full range of shoulder and elbow motion is possible after minimally invasive plate osteosynthesis for humeral shaft fractures J Orthop Trauma 2010 24 4 212 6 10.1097/BOT.0b013e3181c2fe49 20335753 14. Kontakis GM Tosounidis T Pagkalos J Humeral diaphyseal aseptic non-unions: an Algorithm of management Injury 2007 38 Suppl 2 S39 49 10.1016/S0020-1383(07)80008-8 17920417 15. Nagel A The clinical significance of the nutrient artery Orthop Rev 1993 22 5 557 61 8316419 16. Pidhorz L Acute and chronic humeral shaft fractures in adults Orthop Traumatol Surg Res 2015 101 1 Suppl S41 9 10.1016/j.otsr.2014.07.034 25604002 17. Rhinelander FW Tibial blood supply in relation to fracture healing Clin Orthop Relat Res 1974 105 34 81 10.1097/00003086-197411000-00005 4609655 18. Wilson JW Vascular supply to normal bone and healing fractures Semin Vet Med Surg 1991 6 1 26 38 19. Gainor BJ Metzler M Humeral shaft fracture with brachial artery injury Clin Orthop Relat Res 1986 204 154 61 3956006 20. Gardner MJ Voos JE Wanich T Helfet DL Lorich DG Vascular implications of minimally invasive plating of proximal humerus fractures J Orthop Trauma 2006 20 9 602 7 10.1097/01.bot.0000246412.10176.14 17088661

PMC005xxxxxx/PMC5002094.txt

==== Front Lipids Health DisLipids Health DisLipids in Health and Disease1476-511XBioMed Central London 31410.1186/s12944-016-0314-4ResearchThe therapeutic efficacy of intensive medical therapy in ameliorating high-density lipoprotein dysfunction in subjects with type two diabetes Kashyap Sangeeta 216-445-2679Kashyas@ccf.org 1Kheniser Karim Kheniser.3@gmail.com 1Li Ling lil5@ccf.org 2Bena James BenaJ@ccf.org 3Kasumov Takhar 330-325-6552tkasumov@neomed.edu 451 Departemnt of Endocrinology and Metabolism, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44195 USA 2 Department of Core Facilities, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44195 USA 3 Department of Quantitative Health Sciences, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44195 USA 4 Department of Hepatology, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44195 USA 5 Present address: Department of Pharmaceutical Sciences, Northeast Ohio Medical University, 4209 St. R. 44, PO Box 95, Rootstown, OH 44272 USA 27 8 2016 27 8 2016 2016 15 1 1415 5 2016 23 8 2016 © The Author(s). 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Background To determine whether 12 months of intensive medical therapy (IMT) improves HDL functionality parameters in subjects with type II diabetes (T2D). Methods Retrospective, randomized, and controlled 12-month IMT intervention trial that enrolled 13-subjects with T2D (age 51- years, fasting glucose 147 mg/dL, body mass index [BMI] 36.5 kg/m2) and nine healthy control (46-years, fasting glucose 90 mg/dL, BMI 26.5 kg/m2). Subjects with T2D underwent IMT and HDL functionality measures (pro-inflammatory index of high-density lipoprotein (pHDL)), paraoxonase one (PON1), ceruloplasmin (Cp), and myeloperoxidase (MPO) activity were performed on samples at baseline and at 12-months following IMT. Results At baseline, pHDL index was significantly higher in subjects with T2D (p < 0.001) and apolipoprotein A-1 levels were significantly lower (p = 0.013) vs. controls. After 12-months, there was a trend for improved pHDL activity (p = 0.083), as indicated by intent-to-treat analysis, but when the non-adherent subject was omitted (per-protocol), significant attenuations in pHDL activity (p = 0.040) were noted; Δ pHDL activity at 12-months was associated with Δ weight (r = 0.62, p = 0.032) and Δ fasting glucose (r = 0.65, p = 0.022). Moreover, PON1 activity significantly improved (p < 0.001). The aforementioned occurred in association with improvements in inflammatory markers (i.e., C-reactive protein & tumor necrosis factor), hemoglobin A1C, fasting glucose, triglycerides, high-density lipoprotein levels and adipokines. Conclusion IMT ameliorates pHDL index and significantly improves anti-oxidative function, as measured by PON1. Improvements in weight and fasting glucose mediated the decrease in pHDL index. Pharmacological aids and lifestyle modification are required to improve cardiovascular risk factors, subsequent mortality risk, and promote T2D remission. Application of either form of therapy alone may only have relatively miniscule effects on the aforementioned factors, in relation to the aggregate. Keywords Intensive medical therapyParaoxonase onePro-inflammatory high-density lipoproteinsMyeloperoxidaseissue-copyright-statement© The Author(s) 2016 ==== Body Background Undoubtedly, physical inactivity, mass consumption of calorically dense and vitamin deficient processed foods, and environmental factors have contributed to the unabated rise in obesity and type II diabetes mellitus (T2D). In the ensuing few decades, the prevalence of T2D is expected to rise from 2.8 to 4.4 % [1]. As a result, the incidence of cardiovascular related deaths will increase correspondingly [2]. The heightened prevalence of cardiovascular disease (CVD) and cardiovascular-related deaths among subjects with T2D can be partially explained by reductions in high-density lipoprotein (HDL) levels [3]. On a particle number basis, diminutions in HDL would concomitantly reduce the ability of HDL to exert their physiochemical functions, which include potent anti-atherogenic, anti-oxidative, cholesterol efflux, vasodilatory, and anti-inflammatory functions [4]. Consequently, given their constellation of pertinent functions, improving HDL levels in subjects with T2D has been one of the focal points of medical therapy. However, although empirical evidence has indicated that HDL levels are negatively correlated with CVD [5], the association may become distorted in subjects with T2D, due to the presence of HDL dysfunction. Indeed, T2D is marked by depressed anti-oxidative function and increased pro-inflammatory HDL activity (pHDL); not only does pHDL precipitate the formation of oxidized LDL, but they are incapable of preventing and inactivating the formation of LDL-derived oxidized phospholipids [6–8]. The aforesaid is precipitated by a milieu that engenders deleterious morphological alterations to HDL, such as promoting the disassociation of apolipoprotein A-1 (apoA-1) from HDL [9]. Specifically, its etiology directly manifests from increased oxidative stress, chronic inflammation, hyperglycemia, and hypertriglyceridemia [4]. The aggregate perturbs HDL function by promoting HDL glycosylation [10], reducing hepatic synthesis of apoA-1 [11], replacing apoA-1 and paraoxonase (PON1) with serum amyloid A (SAA) [12], and post-translation modification (PTM) of apoA-1 [13]; concurrently, HDL enrichment in triglycerides (TG) and the depletion of cholesterol ester in its hydrophobic core, which is a consequent of increased adipose tissue lipolysis, increased hepatic TG synthesis and the aberrant activity of cholesterol ester transfer protein (CETP), lipoprotein lipase (LPL), lecithin cholesterol acyl-transferase (LCAT) and hepatic lipase (HL) are also contributory [14, 15]. As a result, HDL antioxidative, cholesterol efflux, and anti-inflammatory functions become attenuated [4]. Even more alarming is that dysfunctional HDL have been demonstrated to propagate the inflammatory cascade and consequently may play a pivotal role in inducing plaque progression [16]; pHDL are associated with diminished levels of PON1 [6], which would reduce HDL ability to inhibit low-density lipoproteins (LDL) oxidation by prooxidant enzymes such as myeloperoxidase (MPO) [17]. Therefore, assaying HDL function through qualitative means has supplanted quantitative measurements because the latter fails to provide any insight into its true physiochemical functions. Several lines of evidences suggest that pharmacological and/or lifestyle interventions ameliorate HDL dysfunction. For instance, CETP inhibitors can improve HDL functionality by increasing the ratio of cholesterol esters (CE) to TG [18]. Administration of these drugs with niacin have been demonstrated to reduce apoA-1 fractional catabolic rate in patients with T2D [19]. Further, statins have been noted to attenuate HDL-inflammatory index levels in subjects with coronary heart disease [20]. Correspondingly, physical activity has been demonstrated to improve LCAT, cholesterol efflux, and PON1 activity, collectively conferring potent anti-atherogenic effects [21–24]. Similarly, a three week dietary and physical activity intervention has been noted to significantly improve pHDL activity [25]. However, the aforementioned trials utilized interventions that were of relatively short duration (≤ 3 months) and none analyzed the sustained effects of intensive medical therapy (IMT) (i.e., behavioral and pharmacological treatment) on subjects who are obese and primarily stricken with T2D. Therefore, the purpose of this retrospective study was to ascertain how and to what extent a long-term (one year) IMT intervention would ameliorate HDL dysfunction in obese subjects with T2D; specifically, elucidating if pro-inflammatory HDL and PON1 activity are improved and if they are associated with reductions in MPO, cholesterol, TG, glycaemia, and body mass index (BMI) is most prudent. The central hypothesis was that hyperglycemia induced oxidative stress in subjects with T2D causes loss of anti-oxidative and anti-inflammatory functions of HDL. Moreover, intensive glycemic control will restore HDL functionality. Therefore, we assessed the in vitro functionality of HDL in subjects with T2D vs. healthy control subjects at baseline and following twelve months of intensive diabetes treatment with lifestyle counseling and glucose lowering medications. Methods Retrospectively, we conducted a study on subjects with T2D who were enrolled in a randomized, controlled, single center trial and received intensive medical therapy (IMT) for 12 months. Specifically, IMT was denoted as referencing the most novel therapeutic guidelines set forth by the American Diabetes Association, which included pharmacological treatment and lifestyle modification counseling (hypocaloric, carbohydrate controlled diet and moderate physical activity) provided by quarterly visits with study endocrinologists and an annual visit with diabetic educators. Prescribed pharmacological agents included biguanides (BG), incretin mimetics, thiazolidinediones (THZ), sulfonylureas (SF), and insulin analogs. Of which, BG, incretin mimetics, and insulin were the most commonly administered. Inclusion criteria consisted of an age of 20 to 60 years, HGBA1C levels of ≥ 7.0 %, and a BMI of 27 to 43 kg/m2. The research was performed in accordance with the Declaration of Helsinki and all subjects provided written informed consent. The trial was approved by the institutional review board at the Cleveland Clinic. All chemicals were purchased from Sigma-Aldrich Chemical Company (St Louis, MO) except where indicated otherwise. Blood glucose was measured using the glucose oxidase method (Beckman glucose analyzer, Beckman Instruments, Fullerton, CA), and serum insulin levels were measured using a commercial enzyme-linked immunosorbent assay kit (Linco Research, St Charles, MO). Homeostasis model assessment (HOMA) was calculated as a measure of insulin resistance. Adipokines (adiponectin and leptin) were measured using a high-sensitivity human cytokine multiplex kit (LINCOplex; Linco Research, St Charles, MO). Serum C-reactive protein (CRP) concentration was measured with high-sensitivity sandwich enzyme-linked immunosorbent assay. Total cholesterol, HDL-cholesterol, TG, and glycosylated hemoglobin (HGBA1C) were measured by standard methods in the certified clinical laboratory. Outcomes Primary outcome measures were assessed at baseline and one year after intervention, for T2D, while they were assayed at baseline, in controls. Human ApoA-1 quantification Human ApoA-1 was quantified by immunoassay method on the Abbott ARCHITECT ci8200 Integrated Analyzer System (Abbott Labs, Abbott Park, IL). PON1 activity PON1 activity in 5 μl serum was assayed based on a fluorescence assay (excitation at 360 nm and emission at 450 nm) using EnzChek (Molecular Probes, Inc. Eugene, OR) Paraoxonase Assay Kit protocol. Paraoxon was used as a substrate. Pro-inflammatory index of HDL A modification of the cell-free assay developed by Navab and colleagues [8] was used to quantify pro-inflammatory HDL [26]. This assay measures the anti-oxidant capacity of plasma proteins to prevent Cu2-induced oxidative stress. Briefly, HDL oxidation in apoB-depleted plasma was initiated with Cu2+ and rates of HDL oxidation quantified with 2′,7′-dichlorodihydrofluorescein (DCFH) in a microtiter plate at 37 °C. Fluorescent emission with 530 nm wavelength was measured after serial excitation at 485 nm. MPO activity The peroxidase activity of MPO in serum was measured by spectrophotometer at 650 nm with 3,3′5,5′-tetramethylbenzidine (TMB) as a substrate. Ceruloplasmin (Cp) activity The amino oxidase activity of Cp in serum was measured by spectrophotometer at 530 nm with p-phenylenediamine (based on the article by Wei et al. [27], with modifications based on ref. Lehmann et al. [28]). Statistical analysis Analysis was based on an intention-to-treat and per-protocol. Due to nonparametric data, continuous variables were summarized with medians and quartiles, whilst categorical factors were summarized via frequency and percentiles. To assess group differences, a Fisher’s exact test was used, with respect to categorical variables. Independent group differences for continuous variables were assessed using the Wilcoxon rank sum test, whereas time-related within-group differences for continuous variables were ascertained via the Wilcoxon signed rank test. Lastly, spearman correlation was used to deduce associations between parameters. Data was analyzed via SAS software (version 9.3; Cary, NC). Results Study cohort and baseline characteristics Thirteen subjects with T2D were identified as subjects who met the inclusion criteria, while nine subjects served as the control. Table 1 depicts baseline differences between subjects with T2D and controls. Although BMI was significantly greater among subjects with T2D (p < 0.001), there were no differences in demographic variables between controls (n = 7 [77.8 %] females, age 46 [30.0, 57.2] years, BMI 25.6 [21.2, 26.9] kg/m2) and subjects with T2D (n = 8 [61.5 %] females, age 51.1 [43.7, 55.3] years, BMI 36.5 [33.4, 37.5] kg/m2). For primary outcome measures, pHDL activity was significantly higher (0.49 [0.42, 0.54] vs. 0.26 [0.25, 0.34], p < 0.001) and apoA-1 levels were significantly lower (124.5 [111.4, 128.7] vs. 144.2 [129.4, 174.5], p = 0.013), while there were non-significant differences in PON1 (0.01 [0.01, 0.01] vs. 0.01 [0.01, 0.01], p = 0.15), Cp (0.16 [0.14, 0.17] vs. 0.18 [0.14, 0.25], p = 0.22), and MPO activity (6.5 [5.6, 20.9] vs. 11.8 [4.4, 14.6], p = 0.31). With respect to metabolic parameters, HDL-cholesterol (41.0 [37.0, 45.0] vs. 63.0 [45.0, 72.0], p = 0.005), fasting insulin ([INS-0], 32.4 [16.9, 41.5] vs. 5.9 [5.4, 7.1], p < 0.001), fasting glucose ([GLUC], 180.0 [161.0, 240.0] vs. 90.0 [83.0, 93.0], p < 0.001), HOMA (14.4 [6.8, 22.0] vs. 1.2 [1.1, 1,5], p < 0.001) were potentiated, whilst TG (174.0 [152.0, 337.0] vs. 74.0 [50.0, 86.0], p = 0.008) concentrations were attenuated among T2D; non-significant differences were observed in total cholesterol (187.0 [172.0, 202.0] vs. 175.0 [143.0, 195.0], p = 0.32) and LDL-cholesterol (104.0 [82.0, 114.0] vs. 85.0 [65.5, 95.5], p = 0.15). Finally, systolic blood pressure ([SBP], 137.0 [133.0, 156.0] vs. 108.0 [104.0, 115.0], p < 0.001), and diastolic blood pressure ([DBP], 89.0 [83.0, 91.0] vs. 66.0 [65.0, 69.0], p < 0.001) were significantly greater among T2D.Table 1 Differences in baseline characteristics between subjects with T2D and controls Overall (n = 22) Control (n = 9) D2M (n = 13) Parameter n Summary n Summary n Summary p-value Fasting glucose (mg/dl) 22 147.0 [91.0,198.0] 9 90.0 [83.0,93.0] 13 180.0 [161.0,240.0] < 0.001 b BMI (kg/m2) 22 31.9 [26.4,37.0] 9 25.6 [21.2,26.9] 13 36.5 [33.4,37.5] < 0.001 b HDL-cholesterol (mg/dl) 22 44.5 [37.0,56.0] 9 63.0 [45.0,72.0] 13 41.0 [37.0,45.0] 0.005 b Triglycerides (mg/dl) 22 141.5 [82.0,187.0] 9 74.0 [50.0,86.0] 13 174.0 [152.0,337.0] 0.008 b HOMA 22 4.8 [1.3,14.8] 9 1.2 [1.1,1.5] 13 14.4 [6.8,22.0] < 0.001 b Total cholesterol (mg/dl) 22 180.0 [158.0,202.0] 9 175.0 [143.0,195.0] 13 187.0 [172.0,202.0] 0.32b LDL-cholesterol (mg/dl) 21 95.0 [79.0,109.0] 8 85.0 [65.5,95.5] 13 104.0 [82.0,114.0] 0.15b Fasting insulin (mU/l) 22 10.6 [6.3,34.3] 9 5.9 [5.4,7.1] 13 32.4 [16.9,41.5] <0.001 b Age (years) 22 49.0 [41.7,55.6] 9 46.0 [30.0,57.2] 13 51.1 [43.7,55.3] 0.37b Female 22 15 (68.2) 9 7 (77.8) 13 8 (61.5) 0.65d Systolic blood pressure (mmHG) 22 126.0 [115.0,138.0] 9 108.0 [104.0,115.0] 13 137.0 [133.0,156.0] < 0.001 b Diastolic blood pressure (mmHG) 22 76.0 [68.0,89.0] 9 66.0 [65.0,69.0] 13 89.0 [83.0,91.0] < 0.001 b MPO activity (mOD/μL/min) 21 8.3 [5.1,14.8] 9 11.8 [4.4,14.6] 12 6.5 [5.6,20.9] 0.31b pHDL assay (RFU mg HDLc/min) 21 0.42 [0.28,0.50] 9 0.26 [0.25,0.34] 12 0.49 [0.42,0.54] < 0.001 b Cp activity (μmol.min-1/mL) 21 0.16 [0.14,0.18] 9 0.18 [0.14,0.25] 12 0.16 [0.14,0.17] 0.22b PONI1 activity (nmole.min-1/μL) 21 0.01 [0.01,0.01] 9 0.01 [0.01,0.01] 12 0.01 [0.01,0.01] 0.15b apoA-1 (mg/dl) 20 128.6 [114.8,143.7] 9 144.2 [129.4,174.5] 11 124.5 [111.4,128.7] 0.013 b Intent-to-treat analysis. Values presented as median [P25, P75] or n (column %) p-values: b = Wilcoxon rank sum test, d = Fisher’s exact test Efficacy of IMT Table 2 shows the effects of 12 months of IMT on primary outcome measures, metabolic parameters, anthropometric characteristics, inflammatory markers, and endocrine parameters in subjects with T2D, in relation to baseline. For primary outcome measures, there was a trend for reduced pHDL activity (0.43 [0.35, 0.51] vs. 0.49 [0.42, 0.54], p < 0.083), but when subjected to a per-protocol analysis the p-value became significant (0.40 [0.34,0.51] vs. 0.49 [0.42,0.54], p = 0.040) (Fig. 1); further, significant increases in PON1 activity (0.02 [0.01, 0.03] vs. 0.01 [0.01, 0.01], p < 0.001), whereas MPO (5.9 [4.9, 8.6] vs. 6.5 [5.6, 20.9], p = 0.18) and Cp (0.14 [0.13, 0.19] vs. 0.16 [0.14, 0.17], p = 0.98] were not significantly different. As it pertains to metabolic variables, HDL-cholesterol (50.0 [46.0, 56.0] vs. 41.0 [37.0, 45.0], p = <0.001), GLUC (108.0 [88.0, 128.0] vs. 180.0 [161.0, 240.0], p = 0.002), HGBA1C (6.1 [6.0, 6.7] vs. 9.1 [8.9, 10.2], p < 0.001), and TG (119.0 [94.0, 152.0] vs. 174.0 [152.0, 337.0], p < 0.001) significantly improved, while total cholesterol (182.0 [152.0, 209.0] vs. 187.0 [172.0, 202.0], p = 0.53) and LDL-cholesterol (91.0 [73.0, 127.0] vs. 104.0 [82.0, 114.0], p = 0.58) were not significantly different. Moreover, IMT induced significant decreases in BMI (28.0 [26.4, 30.7] vs. 36.5 [33.4, 37.5], p < 0.001) and weight (83.5 [69.4, 92.1] vs. 96.8 [93.2, 111.1], p < 0.001), while congruently reducing inflammatory markers such as CRP (1.00 [0.50, 3.1] vs. 5.5 [3.1, 6.2], p < 0.001) and tumor necrosis factor ([TNF] 1.04 [0.61, 1.1] vs. 1.4 [1.02, 2.0], p < 0.006). Lastly, adiponectin levels increased (6.1 [4.7, 9.2] vs. 2.8 [2.4, 4.0], p < 0.001), whilst leptin concentrations decreased (13.6 [9.3, 23.6] vs. 22.7 [15.8, 32.4], p = 0.006).Table 2 Differences in clinical characteristics in subjects with T2D after twelve-months of IMT Baseline (n = 13) Follow-up (n = 13) Change (n = 13) Parameter n Summary n Summary n Summary p-value Fasting glucose (mg/dl) 13 180.0 [161.0,240.0] 13 108.0 [88.0,128.0] 13 −87.0 [−127.0,−41.0] 0.002 HGBA1C (%) 13 9.1 [8.9,10.2] 13 6.1 [6.0,6.7] 13 −3.0 [−4.1,−2.1] < 0.001 Body weight (kg) 13 96.8 [93.2,111.1] 13 83.5 [69.4,92.1] 13 −14.7 [−37.9,−5.9] < 0.001 BMI (kg/m2) 13 36.5 [33.4,37.5] 13 28.0 [26.4,30.7] 13 −5.9 [−9.2,−2.0] < 0.001 HDL-cholesterol (mg/dl) 13 41.0 [37.0,45.0] 13 50.0 [46.0,56.0] 13 7.0 [4.0,16.0] < 0.001 Triglycerides (mg/dl) 13 174.0 [152.0,337.0] 13 119.0 [94.0,152.0] 13 −42.0 [−135.0,−13.0] < 0.001 Tumor necrosis factor 13 1.4 [1.02,2.0] 13 1.04 [0.61,1.1] 13 −0.39 [−0.68,−0.23] 0.006 C-reactive protein (mg/dl) 13 5.5 [3.1,6.2] 13 1.00 [0.50,3.1] 13 −2.5 [−4.5,−1.7] < 0.001 Adiponectin (ug/ml) 13 2.8 [2.4,4.0] 13 6.1 [4.7,9.2] 13 2.9 [2.2,5.5] 0.001 Leptin (ng/ml) 13 22.7 [15.8,32.4] 13 13.6 [9.3,23.6] 13 −11.5 [−16.6,−2.0] 0.006 Total cholesterol (mg/dl) 13 187.0 [172.0,202.0] 13 182.0 [152.0,209.0] 13 2.0 [−38.0,13.0] 0.53 LDL-cholesterol (mg/dl) 13 104.0 [82.0,114.0] 13 91.0 [73.0,127.0] 13 −2.0 [−18.0,11.0] 0.58 MPO activity (mOD/μL/min) 12 6.5 [5.6,20.9] 12 5.9 [4.9,8.6] 12 −1.3 [−9.0,0.30] 0.18 pHDL assay (RFU mg HDLc/min) 12 0.49 [0.42,0.54] 12 0.43 [0.35,0.51] 12 −0.11 [−0.13,0.01] 0.083 pHDL assay (RFU mg HDLc/min) 11 0.49 [0.42,0.54] 11 0.40 [0.34,0.51] 11 −0.13 [−0.13, −0.02] 0.040 a Cp activity (μmol.min-1/mL) 12 0.16 [0.14,0.17] 12 0.14 [0.13,0.19] 12 0.00 [−0.02,0.02] 0.98 PONI activity (nmole.min-1/μL) 12 0.01 [0.01,0.01] 12 0.02 [0.01,0.03] 12 0.01 [0.00,0.02] < 0.001 Systolic blood pressure (mmHG) 13 137.0 [133.0,156.0] 13 129.0 [117.0,141.0] 13 −8.0 [−22.0,3.0] 0.051 Diastolic blood pressure (mmHG) 13 89.0 [83.0,91.0] 13 79.0 [72.0,85.0] 13 −8.0 [−11.0,−1.00] 0.005 Values presented as median [P25, P75] p-values: Wilcoxon signed rank test. a = per-protocol analysis Fig. 1 Box plot depicting differences in pHDL activity at baseline and after 12 months of IMT. As illustrated, there were significant differences at baseline between subjects with T2D and controls. After 12 months of IMT, a per-protocol analysis, which omitted the non-adherent subject, indicated that pHDL activity significantly decreased. p-values: Wilcoxon rank sum test for intergroup differences, whereas within-group differences were analyzed via the Wilcoxon signed rank test. DM, type II diabetics; pHDL, pro-inflammatory high-density lipoprotein activity Association between primary HDL functionality outcome variables and clinical/metabolic parameters Table 3 illustrates relations between primary outcome variables and clinical marker at baseline and after 12 months of IMT for subjects with T2D and among the aggregate of both groups. Baseline combined group correlations indicated that pHDL was positively correlated with BMI (r = 0.74, p < 0.001, 95 % confidence interval [CI] 0.42–1.00), INS-0 (r = 0.73, p < 0.001, CI 0.41–1.00), GLUC (r = 0.71, p < 0.001, CI 0.37–1.00), HOMA (r = 0.72, p < 0.001, CI 0.38–1.00), and TG (r = 0.53, p = 0.014, CI 0.12–0.94), whereas it was negatively associated with HDL-cholesterol levels (r = −0.65, p = 0.001, CI −1.00–−0.29). Moreover, MPO and PON1 activity were positively correlated with total cholesterol (r = 0.55, p = 0.011, CI 0.14–0.95; r = 0.46, p = 0.036, CI 0.03–0.89, respectively), whereas Cp activity was negatively associated with age (r = −0.46, p = 0.037, CI −0.88–−0.03). Conversely, with respect to subjects with T2D at baseline, PON1 and Cp activity were correlated with HDL-cholesterol (r = 0.78, p = 0.003, CI 0.35–1.00) and CRP (r = 0.66, p = 0.019, CI 0.13–1.00), respectively; further, pHDL was significantly associated with LDL-cholesterol (r = 0.59, p = 0.042, CI 0.03–1.00). After 12 months of IMT, the difference in (Δ) Cp activity was significantly associated with Δ CRP (r = 0.69, p = 0.012, CI 0.19–1.00) and Δ TNF (r = 0.79, p = 0.002, CI 0.36–1.00), while Δ MPO was correlated with Δ total cholesterol levels (r = 0.59, p = 0.044, CI 0.02–1.00); furthermore, Δ pro-inflammatory index of HDL was related with Δ weight and Δ glucose (r = 0.62, p = 0.032, CI 0.07–1.00; r = 0.65, p = 0.022, CI 0.12–1.00, respectively) (Fig. 2).Table 3 Correlations between primary outcome variables and clinical characteristics at baseline and twelve-months Parameter n rho 95 % CI p value Baseline Combined Group Correlations  Cp Activity Age 21 −0.46 (−0.88,−0.03) 0.037  MPO Activity Total cholesterol 21 0.55 (0.14,0.95) 0.011  PON1 Activity Total cholesterol 21 0.46 (0.03,0.89) 0.036 Systolic blood pressure 21 0.45 (0.03,0.88) 0.038  pHDL Assay Body mass index 21 0.74 (0.42,1.00) < 0.001 Fasting insulin 21 0.73 (0.41,1.00) < 0.001 Fasting glucose 21 0.71 (0.37,1.00) < 0.001 HOMA 21 0.72 (0.38,1.00) < 0.001 High-density lipoproteins 21 −0.65 (−1.00,−0.29) 0.001 Triglycerides 21 0.53 (0.12,0.94) 0.014 Systolic blood pressure 21 0.74 (0.41,1.00) < 0.001 Diastolic blood pressure 21 0.61 (0.22,0.99) 0.004 Baseline Correlations among T2D  Cp Activity C-reactive protein 12 0.66 (0.13,1.00) 0.019  PON1 Activity High-density lipoproteins 12 0.78 (0.35,1.00) 0.003  pHDL Assay Low-density lipoproteins 12 0.59 (0.03,1.00) 0.042 One Year Correlations among T2D  Cp - Difference C-reactive protein – difference 12 0.69 (0.19,1.00) 0.012 Tumor necrosis factor – difference 12 0.79 (0.36,1.00) 0.002  MPO - Difference Fasting cholesterol – difference 12 0.59 (0.02,1.00) 0.044  pHDL - Difference Weight – difference 12 0.62 (0.07,1.00) 0.032 Fasting glucose – difference 12 0.65 (0.12,1.00) 0.022 Intent-to-treat analysis, r-values: Spearman correlation Cp ceruloplasmin, MPO myeloperoxidase, PON1 paraoxonase one, pHDL pro-inflammatory high-density lipoproteins Fig. 2 a Positive correlation between Δ pHDL activity and Δ weight in subjects with T2D, after 12 months of IMT. b Positive association between Δ pHDL activity Δ fasting glucose in subjects with T2D, after 12 months of IMT. The figure highlights that large decreases in weight or glucose were related with heightened decrements in pHDL activity. Intent-to-treat analysis; r-value: Spearman correlation. pHDL, pro-inflammatory high-density lipoproteins Discussion While others have noted the presence of pHDL activity and defective HDL anti-oxidative function in subjects with T2D [6, 7], the principal and novel outcome was that pHDL activity may be ameliorated with IMT. Particularly, although there was a two-fold difference in pHDL activity at baseline between controls and subjects with T2D, HDL dysfunction was ameliorated, such that there was a trend for reduced pHDL activity after 12 months of IMT. However, when subjected to a per-protocol analysis, which removed the non-adherent subject, the results became modestly significant (p = 0.040). In particular, IMT fostered reductions in pHDL activity after 12 months, in nine of the twelve type II diabetic subjects, in which it was measured, whereas it was noted to increase in three. Along with pHDL activity, one subject’s metabolic profile was exacerbated at 12 months, thereby demonstrating that the subject was non-adherent to therapy; while pHDL activity was refractory, even in the midst of decreases in blood glucose and TG, in the remaining two subjects. Further, in the aforementioned non-responders, weight was modestly reduced or maintained at 12 months. Consequently, attenuations in pHDL activity may necessitate concomitant significant decreases in blood glucose and weight. Reductions in GLUC would attenuate HDL glycosylation and oxidation, which would improve HDL functionality [10]. As such, correlational analysis indicated that Δ blood glucose and Δ weight were significantly correlated with Δ pHDL activity at 12 months, which affirmed the hypothesis that there would be a relationship between glycaemia and pHDL activity. Of note, the results from this study advance the knowledge on the effectiveness of dietary and physical activity paradigms that foster weight-loss by demonstrating that additional benefits (i.e., improved HDL function, etc.) can be accrued from these modalities. Similarly, pharmacotherapy that attenuates blood glucose levels may provide additional benefits outside of the effects on GLUC and insulin control; this has been demonstrated with CETP inhibitors, statins, and niacin medications [18–20, 29]. Consequently, the synergistic actions and thus the aggregate of physical activity, pharmacotherapy, and dietary adherence would provide the most benefit than one modality of IMT alone, given that HDL dysfunction is caused by an overabundance of factors. Baseline combined group analysis demonstrated that pHDL activity was related with BMI, INS-0, blood glucose, HOMA, HDL-cholesterol, and TG, while it was associated with LDL-cholesterol in subjects with T2D, when diabetes was uncontrolled. Therefore, pHDL activity may be associated with divergent parameters, depending on the population and severity of disease, and a myriad of variables may precipitate HDL dysfunction. The association between PON1 and HDL at baseline for T2D underpins them as mutualistic beneficiaries, but this correlation is more frequently noted among controls [30]. Interestingly, the association was not evident at one year, which suggests that other factors mediate their association [30]. Noteworthy, HDL antioxidant activity as assayed by PON1 was not different between subjects with T2D and controls at baseline, which has been corroborated in other trials [20, 31, 32]. The results indicate that PON1 activity may be resilient to the hyperglycemic, hypertriglyceridemic, and pro-inflammatory milieu that is present in subjects with T2D; further, PON1 activity may only be significantly reduced in subjects with long-standing T2D. In addition, the non-significant intergroup differences in PON1 activity and the heightened levels of pHDL activity at baseline attests that other factors mediate HDL anti-inflammatory dysfunction. Indeed, Navab et al. [33] observed that PON1 activity alone did not explain the divergences in HDL function. Further, PAF activity is concomitantly diminished [6]. Therefore, decrement activity in the aforesaid parameter may have also been causal to HDL dysfunction. However, PON1 activity improved after 12 months of IMT, which indicates that it may be more amenable to IMT, given that it improved significantly, while there was a trend for improved pHDL activity. Potentiated PON1 activity would heighten HDL ability to inhibit LDL oxidation and mildly oxidized LDL, which results in reduced monocyte chemotaxis [8, 16, 33], and assisting in inhibiting the formation of the one of the prime initial catalysts to the development of atheroma: oxidized LDL [34]. Further, HDL is less prone to glucose-induced lipid peroxidation when PON1 activity is elevated [10]. Albeit non-significant, Cp and MPO activity were lower at baseline and remained so even after 12 months of IMT, relative to controls. Interestingly, the aforementioned did not significantly decrease, while inflammatory markers (i.e., CRP and TNF) decreased significantly in subjects with T2D. Inflammation is requisite to increased acute phase protein levels [4], and improvements in PON1 activity would expectedly attenuate MPO activity, but this was not significant; MPO and PON1 reciprocate each others actions [35]. Further, acute phase proteins (e.g., Cp) have been demonstrated to replace PON1, LCAT, and PAF, which precipitates pHDL activity [8, 16]. Cp has also been noted to facilitate LDL oxidation [36]. Moreover, MPO activity can negatively modulate HDL function by impairing its antiapototic and anti-inflammatory capacity, while congruently increasing its pro-inflammatory activity [37]. Similarly, MPO intermediates (e.g., hypothiocyanous and hypochlorous acid) induce apoA-1 oxidation, which would impair its cholesterol efflux capacity and consequently its anti-atherogenic functions [13, 38]. The former selectively targets tryptophan residues [13]. However, even though elevated levels of Cp and MPO activity may cause HDL dysfunction and contribute to pHDL activity, the results provide credence to the fact that it is unlikely that they had a significant effect on pHDL activity. Indeed, pHDL activity improved irrespectively of significant reductions in Cp and MPO and, relative to controls, their activity was lower at baseline, when pHDL activity was elevated. Conversely, apoA-1 levels were significantly attenuated among subjects with T2D at baseline [39]. Indeed, subjects with T2D displayed elevated levels of GLUC, TG, and inflammatory markers. Due to the hypertriglyceridemic and hyperglycemic milieu, HDL become enriched in TG and apoA-1 becomes glycosylated [4]. The decreased CE/TG ratio engenders apoA-1 instability and alters the conformation of the central and C-terminal domains [40]. Further, due to chronic inflammation and oxidative stress, apoA-1 is replaced by SAA and undergoes PTM [4]. The aggregate causes reductions in apoA-1 levels and perturbs apoA-1 functionality. Subsequently, their ability to promote cholesterol efflux, remove seeding molecules (i.e., formation of metabolites of linoleic and arachidonic acid by endothelial cells such as 12-lipoxygenase) from LDL, and inhibit LDL oxidation will be abrogated [33, 41, 42]. With respect to the latter, apoA-1 confers antioxidative functions directly, as mentioned, and indirectly via improvements in PON1 activity [42]. Of importance, cholesterol efflux capacity is inversely related with CVD [43], and oxidized LDL promote monocyte adherence and chemotaxis [33]. However, it is plausible that apoA-1 levels increased, given that HDL levels improved after 12 months of IMT [44]. Weight-loss, PON1 activity, and pharmacological aids appreciate HDL levels [9, 30, 45]. Further, it may have been a consequent of reductions in CETP activity, increased hepatic synthesis and lipidation of apoA-1, and increased cholesterol efflux [4]. In regard to CETP inhibitors, torcetrapib, dalcetrapib and evacetrapib have been discontinued either due to adverse effects in the case of the former or modest effects on cardiovascular outcomes in the latter two medications; the effects of novel CETP inhibitors on cardiovascular outcomes is yet to be elucidated [46]. Concurrently, physical activity has been documented to improve cholesterol efflux and LCAT activity, which would increase the ratio of CE to TG [21–24]; thereby, improving HDL stability [40]. Moreover, the significant reductions in TG would favor up-regulated levels of CE, relative to TG in HDL hydrophobic core [4]. Therefore, on a particle number basis, increased HDL concentrations improves their ability to exert their critical physiochemical functions. Conclusion While discerning between the effects of a multidimensional therapeutic paradigm of IMT on HDL function represented a core limitation to the study, the trial sample size was also finite (n = 13). However, rather than strictly looking at the acute effects of pharmacotherapy and lifestyle modification on HDL dysfunction, the present study gleaned additional insight into the long-term benefits and efficacy of IMT. To our knowledge, we are first to report the effects of IMT on HDL function during a long-term trial and particularly in obese subjects with T2D. Specifically, we noted that IMT is a feasible approach to mitigate and possibly ameliorate the deleterious effects of T2D. After all, retarding the progression of T2D requires the synergistic application of lifestyle modification and pharmacotherapy, which more aptly treats the disease. Abbreviations apoA-1Apolipoprotein A-1 BGBiguanides BMIBody mass index CECholesterol esters CETPCholesterol ester transfer protein CPCeruloplasmin CRPC-reactive protein DBPDiastolic blood pressure DCFH2′,7′-dichlorodihydrofluorescein GLUCFasting blood glucose HDLHigh-density lipoprotein HGBA1CGlycosylated hemoglobin HLHepatic lipase HOMAHomeostasis model assessment IMTIntensive medical therapy LCATLecithin cholesterol acyl-transferase LDLLow-density lipoproteins LPLLipoprotein lipase MPOMyeloperoxidase MSMetabolic syndrome PAFPlatelet-activity factor-acetyl hydrolase pHDLPro-inflammatory high-density lipoproteins PON1Paraoxonase one PTMPost-translation modification SAASerum amyloid A SBPSystolic blood pressure SFSulfonylureas T2DType II diabetes TGTriglycerides THZThiazolidinediones TNFTumor necrosis factor Acknowledgements Not applicable. Funding With respect to funding sources, there is no information to disclose. Authors’ contributions SK made substantial contributions to the design, conception, interpretation, drafting, and revision of the manuscript. KK made substantial contributions to the drafting of the manuscript. LL made substantial contributions to the acquisition of data, while JB made substantial contributions to the data analyses and interpretation. TK made substantial contributions to the revision, interpretation, and appropriate formatting of the paper. All authors read and approved the final manuscript. Competing of interests The authors declare that they have no competing interests. Consent for publication Not applicable. Ethics approval and consent to participate The trial was approved by the Institutional Review Board at the Cleveland Clinic. All participants signed an informed consent form, prior to the outset of the trial. ==== Refs References 1. Wild S Roglic G Green A Sicree R King H Global prevalence of diabetes: estimates for the year 2000 and projections for 2030 Diabetes Care 2004 27 5 1047 1053 10.2337/diacare.27.5.1047 15111519 2. Laakso M Cardiovascular disease in type 2 diabetes: challenge for treatment and prevention J Intern Med 2001 249 3 225 235 10.1046/j.1365-2796.2001.00789.x 11285042 3. Grant RW Meigs JB Prevalence and treatment of low HDL cholesterol among primary care patients with type 2 diabetes: an unmet challenge for cardiovascular risk reduction Diabetes Care 2007 30 3 479 484 10.2337/dc06-1961 17327308 4. Kontush A Chapman MJ Functionally defective high-density lipoprotein: a new therapeutic target at the crossroads of dyslipidemia, inflammation, and atherosclerosis Pharmacol Rev 2006 58 3 342 374 10.1124/pr.58.3.1 16968945 5. Maron DJ The epidemiology of low levels of high-density lipoprotein cholesterol in patients with and without coronary artery disease Am J Cardiol 2000 86 12 11 14 10.1016/S0002-9149(00)01462-4 6. Nobécourt E Jacqueminet S Hansel B Chantepie S Grimaldi A Chapman MJ Kontush A Defective antioxidative activity of small dense HDL3 particles in type 2 diabetes: relationship to elevated oxidative stress and hyperglycaemia Diabetologia 2005 48 3 529 538 10.1007/s00125-004-1655-5 15729582 7. Morgantini C Natali A Boldrini B Imaizumi S Navab M Fogelman AM Ferrannini E Reddy ST Anti-inflammatory and antioxidant properties of HDLs are impaired in type 2 diabetes Diabetes 2011 60 10 2617 2623 10.2337/db11-0378 21852676 8. Navab M Hama SY Hough GP Subbanagounder G Reddy ST Fogelman AM A cell-free assay for detecting HDL that is dysfunctional in preventing the formation of or inactivating oxidized phospholipids J Lipid Res 2001 42 8 1308 1317 11483633 9. Barter PJ High density lipoprotein: a therapeutic target in type 2 diabetes Endocrinol Metab 2013 28 3 169 177 10.3803/EnM.2013.28.3.169 10. Ferretti G Bacchetti T Marchionni C Caldarelli L Curatola G Effect of glycation of high density lipoproteins on their physicochemical properties and on paraoxonase activity Acta Diabetol 2001 38 4 163 169 10.1007/s592-001-8074-z 11855794 11. Lowell CA Stearman RS Morrow JF Transcriptional regulation of serum amyloid a gene expression J Biol Chem 1986 261 18 8453 8461 2424896 12. Parks JS Rudel LL Alteration of high density lipoprotein subfraction distribution with induction of serum amyloid a protein (SAA) in the nonhuman primate J Lipid Res 1985 26 1 82 91 3919135 13. Hadfield KA Pattison DI Brown BE Hou L Rye KA Davies MJ Hawkins CL Myeloperoxidase-derived oxidants modify apolipoprotein A-I and generate dysfunctional high-density lipoproteins: comparison of hypothiocyanous acid (HOSCN) with hypochlorous acid (HOCl) Biochem J 2013 449 2 531 542 10.1042/BJ20121210 23088652 14. Khovidhunkit W Kim MS Memon RA Shigenaga JK Moser AH Feingold KR Grunfeld C Effects of infection and inflammation on lipid and lipoprotein metabolism: mechanisms and consequences to the host J Lipid Res 2004 45 7 1169 1196 10.1194/jlr.R300019-JLR200 15102878 15. Esteve E Ricart W Fernández-Real JM Dyslipidemia and inflammation: an evolutionary conserved mechanism Clin Nutr 2005 24 1 16 31 10.1016/j.clnu.2004.08.004 15681098 16. Navab M Berliner JA Subbanagounder G Hama S Lusis AJ Castellani LW Reddy S Shih D Shi W Watson AD HDL and the inflammatory response induced by LDL-derived oxidized phospholipids Arterioscler Thromb Vasc Biol 2001 21 4 481 488 10.1161/01.ATV.21.4.481 11304461 17. Feingold KR Memon RA Moser AH Grunfeld C Paraoxonase activity in the serum and hepatic mRNA levels decrease during the acute phase response Atherosclerosis 1998 139 2 307 315 10.1016/S0021-9150(98)00084-7 9712337 18. Mohammadpour AH Akhlaghi F Future of cholesteryl ester transfer protein (CETP) inhibitors: a pharmacological perspective Clin Pharmacokinet 2013 8 52 615 626 10.1007/s40262-013-0071-8 23658137 19. Pang J Chan DC Hamilton SJ Tenneti VS Watts GF Barrett PH Effect of niacin on high-density lipoprotein apolipoprotein A-I kinetics in statin-treated patients with type 2 diabetes mellitus Arterioscler Thromb Vasc Biol 2014 34 2 427 432 10.1161/ATVBAHA.113.302019 24285582 20. Ansell BJ Navab M Hama S Kamranpour N Fonarow G Hough G Rahmani S Mottahedeh R Dave R Reddy ST Inflammatory/antiinflammatory properties of high-density lipoprotein distinguish patients from control subjects better than high-density lipoprotein cholesterol levels and are favorably affected by simvastatin treatment Circulation 2003 108 22 2751 2756 10.1161/01.CIR.0000103624.14436.4B 14638544 21. Králová LI Suchánek P Kovár J Poledne R Life style change and reverse cholesterol transport in obese women Physiol Res 2009 58 1 S33 S38 22. Riedl I Yoshioka M Nishida Y Tobina T Paradis R Shono N Tanaka H St-Amand J Regulation of skeletal muscle transcriptome in elderly men after 6 weeks of endurance training at lactate threshold intensity Exp Gerontol 2010 45 11 896 903 10.1016/j.exger.2010.08.014 20813182 23. Campbell SC Moffatt RJ Kushnick MR Continuous and intermittent walking alters HDL(2)-C and LCATa Atherosclerosis 2011 218 2 524 529 10.1016/j.atherosclerosis.2011.07.002 21798542 24. Casella-Filho A Chagas AC Maranhão RC Trombetta IC Cesena FH Silva VM Tanus-Santos JE Negrão CE da Luz PL Effect of exercise training on plasma levels and functional properties of high-density lipoprotein cholesterol in the metabolic syndrome Am J Cardiol 2011 107 8 1168 1172 10.1016/j.amjcard.2010.12.014 21310370 25. Roberts CK Ng C Hama S Eliseo AJ Barnard RJ Effect of a short-term diet and exercise intervention on inflammatory/anti-inflammatory properties of HDL in overweight/obese men with cardiovascular risk factors J Appl Physiol 2006 101 6 1727 1732 10.1152/japplphysiol.00345.2006 16902063 26. Ji X Xu H Zhang H Hillery CA Gao HQ Pritchard KA Anion exchange HPLC isolation of high-density lipoprotein (HDL) and on-line estimation of proinflammatory HDL Plos One 2014 9 3 e91089 10.1371/journal.pone.0091089 24609013 27. Wei H Frei B Beckman JS Zhang WJ Copper chelation by tetrathiomolybdate inhibits lipopolysaccharide-induced inflammatory responses in vivo Am J Physiol Heart Circ Physiol 2011 301 3 H712 H720 10.1152/ajpheart.01299.2010 21724870 28. Lehmann HP Schosinsky KH Beeler MF Standardization of serum ceruloplasmin concentrations in international enzyme units with o-dianisidine dihydrochloride as substrate Clin Chem 1974 20 12 1564 1567 4430134 29. Vergès B Florentin E Baillot-Rudoni S Petit JM Brindisi MC de Barros JP P Lagrost L Gambert P Duvillard L Rosuvastatin 20 mg restores normal HDL-apoA-I kinetics in type 2 diabetes J Lipid Res 2009 50 6 1209 1215 10.1194/jlr.P800040-JLR200 19168444 30. Blatter Garin M Moren X James R Paraoxonase-1 and serum concentrations of HDL-cholesterol and apoA-I J Lipid Res 2006 47 3 515 520 10.1194/jlr.M500281-JLR200 16327022 31. Sanguinetti SM Brites FD Fasulo V Verona J Elbert A Wikinski RL Schreier LE HDL oxidability and its protective effect against LDL oxidation in Type 2 diabetic patients Diabetes Nutr Metab 2001 14 1 27 36 11345163 32. Kopprasch S Pietzsch J Kuhlisch E Graessler J Lack of association between serum paraoxonase 1 activities and increased oxidized low-density lipoprotein levels in impaired glucose tolerance and newly diagnosed diabetes mellitus J Clin Endocrinol Metab 2003 88 4 1711 1716 10.1210/jc.2002-021561 12679462 33. Navab M Hama SY Anantharamaiah GM Hassan K Hough GP Watson AD Reddy ST Sevanian A Fonarow GC Fogelman AM Normal high density lipoprotein inhibits three steps in the formation of mildly oxidized low density lipoprotein: steps 2 and 3 J Lipid Res 2000 41 9 1495 1508 10974057 34. Yoshida H Kisugi R Mechanisms of LDL oxidation Clin Chim Acta 2010 411 23–24 1875 1882 10.1016/j.cca.2010.08.038 20816951 35. Huang Y Wu Z Riwanto M Gao S Levison B Gu X Fu X Wagner M Besler C Gerstenecker G Myeloperoxidase, paraoxonase-1, and HDL form a functional ternary complex J Clin Invest 2013 123 9 3815 3828 10.1172/JCI67478 23908111 36. Ehrenwald E Chisolm GM Fox PL Intact human ceruloplasmin oxidatively modifies low density lipoprotein J Clin Invest 1994 93 4 1493 1501 10.1172/JCI117127 8163654 37. Undurti A Huang Y Lupica JA Smith JD DiDonato JA Hazen SL Modification of high density lipoprotein by myeloperoxidase generates a pro-inflammatory particle J Biol Chem 2009 284 45 30825 30835 10.1074/jbc.M109.047605 19726691 38. Bergt C Pennathur S Fu X Byun J O’Brien K McDonald TO Singh P Anantharamaiah GM Chait A Brunzell J The myeloperoxidase product hypochlorous acid oxidizes HDL in the human artery wall and impairs ABCA1-dependent cholesterol transport Proc Natl Acad Sci U S A 2004 101 35 13032 13037 10.1073/pnas.0405292101 15326314 39. Hu A Luo Y Li T Guo X Ding X Zhu X Wang X Tang S Low serum apolipoprotein A1/B ratio is associated with proliferative diabetic retinopathy in type 2 diabetes Graefes Arch Clin Exp Ophthalmol 2012 250 7 957 962 10.1007/s00417-011-1855-x 22327732 40. Sparks DL Davidson WS Lund-Katz S Phillips MC Effects of the neutral lipid content of high density lipoprotein on apolipoprotein A-I structure and particle stability J Biol Chem 1995 270 45 26910 26917 10.1074/jbc.270.45.26910 7592936 41. Zhang Y Zanotti I Reilly MP Glick JM Rothblat GH Rader DJ Overexpression of apolipoprotein A-I promotes reverse transport of cholesterol from macrophages to feces in vivo Circulation 2003 108 6 661 663 10.1161/01.CIR.0000086981.09834.E0 12900335 42. Navab M Hama SY Cooke CJ Anantharamaiah GM Chaddha M Jin L Subbanagounder G Faull KF Reddy ST Miller NE Normal high density lipoprotein inhibits three steps in the formation of mildly oxidized low density lipoprotein: step 1 J Lipid Res 2000 41 9 1481 1494 10974056 43. Saleheen D Scott R Javad S Zhao W Rodrigues A Picataggi A Lukmanova D Mucksavage ML Luben R Billheimer J Association of HDL cholesterol efflux capacity with incident coronary heart disease events: a prospective case-control study Lancet Diabetes Endocrinol 2015 3 7 507 513 10.1016/S2213-8587(15)00126-6 26025389 44. Barter PJ Brandrup-Wognsen G Palmer MK Nicholls SJ Effect of statins on HDL-C: a complex process unrelated to changes in LDL-C: analysis of the VOYAGER Database J Lipid Res 2010 51 6 1546 1553 10.1194/jlr.P002816 19965573 45. Gomaraschi M Adorni MF Banach M Bernini F Franceschini G Calabresi L Effects of established hypolipidemic drugs on HDL concentration, subclass distribution, and function Handb Exp Pharmacol 2015 224 593 615 10.1007/978-3-319-09665-0_19 25523003 46. Filippatos TD Klouras E Barkas F Elisaf M Cholesteryl ester transfer protein inhibitors: challenges and perspectives Expert Rev Cardiovasc Ther 2016 14 953 10.1080/14779072.2016.1189327 27171534

PMC005xxxxxx/PMC5002095.txt

==== Front BMC Public HealthBMC Public HealthBMC Public Health1471-2458BioMed Central London 357310.1186/s12889-016-3573-0DebateThe future of financing for HIV services in Uganda and the wider sub-Saharan Africa region: should we ask patients to contribute to the cost of their care? Kakaire Tom 1Schlech Walter 2Coutinho Alex 1Brough Richard 1Parkes-Ratanshi Rosalind +256-414307238rp549@medschl.cam.ac.uk 11 Infectious Diseases Institute, Makerere University College of Health Sciences, PO Box 22418, Kampala, Uganda 2 Dalhousie University Faculty of Medicine, Nova Scotia, Canada 27 8 2016 27 8 2016 2016 16 1 89622 5 2015 23 8 2016 © Kakaire et al. 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Whilst multi-lateral funding for HIV/AIDS dramatically increased from 2004 to 2008, it has largely plateaued in the last 8 years. Across sub-Saharan Africa, up to 20 % of total spending on health is used for HIV services, and of this over 85 % is estimated to come from international funding rather than in-country sources. In Uganda, the fiscal liability to maintain services for all those who are currently receiving it is estimated to be as much as 3 % of Gross Domestic Product (GDP). In order to meet the growing need of increased patient numbers and further ART coverage the projected costs of comprehensive HIV care and treatment services will increase substantially. Current access to HIV care includes free at point of delivery (provided by Ministry of Health clinics), as well as out-of-pocket financing and health insurance provided care at private for- and not for- profit facilities. The HIV response is funded through Ugandan Ministry of Health national budget allocations, as well as multilateral donations such as the President’s Emergency Plan for AIDS in Africa (PEPFAR) and Global Fund (GF) and other international funders. We are concerned that current funding mechanism for HIV programs in Uganda may be difficult to sustain and as service providers we are keen to explore ways in which provide lifelong HIV care to as many people living with HIV (PLHIV) as possible. Until such time as the Ugandan economy can support universal, state-supported, comprehensive healthcare, bridging alternatives must be considered. We suggest that offering patients with the sufficient means to assume some of the financial burden for their care in return for more convenient services could be one component of increasing coverage and sustaining services for those living with HIV. Keywords FinancingSustainingUgandaHIVTreatmentissue-copyright-statement© The Author(s) 2016 ==== Body Background Since 1990, 458 billion United States dollars (USD) has been spent in development aid for health. This aid comes primarily from donor governments and philanthropic organizations. Of this, 23.2 % was targeted for HIV/AIDs treatment and support [1]. Whilst multi-lateral funding for HIV/AIDS dramatically increased from 2004 to 2008, it has largely plateaued in the last 8 years, with a drop in 2010 and marginal 1 % increase from 2013 to 14 [2]. Across sub-Saharan Africa, up to 20 % of total spending on health is used for HIV services, and of this over 85 % is estimated to come from international funding rather than in-country sources [3]. Uganda is among 51 countries that rely on international sources to fund over 75 % of their HIV response; over 20 donors contribute over USD 400 million annually [4]. There were an estimated 1,500,000 people living with HIV in Uganda in 2012 and by the end of June 2013 567,000 eligible persons were estimated to be receiving ART. Unfortunately despite the scale up of HIV services in Uganda, there is an increased HIV prevalence (from 6.8 % in 2001 to 7.2 % in 2012) and only 40 % of people eligible for ART are currently receiving it [5]. Even at this level of ART coverage the fiscal liability for Uganda to maintain services for all those who are currently receiving it is estimated to be as much as 3 % of GDP [5, 6]. In order to meet the growing need of increased patient numbers and further ART coverage the projected costs of comprehensive HIV care and treatment services will increase substantially. Whilst this has been deemed to be cost effective in modeling in sub-Saharan African epidemics (Zambia, South Africa) expansion of ART coverage to those with CD4 counts over 350 cells/ul will still cost between $237 to $1691 per DALY averted [7]. Consequently the sustainability of these services is extremely uncertain, as international funding has plateaued and there is no clear long term plan for the Government of Uganda (GOU) to fill the potential gap. This is compounded by the growing view that HIV/AIDS care gets a disproportionate allocation of national health resources relative to other chronic conditions and that it should be “normalized” so that the financing model for HIV/AIDs more closely reflects the economic and social reality of Uganda. This article reviews the current literature on sources of HIV service financing in Uganda and the possible options to expand and sustain life-long HIV treatment in Uganda. Models of access to health care and HIV services in Uganda Table 1 summarizes the ongoing local financing models in Uganda. It describes the current status of these options mentioned above, and the action required from stakeholders to expand and strengthen these options.Table 1 Future HIV financing options for PNFPs Option Current status Government action required PNFP action required Direct Government funding of services and/or centralized HIV/AIDS Fund [2, 6, 10] Inadequate resources and inadequate budget allocation • Streamlined, transparent public- private partnership (PPP) frameworks for health • Long term budget commitment • Engagement of donors for medium term budget support • Innovative tax/revenue collection and allocation measures • Stronger accountability and performance/quality measurement systems • Systems to meet minimum public private partnership guidelines • Less vertical /“silo” based programs • Health system wide capacity building • Quality Assurance and Performance measurement systems National Health Insurance Scheme [6, 11, 13] In discussion • Stakeholder involvement; generating a consensus on nature and scope of coverage • Fund management capacity • Governance, accountability and confidence building • Legislation and regulation • Definition and implementation of systems to meet minimum NHIS requirements for PNFP providers • Capacity building for NHIS compliance • Quality Assurance and Performance measurement systems Out-of-Pocket Service Currently funds 50 % of health care [32] • Stronger accountability and performance/quality measurement systems • Direct cash transfers • Quality Assurance and Performance measurement systems Private Insurance [16, 17] Very low coverage, inadequate regulation • Stronger Regulation • Stronger accountability and performance/quality measurement systems • Creating an attractive business proposition for private insurance firms • Creating systems to meet minimum private health insurance company requirements Community Health Insurance Schemes [11] Very low coverage, inadequate regulation • Legislation and regulation • Community engagement and confidence building • Community accountability mechanisms • Differentiated care for scheme members Co-Payment to subsidize overall costs of care Used mainly by faith based organizations and NGOs, but little documentation on coverage and best practices • No additional legislation and/or regulation • Creating capacity to manage and report separate (paying and non-paying client) income streams • Mobilizing paying and non-paying client support • Mobilizing external stakeholder (eg donor and government) support • Maintenance of standard quality care for all clients Free at point of delivery Following the abolition of user fees in Uganda in 2001 [8], health care at government clinics is free at the point of delivery, for a menu of selected services (mainly out-patient) and informal charging for additional add on services which are sometimes present. This could be termed a “publically provided health insurance” financed by taxes and other sources of government revenue. The Ministry of Health aim is to provide all necessary HIV/AIDS services free in the 2655 government health facilities it owns and manages [9]. However, the total health allocation in the national budget fell from 8.3 % in FY 2011/2012 to 7.4 % in FY 2012/2013 which means that the overall level of expenditure on health is significantly below the Abuja target of a 15 % allocation of a government’s budget to health [10], and finding sufficient funding for HIV and other health services is a challenge. The GOU has taken some steps to introduce a National Health Insurance Scheme (NHIS) since 1995, but this has made little progress largely due to negative stakeholder influence [11]. Even if it was implemented, it’s not clear how the NHIS in its current proposed form would incorporate existing health financing systems in order to avoid fragmentation and create mechanisms for cross-subsidization [12]. Moreover, feasibility studies generally point to public mistrust, as well as a lack of key systems and capacities (particularly relating to governance and accountability) to implement the NHIS scheme in Uganda which would need to be addressed [13]. Out-of-pocket financing In addition to government clinic there are 994 private for profit (PFP) and 801 private not for profit (PNFP) (including faith based) health facilities, of varying quality, size and geographical distribution [9]. The 2010 National Health Accounts (NHA) Study in Uganda for health expenditures estimated that 50 % of health expenditure came from private funds (42 % from households) [14]. Many people seek HIV services with a fee for service in PFP and PNFP facilities and factors associated with this include distance to service provided, perceived better quality of service and avoidance of long waiting times [15]. Health Insurance Data on health insurance is challenging to come by. In 2004 WHO estimated 0.2 % of total health expenditure was on private medical insurance [16]. Estimates from 2008 suggest that only 0.5 % of the population in Uganda has any health insurance coverage [12]. Private health insurance schemes cover only 0.2 % of the population and are largely available only for people in formal employment whose employers subsidize the cost [17]. The 30 registered community insurance schemes in Uganda are mostly health facility-based schemes in which premiums are paid by individual community members. These schemes cover only an estimated additional 100,000 people (about 0.3 % of the current population) [8]. Current Sources of funding and future options for HIV/AIDs services in Uganda Government funding In FY 2012/2013 the GOU spent 7.4 % of its total annual budget on financing health systems, in particular human resources in the health sector, on which the delivery of HIV/AIDs services relies heavily. Of this only about 3 % of the health budget is for targeted to HIV/AIDS programmes, but the remainder of this funding indirectly supports HIV programmes through government health systems strengthening. Policy innovations to increase the budget for health (and specifically for HIV services) are being considered such as an HIV tax on selected services and/or a cash transfers. The President’s Emergency Plan for AIDS in Africa (PEPFAR) funding The United States government Presidents Emergency Fund for AIDS Relief (PEPFAR) funded programs invested over $1.8 billion in HIV-related financing for Uganda between 2004 and 2011. PEPFAR currently contributes over 85 % of the national HIV response through projects implemented by international and local organizations (PNFP organizations) which provide free HIV/AIDS services or significantly support such services at public facilities. In practice, this means PNFPs housed within government health facilities, but running a siloed service, with directly employed health care workers but reporting to the national Ministry of Health system through the Health Information Management Service (HMIS). PEPFAR funding has already plateaued and PEPFAR now requires “cost-sharing assurances” from governments [18]. As governments assume more cost-sharing commitments, PEPFAR plans to more actively cede control over priority-setting and accounting for results to partner countries. Global Fund (GF) and other funders Since its inception, the Global Fund to Fight AIDS, Tuberculosis and Malaria (Global Fund) has disbursed over USD 230 million to support various aspects of HIV/AIDS programs in Uganda through government and civil society mechanisms [19]. While the Global Fund has mainly utilized the mainstream government channels for delivering its support for the HIV effort, a substantial part of downstream resources that it provides (particularly ARVs) are delivered either directly through PNFPs or with significant support from PNFPs at government facilities [20]. The Global Fund has prioritized lower income, high burden countries and streamlined the funding application process through its new funding mechanism to make funding more predictable and flexible [21]. Nonetheless, developed country contributions to the Global Fund and other funding mechanisms are on a downward trend, triggering similar cutbacks in funding to many countries’ HIV programs. This signals potential exposure to the risk of a serious financing gap particularly for PNFPs providing free HIV services to the public. The case for voluntary co-pay: Can Ugandans help each other to pay for HIV care? A National Health Insurance scheme that integrates all available external, private and public providers and financing agencies would create the opportunity reduce administrative costs, create less fragmented, full service provision points, and provide the impetus for introduction of comprehensive cross-subsidized contributory financing for health in general and HIV services in particular. However, in reality this may not be developed in time to fill the gap between drop in multi-lateral donations and a definite government commitment to funding for HIV services. As PNFP we are front line providers of HIV care. As such our organization as well as others will feel the early effects of a reduction or plateauing in funding, and will be unable to expand our services further, or will need to cut back. We believe that the option that may be most expedient for PNFP providers such as ourselves to continue to provide care in the short term is development of voluntary co-pay services. Uganda has an improving economy with a rising middle class; the number of Ugandans in the middle class (defined as those whose ability to meet their basic needs is relatively stable and who can afford to save for the future) increased from1.8 million 1992 to 12.6 million in 2012 (10 to 37 % of the population) [22]. In Uganda the 2011 AIDS indicator survey found that HIV prevalence increases with higher socio-economic status [23], and many of these PLHIV will access care in the private sector due to reduced waiting times and longer opening hours. However, the private health sector which already charges a fee for service, is a profit making sector, which has no obligation to help the lower income strata. In addition, as they are independent of the government system there is no quality control on their care [24], records are often absent and retention in care is poor [25]. We suggest that as a short term measure to mitigate against the emerging risk of financing gaps, PNFPs could explore the introduction of voluntary co-pay schemes so that those who can afford pay for services that can help subsidize the cost for those who cannot afford to pay. However, previous experience with co-pay systems both within the general health service suggest that they may affect utilization of services, as in Uganda, their abolition in 2001 increase attendance at health clinics [26]. Of particular concern in HIV care is lack of adherence to ART drugs and development of drug resistance, which has been noted when patients are paying for care [27, 28]. Therefore, in order mitigate against these issues, and to maintain a reasonable level of equity, we propose that such schemes could differentiate services through convenience and accessibility, whilst maintaining essential services at an acceptable standard for both groups of clients. For example, patients pay a standard cost for consultation in order to see the same doctor at each appointment or to access care at weekends or evenings, but are not charged for tests or medication. At the Infectious Diseases Institute in Kampala we are currently studying the feasibility of this concept by piloting a co-pay component to our large, multidisciplinary HIV clinic. A routine customer care survey of 387 of our 8500 patients suggested that over 40 % would be willing to pay a fee for greater convenience with longer (after work) opening hours and shorter waiting times. Qualitative research in our clinic has shown general acceptability of some patients for co-pay services for convenience, and also a willingness of those patients to contribute to the cost of care for other patients with less resources [29] and preliminary data suggests that access to a co-pay clinic improves retention to care in those who were lost to follow up within the private sector [30]. However, we will need to determine if this model is actually cost saving and can contribute funds to support sustaining of services, as previous experience with user fees in Africa suggest administration costs can be greater than fees collected [31]. Conclusion We are concerned that current funding mechanism for HIV programs in Uganda is not sustainable and as service providers we are keen to explore ways in which provide lifelong HIV care to as many people living with HIV (PLHIV) as possible. In reality, it is likely that only a model that combines several different financing approaches will provide the necessary care and treatment for our PLHIV in the future. Until such time as the Ugandan economy can support universal, state-supported, comprehensive healthcare bridging alternatives must be considered. We suggest that offering patients with the sufficient means to assume some of the financial burden for their care in return for more convenient services could be one component of preparing for a “Post-PEPFAR” funding environment. We urge other care providers to share their ideas and experiences in sustaining services so that we can learn from best practices of others in maintaining and expanding quality HIV services in resource limited settings. Acknowledgements We would like to thank the patients and staff of the Infectious Diseases Clinic, Kampala, Uganda. Funding This work was funded by a grant to Dalhousie University by Grand Challenges Canada grant number 2078-01. Availability of data and materials Not applicable. Authors’ contributions TK – background data collection, manuscript drafting, RPR – background data checking, manuscript review and revisions, WS, RB, AC – manuscript review and revisions. All authors read and approved the final manuscript. Authors’ information TK is head of grants and strategic planning at IDI in Kampala. RPR is the immediate past head of clinical services at the Infectious Diseases Institute (IDI), Kampala, Uganda. RB is the current Executive Director of IDI and AC the immediate past ED. WS is a Professor of Infectious Diseases at Dalhousie University and has served as a visiting professor to IDI for over 10 years. The IDI has been involved in care and treatment of patients living with HIV since 2004 and is involved directly or indirectly in care for over 10 % of all patients on HIV treatment in Uganda. Competing interests Rosalind Parkes-Ratanshi has received funding for consultancy for Janssen Global Public Health. No other authors declare competing interests. Consent for publication Not applicable. Ethics approval and consent to participate Not applicable. ==== Refs References 1. Dieleman JL Graves C Johnson E Templin T Birger M Hamavid H Freeman M Leach-Kemon K Singh L Haakenstad A Sources and Focus of Health Development Assistance, 1990–2014 JAMA 2015 313 23 2359 2368 10.1001/jama.2015.5825 26080340 2. UNAIDS THJKF Financing the Response to HIV in Low- and Middle-Income Countries: International Assistance from Donor Governments in 2014 2015 3. Amico P Aran C Avila C HIV spending as a share of total health expenditure: an analysis of regional variation in a multi-country study PLoS One 2010 5 9 0012997 10.1371/journal.pone.0012997 4. UNAIDS. Global report: UNAIDS report on the global AIDS epidemic 2013. In., vol. http://www.unaids.org/sites/default/files/media_asset/UNAIDS_Global_Report_2013_en_1.pdf. Geneva: Joint United Nations Programme on HIV/AIDS (UNAIDS); 2013. p.198. 5. UNAIDS. Update on HIV treatment 2013: Results, Impact and Opportunities. In. Geneva: Joint United Nations Programme on HIV/AIDS; 2013. 6. WHO. The World Health Report: health systems financing: the path of universal coverage. In., vol. http://www.who.int/whr/2010/en/. Geneva: World Health Organization (2010); 2010. p.20. 7. Eaton JW Menzies NA Stover J Cambiano V Chindelevitch L Cori A Hontelez JA Humair S Kerr CC Klein DJ Health benefits, costs, and cost-effectiveness of earlier eligibility for adult antiretroviral therapy and expanded treatment coverage: a combined analysis of 12 mathematical models Lancet Global Health 2014 2 1 e23 34 10.1016/S2214-109X(13)70172-4 25104632 8. Basaza RK Criel B Van der Stuyft P Community health insurance amidst abolition of user fees in Uganda: the view from policy makers and health service managers BMC Health Serv Res 2010 10 33 1472 6963 9. Health Mo: Statistical Abstract 2010. http://www.ubosorg/onlinefiles/uploads/ubos/pdf%20documents/PNSD/2010MOHStatAbstpdf 2010. Accessed 15 Dec 2015. 10. Hecht R Stover J Bollinger L Muhib F Case K de Ferranti D Financing of HIV/AIDS programme scale-up in low-income and middle-income countries, 2009-31 Lancet 2010 376 9748 1254 1260 10.1016/S0140-6736(10)61255-X 20934597 11. UCBHFA. Strategic Plan for the Uganda Community Based Health Financing Association (UCBHFA) 2013/2014-2017/2018 In., vol. http://ucbhfa.org/wp-content/uploads/2010/12/Strategic-Plan-.df_.pdf. Kampala: UCBHFA; 2013. p.29. 12. Basaza RK O’Connell TS Chapcakova I Players and processes behind the national health insurance scheme: a case study of Uganda BMC Health Serv Res 2013 13 357 1472 6963 13. Orem JN Zikusooka CM Health financing reform in Uganda: How equitable is the proposed National Health Insurance scheme? Int J Equity Health 2010 9 23 1475 9276 14. MOH National Health Accounts FY 2008/09 and FY 2009/10 : general health accounts, reproductive health sub-accounts, and child health sub-accounts 2013 Kampala Ministry of Health 79 15. Rhoda K Wanyenze GW Alamo S Amanyire G Ouma J Kwarisima D Sunday P Wabwire-Mangen F Kamya M Evaluation of the Efficiency of Patient Flow at Three HIV Clinics in Uganda AIDS Patient Care STDs 2010 24 7 441 446 10.1089/apc.2009.0328 20578908 16. WHO Private health insurance: Implications for developing countries 2004 17. Charlotte M Zikusooka RK Private medical pre-payment and insurance schemes in Uganda: What can the proposed SHI policy learn from them? 2008 18. PEPFAR. Uganda Operational Plan Report. FY 2010. In., vol. http://www.pepfar.gov/documents/organization/145738.pdf. Kampala, Uganda: PEPFAR; 2010. 19. Coutinho A, Roxo U, Epino H, Muganzi A, Dorward E, Pick B. The expanding role of civil society in the global HIV/AIDS response: what has the President’s Emergency Program For AIDS Relief’s role been? J Acquir Immune Defic Syndr. 2012;60:152–157. 20. MOH. Status report on the global fund grants in Uganda. In., vol. http://www.health.go.ug/MoH/docs/global_fund.pdf. Kampala: Ministry of Health; 2011. p 32. 21. Globalfund New Funding Model: Eligibility, Counterpart Financing and Prioritization Policy Revision Thirtieth Board Meeting GF/B30/6 7-8 November 2013 2013 2013 Geneva, Switzerland Global Fund 22. MOF Poverty Status Report (PSR) 2012 2012 23. Health UMo AIDS Indicator Survey 2011 24. USAID AIDSTAR-One Project: Private Sector Involvement in HIV 2009 25. Project UA Assessment of quality of ART services in the private sector in Uganda 2009 26. Burnham GM Pariyo G Galiwango E Wabwire-Mangen F Discontinuation of cost sharing in Uganda Bull World Health Organ 2004 82 3 187 195 15112007 27. Byakika-Tusiime J Oyugi JH Tumwikirize WA Katabira ET Mugyenyi PN Bangsberg DR Adherence to HIV antiretroviral therapy in HIV+ Ugandan patients purchasing therapy Int J STD AIDS 2005 16 1 38 41 10.1258/0956462052932548 15705271 28. Alsan M Schoemaker L Eggleston K Kammili N Kolli P Bhattacharya J Out-of-pocket health expenditures and antimicrobial resistance in low-income and middle-income countries: an economic analysis Lancet Infect Dis 2015 15 10 1203 1210 10.1016/S1473-3099(15)00149-8 26164481 29. Adelline T King R Schlech WF Faridah M Kakaire T Rosalind Parkes-Ratanshi P Exploring Attitudes and Perceptions of Patients and Staff Towards a Fee For Service “After Hours” Clinic Supplementing Free Hiv Services in Uganda: A Qualitative Study Value Health 2015 18 7 A590 10.1016/j.jval.2015.09.1516 26533314 30. Parkes-Ratanshi R, Mukisa G, Kakaire T, Mayanja F, Tumikye A, Mitchell B, Nakalema S, Schlech W. Characteristics and outcomes of patients seeking care at a new “co-pay” convenience clinic established to explore sustainable funding models in Uganda. In: Conference on Retroviruses and Opportunistic Infections February 23-26, Seattle, Washington; 2015. Available at http://www.croiconference.org/sessions/characteristics-and-outcomes-patients-seeking-care-new-%E2%80%9Cco-pay%E2%80%9D%C2%9D-convenience-clinic. 31. Gilson L The lessons of user fee experience in Africa Health Policy Plan 1997 12 4 273 285 10.1093/oxfordjournals.heapol.a018882 10176263

PMC005xxxxxx/PMC5002096.txt

==== Front BMC PediatrBMC PediatrBMC Pediatrics1471-2431BioMed Central London 68010.1186/s12887-016-0680-0Research ArticleRefractory wheezing in Chinese children under 3 years of age: bronchial inflammation and airway malformation Gu Wenjing guwenjing999@126.com 1Jiang Wujun eggjwjsjw@163.com 1Zhang Xinxing zhangnewstar@126.com 1Chen Zhengrong chen_zheng_rong@163.com 1Yan Yongdong yyd3060@163.com 1Huang Li szdalv@163.com 1Wang Meijuan wangmeijuan812@sina.com 1Shao Xuejun xuejunshao@hotmail.com 2Wang Shuhui wangshuhui158158@163.com 1Ji Wei +86 0512 80698303szdxjiwei@163.com 11 Department of Respiration, Children’s Hospital of Soochow University, Suzhou, 215003 China 2 Department of Clinical laboratory, Children’s Hospital of Soochow University, Suzhou, 215003 China 27 8 2016 27 8 2016 2016 16 1 14512 5 2015 16 8 2016 © Gu et al. 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Background Wheezing is a common symptom in early childhood. However, refractory wheezing is difficult to treat, and it may thus account for extensive use of medical resources. It is therefore important to improve our understanding of the pathophysiology of refractory childhood wheezing. Methods In this descriptive study, we studied 156 children with refractory wheezing using fiberoptic bronchoscopy and bronchoalveolar lavage (BAL), and compared the results with a control group of 46 children with various pulmonary diseases but no wheezing. Etiology and cell classification were analyzed for each BAL sample. Results Overall, 21.8 % of children with refractory wheezing had airway malformations including tracheomalacia, airway stenosis, and tracheal bronchus. The incidence of airway malformations increased to 31 % in infants under 12 months of age. A significant increase in neutrophil ratio and decrease in macrophage ratio were observed in BAL from children with refractory wheezing compared with controls. Pathogen infection led to a higher ratio of neutrophils in the wheezing group compared with controls. However, there were no significant differences in neutrophil ratios among children with various pathogen infections. Furthermore, children with refractory wheezing had a high rate of Mycoplasma pneumoniae infection. Conclusions Airway malformations might play an important role in children under 3 years of age with refractory wheezing, especially in infants under 12 months of age. Neutrophil-mediated airway inflammation was characteristic of refractory wheezing in children under 3 years of age. In addition, infections such as M. pneumoniae may aggravate airway inflammation and affect refractory wheezing. Keywords Bronchoalveolar lavageNeutrophil-mediated inflammationInfantRefractory wheezingAirway malformationissue-copyright-statement© The Author(s) 2016 ==== Body Background Wheezing in early childhood is a common but poorly characterized symptom, with a third of infants reported to experience multiple episodes of wheezing in their first 3 years [1]. Lower-airway inflammation, especially neutrophil-mediated inflammation, occurs in young children with recurrent wheezing [2, 3]. Compared with recurrent wheezing, refractory wheezing is more difficult to treat effectively and often accounts for extensive use of medical resources. It is therefore important to improve our understanding of the pathophysiology of refractory wheezing. Congenital malformations of the lungs and airways are among several causes of irreversible airway obstruction in children who may develop various symptoms such as recurrent wheezing, cough, recurrent lower airway infection, severe dyspnea, and respiratory insufficiency [4–7]. We speculate that airway morphology may play an important role in refractory wheezing. Fiberoptic bronchoscopy (FB) and bronchoalveolar lavage (BAL) are indispensable techniques for investigating pediatric patients with airway abnormalities and pulmonary infiltrates, and both are carried out as routine procedures in many health centers [8]. The aim of this study was to determine the cellular profile of BAL from infants with refractory wheezing in the Suzhou area, China, and to investigate airway malformations and neutrophil ratios to determine if ongoing inflammation plays a role in the development of this condition. Methods Inclusion and exclusion criteria This descriptive, retrospective study enrolled children under 3 years of age with refractory wheezing, defined as persistent wheezing requiring at least 4 weeks of oral corticosteroid treatment after poor responses to an initial inhaled combination of corticosteroids and bronchodilators. Additionally, the symptom-free time during which children had no wheezing symptoms was no longer than 3 days in the study population. Children were excluded from the study group if they met any of the following conditions: 1) wheezing symptoms lasted for <1 month; 2) had an apparent recovery period; 3) were severely sick and unable to tolerate FB; 4) had a family history of smoking; and 5) were premature or low birthweight babies. All the participants’ parents or guardians gave their written informed consent for participation in the study. The study was approved by the ethics committee of Soochow University. Children with wheezing Based on the inclusion and exclusion criteria, 156 children with wheezing who were hospitalized in the Children’s Hospital of Soochow University, China, from September 2011 to May 2014 were chosen for this study. All included patients received oral steroids. A wheezing infant was considered to be atopic if they had a strong family history of atopy (two or more direct relatives with atopy), atopic dermatitis, or a prior positive allergen test consisting of a prick test with an allergen on the skin or a radioallergosorbent test. Control group There are ethical problems associated with conducting BAL cell analysis in healthy children. The control group therefore consisted of children sampled during the same period with the following conditions, in which the BAL cell analysis was relatively normal: 1) recurrent bronchitis without wheezing (n = 11); 2) persistent pulmonary atelectasis for at least 1 month (n = 3); 3) suspicion of foreign-body aspiration (n = 7); 4) persistent cough and final diagnosis of psychogenic cough (n = 12); 5) persistent laryngeal stridor with suspected laryngomalacia (n = 10); or 6) hemoptysis with suspected lung disease and final diagnosis of epistaxis (n = 3). Children in the first three groups underwent FB during remission. Children were excluded from the control group if they met any of the following criteria: 1) history of wheezing; 2) endoscopic bronchial inflammation; 3) endoscopic airway malformation; and/or 4) personal or familial history of atopy. Initial evaluation Demographic, clinical, and laboratory data were collected after admission. White blood cell count, platelets, neutrophil ratio, and lymphocyte ratio were tested within 4 h after admission. Fiberoptic bronchoscopy Parents/guardians were informed about the surgical risks of FB and signed consent was obtained prior to the procedure. Children were fasted for both solids and liquids for at least 6 h prior to the procedure. Intramuscular atropine sulfate 0.01–0.02 mg/kg and midazolam 0.2–0.4 mg/kg were administered as premedication. Upper and lower airway anesthesia was achieved with 2 % lidocaine. A flexible bronchoscope (Olympus CV260, Tokyo, Japan or Fujinon EB-270P, Miyoshi, Japan) was wedged into each lobe. The airways were washed three times with 1 ml/kg of prewarmed sterile 0.9 % saline solution, which was then collected by a sterile sputum-collecting pipe (Falcon 50 ml, Becton-Dickinson, Rutherford, NJ, USA). The collected BAL was used for cell counts, viral analysis, and microbiological analysis (bacteria and Mycoplasma pneumoniae). Cell counts Differential cell counts were obtained using a modified version of Wright–Giemsa staining (Wright–Giemsa Stain, Baso Diagnostics Inc., Taiwan, China). At least 500 cells were examined for each specimen. The ratios of various cell types in total cell counts were reported. Microbiological analysis BAL samples from wheezing and control children were tested for 10 types of viruses and bacteria, as well as M. pneumoniae. Bacteria were tested by inoculating BAL onto blood plates and examining them after incubation for 18–20 h. Bacterial growth >103 colony-forming units/ml was considered significant. Viruses, including respiratory syncytial virus, adenovirus, influenza virus (A, B), and parainfluenza virus (1, 2, 3) were investigated by immunofluorescence tests using D3 Ultra Respiratory Virus Screening and LD Kit (Diagnostic Hybrids, OH, USA). Positive results were defined as more than five inclusion bodies detected under a fluorescence microscope. M. pneumoniae and viruses including rhinovirus, human metapneumovirus, and bocavirus were examined by polymerase chain reaction using a Nucleic Acid Amplification Fluorescent Reagent Kit (Ann Gene Co., Guangdong, China) according to the manufacturer’s instructions. Statistical analysis Cell counts were presented as the mean ± standard deviation (SD) and as medians (25 % to 75 %). The wheezing and control groups were compared using nonparametric Mann–Whitney two-sample U-tests, and Mann–Whitney U-tests were used for unpaired data. χ2 tests were used for categorical variables. A P value < 0.05 was considered statistically significant. Results Demographic information Overall, 5830 children under 3 years old were hospitalized because of wheezing between September 2011 and May 2014. Among these, 356 had refractory wheezing for at least 4 weeks (6.1 %), 54 had an apparent recovery period, 120 had a family history of smoking, and 32 were premature or low-birthweight babies. According to the exclusion criteria, 156 children with refractory wheezing were ultimately enrolled in the study (age range, 3–36 months; mean ± SD, 13.68 ± 7.08 months), of whom 121 (77.6 %) were male. The control group was aged 1–36 months (mean ± SD, 13.08 ± 10.56 months), including 35 male children (76.1 %). The age, sex ratio, and weight were similar in the wheezing and control groups (P > 0.05). The white blood cell count was significantly higher in the wheezing group compared with the control group (11.44 ± 4.97 vs 10.39 ± 4.28, P < 0.001). The neutrophil ratio was significantly higher and the lymphocyte ratio significantly lower in the wheezing group compared with the controls (36.82 ± 14.97 vs 29.00 ± 13.47 and 53.38 ± 14.38 vs 60.31 ± 13.22, respectively, P < 0.001). The blood platelet count was significantly higher in the wheezing group compared with the controls (394.44 ± 120.19 vs 381.48 ± 131.09, respectively, P < 0.001) (Table 1).Table 1 Demographic data for children with refractory wheezing and controls Wheezing group (n = 156) Control group (n = 46) P value Age, months 13.68 ± 7.08 13.08 ± 10.56 0.262 Sex, M/F 121/35 35/11 0.843 Weight, kg 10.56 ± 2.13 9.74 ± 3.72 0.115 Breastfeeding, Y/N 98/58 35/11 0.095 Whole blood cell analysis  White blood cell, ×109 11.44 ± 4.97 10.39 ± 4.28 <0.001  Neutrophils, % 36.82 ± 14.97 29.00 ± 13.47 <0.001  Lymphocytes, % 53.38 ± 14.38 60.31 ± 13.22 <0.001  Blood platelet, ×109 394.44 ± 120.19 381.48 ± 131.09 <0.001 Fiberoptic bronchoscopy findings FB and BAL were well tolerated by both groups of children, and no major complications were observed. Thirty-four children (21.8 %) in the wheezing group had airway malformations including tracheomalacia (n = 24), airway stenosis (n = 6), and tracheal bronchus (n = 7). Among these 34 had airway malformations, 28 (82.4 %) were male and 27 (79.4 %) were under 12 months of age. We further studied the characteristics of the airway inflammation in the wheezing group. A previous study reported that airway malformation alone could cause refractory childhood wheezing [4], and that these children might not have typical airway inflammation. We therefore analyzed the cellular contents of BAL from children with refractory wheezing without airway malformation and compared the results with the control group. We also performed a subject-to-subject comparison in the wheezing group. The median recovery rate was 70 % (range, 45 %–80 %), with very little variation between the two compared groups. However, the recovered cell ratio showed a wide distribution in the wheezing group (Table 2).Table 2 BAL cellular contents in children with refractory wheezing and controlsa Refractory wheezing children without airway malformation (n = 122) Control group (n = 46) P value Alveolar macrophages, % 60.80 ± 29.20, 67 (35.0 ~ 88.25) 87.06 ± 8.22, 90 (82.5 ~ 93.0) <0.001 Lymphocytes, % 7.20 ± 6.46, 5 (2.0–10.0) 6.48 ± 5.52, 5 (2.0 ~ 8.0) 0.314 Neutrophils, % 31.26 ± 28.10, 25 (6.0 ~ 49.25) 6.38 ± 6.59, 4 (2.0 ~ 9.0) <0.001 Eosinophils, % 0.75 ± 2.76, 0.0 (0.0 ~ 0.0) 0.08 ± 0.27, 0.0 (0.0 ~ 0.0) 0.040 aData presented as mean ± SD, median (25th to 75th percentiles) Comparison between wheezing and control subjects The neutrophil ratio was significantly higher (31.26 ± 28.10 % vs 6.38 ± 6.59 %, respectively, P < 0.001) and the macrophage ratio was significantly lower (60.80 ± 29.20 % vs 87.06 ± 8.22 %, respectively, P < 0.001) in the wheezing group compared with the control group. The eosinophil ratio was also higher in the wheezing group (0.75 ± 2.76 % vs 0.08 ± 0.27 %, respectively, P = 0.040). There was no significant difference in BAL lymphocyte ratio between the two groups. Comparisons within the wheezing group Eighty-one of 122 (66.4 %) wheezing children were atopic and had no airway malformations. There was no significant difference in the ratio of macrophages, lymphocytes, eosinophils, or neutrophils in the BAL between atopic and nonatopic children in the wheezing group (Table 3).Table 3 BAL cellular content in atopic and nonatopic childrena Variables Atopic children (n = 81) Nonatopic children (n = 41) P value Alveolar macrophages, % 56.86 ± 30.28, 61.0(33.5 ~ 85.5) 68.56 ± 25.60, 77.0 (53.5 ~ 90.0) 0.052 Lymphocytes, % 7.75 ± 7.28, 5.0 (2.0 ~ 12.0) 6.10 ± 4.30, 5.0 (2.0 ~ 10.0) 0.366 Neutrophils, % 34.89 ± 29.61, 30.0 (7.0 ~ 54.0) 24.10 ± 23.57, 15.0 (5.0 ~ 35.5) 0.062 Eosinophils, % 0.51 ± 1.28, 0.0 (0.0 ~ 0.0) 1.24 ± 4.40, 0.0 (0.0 ~ 0.0) 0.305 aData presented as mean ± SD, median (25th to 75th percentiles) In terms of pathogen analysis, we performed viral analysis, and M. pneumoniae and microbiological cultures for the wheezing and control groups. No pathogens were detected in the control group. Virus detection was positive in 17 (13.9 %) of the wheezing children without airway malformations, and microbiological findings were positive in 78 (63.9 %). Among them, M. pneumoniae showed the highest detection rate (n = 63, 51.6 %). In addition, other bacteria were detected in 30 wheezing children (24.6 %), including Haemophilus influenzae (n = 12), Streptococcus pneumoniae (n = 12), Escherichia coli (n = 2), Burkholderia cepacia (n = 1), Pseudomonas aeruginosa (n = 1), Serratia marcescens (n = 1), and Enterobacter aerogenes (n = 1). We further investigated how the BAL cellular content was affected by pathogens by categorizing wheezing children into three groups: 1) non-infection group (N, n = 40), no pathogens detected; 2) bacterial-infection group (B, n = 78), bacterial infection detected (including M. pneumoniae); and (3) virus-infection group (V, n = 17), virus infection detected. We then compared the BAL cellular content between each of these groups and the control group (C, n = 46) (Fig. 1). The lowest neutrophil ratio was found in the control group, followed by the non-infection group, with the highest neutrophil ratios in the infection groups. There was no significant difference among the various bacterial ratios. The highest alveolar macrophage ratio was found in the control group, and the lowest in the infection groups, while the eosinophil ratio was highest in the infection groups. Of note, there was no significant difference in lymphocyte ratios among the four groups.Fig. 1 BAL cellular content. The percentages of neutrophils (a), alveolar macrophages (b), lymphocytes (c) and eosinophils (d) in the BAL in the control (C, n = 46), non-pathogen (N, n = 40), bacterial-infection (including M. pneumoniae) (B, n = 78), and virus-infection groups (V, n = 17). Horizontal bars indicate the median for each group of patients Discussion Wheezing is a common clinical symptom in infants. Its long disease course, slow recovery, and risk of relapse mean that refractory wheezing is difficult to treat, resulting in high medical-resource utilization. In our department, 6.1 % of children under 3 years of age had refractory wheezing for at least 4 weeks. Although various treatments have been used, including inhaled and systemic corticosteroids, and montelukast, they have little effect on preventing the recurrence of refractory wheezing [9]. These medications may influence eosinophilic airway inflammation but have limited impact on noneosinophilic airway inflammation. It has therefore been speculated that refractory wheezing may have a different pathophysiologic mechanism. Airway malformation is one cause of irreversible airway obstruction in children and is accompanied by many clinical symptoms [10]. The incidence of congenital pulmonary airway malformation is one per 8300–35,000. It usually affects a single lobe and has no sex bias [11]. In this study, 21.8 % of children with refractory wheezing had airway malformations including tracheomalacia, airway stenosis, and tracheal bronchus, and 27 of 87 (31 %) wheezing children under 12 months of age had airway malformations, suggesting that airway malformations may play an important role in refractory wheezing. A literature search [12, 13] revealed differing total BAL cell counts, though they generally fell within the range 0.5–57.1 × 104/ml. Of the total BAL cells, >85 % are alveolar macrophages, <12 % are lymphocytes, <2 % are neutrophils, and <1 % are eosinophils [8, 10]. The classification of BAL cells in pathological conditions varies, and BAL analysis can provide important reference information on the nature of lung diseases, the inflammation process, and disease activity. BAL fluid analysis has recently been used globally for diagnostic and prognostic purposes, and to observe the curative effect of respiratory disease treatments. However, although it is theoretically important to study the cellular profile of BAL in healthy children, this is ethically difficult. We therefore used BAL from relatively normal children as a control, as in previous studies [3]. BAL cell analysis in our control group was comparable to that reported for healthy children [3], suggesting that the results in the control group were representative of the endotracheal cell ratio in a normal child. In the current study, we compared the BAL inflammatory cell profiles between 122 children with severe refractory wheezing and 46 control children. The BAL neutrophil ratio was significantly increased and the macrophage ratio was significantly decreased in the wheezing group compared with controls, suggesting that granulocytes contribute to airway inflammation. The mechanism of neutrophil influx and activation may be mediated by interleukin (IL)-8 secretion [14]. Studies on neutrophilic inflammation and persistent asthma by Gibson et al. revealed that sputum IL-8 levels were higher in patients with noneosinophilic asthma [15]. Neutrophils showed a close correlation with IL-8 levels [15]. Similarly, neutrophil levels were significantly higher in 83 wheezing infants compared with 10 control subjects [3]. Marguet et al. compared BAL cell profiles in children with asthma, infantile wheeze, chronic cough, and cystic fibrosis, and reported a significantly higher level of neutrophils in wheezing infants [2]. Previous studies indicated that airway virus infection and bacterial colonization can increase the risk of wheezing [16, 17]. In this study, virus detection was positive in 17 children with refractory wheezing and no airway malformation, but who had the highest bocavirus ratios. M. pneumoniae was positive in 63 wheezing children, and other culture findings were positive in 30 wheezing children. The M. pneumoniae infection rate was high in children with refractory wheezing in the current study, and previous studies also suggested that M. pneumoniae could cause childhood wheezing [18, 19]. However, the M. pneumoniae status in infants is often ignored in clinical practice. These results suggest that children with refractory wheezing should be prescribed macrolides to treat M. pneumoniae infection and help reduce clinical symptoms. A high level of neutrophils in the BAL might correlate with infection. The relationships between various BAL cells counts and infection indicated a significantly higher neutrophil ratio and lower alveolar macrophage ratio in children with refractory wheezing compared with the control group, but this result was not affected by pathogen infection. However, children with a pathogenic infection had a higher neutrophil ratio and lower alveolar macrophage ratio than those without positive pathogen detection. We speculated that neutrophil-mediated chronic inflammation in the airways may also play an important role in refractory wheezing, regardless of pathogen infection, though infection may potentially aggravate the airway inflammation. Irrespective of the pathogens, the neutrophil ratio was always higher in infected compared with uninfected children. Furthermore, there was no difference in neutrophil numbers between those infected with bacteria (including M. pneumoniae) or virus. Bacteria, viruses, and M. pneumoniae have been shown to increase the BAL neutrophil ratio. A previous study [20] found that BAL neutrophilia was associated with bacterial pulmonary infection. The neutrophil ratio has shown a tendency to increase with the occurrence of viral infection [21, 22]. This may be correlated with IL-8 and leukotriene B4, which favor the recruitment of neutrophils in airways [23–25]. M. pneumoniae has been studied intensively in recent years, and children with M. pneumoniae infection had an increased BAL cell count, which was attributed to an increase in neutrophils [26]. In contrast, other previous studies indicated that the BAL neutrophil ratio was not correlated with bacteriological results, implying that neutrophil-mediated inflammation was independent of bacterial infection [2, 3, 27] in wheezy infants. Cellular BAL levels may be related to allergy. In our study, wheezing children were further categorized as atopic or nonatopic, and there were no significant differences in cell counts (macrophages, lymphocytes, eosinophils, and neutrophils) between the atopic and nonatopic groups. This was similar to the results of Le Bourgeois et al. [3], who observed normal BAL cell levels in atopic children. However, other studies revealed that children with atopy had higher eosinophil levels [2]. Marguet et al. [2] reported that eosinophils, which are characteristic of asthma, were rarely seen in wheezing infants. However, normal eosinophil levels do not exclude the possibility of their involvement in the physiopathology of atopic wheezing, and further studies are needed to examine the roles of activated and degranulated eosinophils. Eosinophil cationic protein (ECP) is released by activated eosinophils and is used as a marker of eosinophilic inflammation. High ECP levels have been reported in patients with active asthma and other allergic diseases [28]. The detection of ECP levels in BAL would thus be valuable [29]. Conclusions It is important to identify the causes associated with the pathogenesis of refractory wheezing in children. In this descriptive study, some children with refractory wheezing had airway malformations, with children under 12 months of age having a particularly high proportion of airway anomalies. Neutrophil-mediated airway inflammation might also play an important role in the pathophysiology of refractory wheezing, while infections might aggravate airway inflammation. The bacterial-detection rate (especially M. pneumoniae) was high among wheezing children. Further research on infants with refractory wheezing is needed to develop specific inflammatory markers. Abbreviations BALBronchoalveolar lavage ECPEosinophil cationic protein FBFiberoptic bronchoscopy We gratefully acknowledge the valuable cooperation of Mrs. Xiuping Gu and the members of her team in helping us accomplish fiberoptic bronchoscopy and collecting bronchoalveolar lavage. Funding This work was supported by the National Natural Science Foundation of China (Grant to Wei Ji, No. 81570016). Authors’ contributions WJ, ZC and WG conceived and designed the experiments. XS, LH, MW and YY performed the experiments. XZ and SW analyzed the data. WG, LH and WJ drafted the manuscript. All authors read and approved the final manuscript. Competing interests The authors declare that they have no competing interests. ==== Refs References 1. Schwerk N Brinkmann F Soudah B Kabesch M Hansen G Wheeze in preschool age is associated with pulmonary bacterial infection and resolves after antibiotic therapy PLoS One 2011 6 10.1371/journal.pone.0027913 22140482 2. Marguet C Jouen-Boedes F Dean TP Warner JO Bronchoalveolar cell profiles in children with asthma, infantile wheeze, chronic cough, or cystic fibrosis Am J Respir Crit Care Med 1999 159 1533 40 10.1164/ajrccm.159.5.9805028 10228122 3. Le Bourgeois M Goncalves M Le Clainche L Benoist MR Fournet JC Scheinmann P Bronchoalveolar cells in children < 3 years old with severe recurrent wheezing Chest 2002 122 791 7 10.1378/chest.122.3.791 12226015 4. Carden KA Boiselle PM Waltz DA Ernst A Tracheomalacia and tracheobronchomalacia in children and adults: an in-depth review Chest 2005 127 984 1005 10.1378/chest.127.3.984 15764786 5. Wood RE Localized tracheomalacia or bronchomalacia in children with intractable cough J Pediatr 1990 116 404 6 10.1016/S0022-3476(05)82833-1 2308032 6. Irwin RS Madison JM The persistently troublesome cough Am J Respir Crit Care Med 2002 165 1469 74 10.1164/rccm.2110097 12045118 7. McNamara VM Crabbe DC Tracheomalacia Paediatr Respir Rev 2004 5 147 54 10.1016/j.prrv.2004.01.010 15135125 8. Efrati O Sadeh-Gornik U Modan-Moses D Barak A Szeinberg A Vardi A Flexible bronchoscopy and bronchoalveolar lavage in pediatric patients with lung disease Pediatr Crit Care Med 2009 10 80 4 10.1097/PCC.0b013e31819372ea 19057431 9. Beigelman A, Isaacson-Schmid M, Sajol G, Baty J, Rodriguez OM, Leege E, et al. Randomized trial to evaluate azithromycin's effects on serum and upper airway IL-8 levels and recurrent wheezing in infants with respiratory syncytial virus bronchiolitis. J Allergy Clin Immunol. 2014. 10. Boogaard R Huijsmans SH Pijnenburg MW Tiddens HA de Jongste JC Merkus PJ Tracheomalacia and bronchomalacia in children: incidence and patient characteristics Chest 2005 128 3391 7 10.1378/chest.128.5.3391 16304290 11. Nadeem M Elnazir B Greally P Congenital pulmonary malformation in children Scientifica (Cairo) 2012 2012 209896 24278678 12. de Blic J Midulla F Barbato A Clement A Dab I Eber E Bronchoalveolar lavage in children. ERS Task Force on bronchoalveolar lavage in children. European Respiratory Society Eur Respir J 2000 15 217 31 10.1183/09031936.00.15121700 10678650 13. Ratjen F Bredendiek M Brendel M Meltzer J Costabel U Differential cytology of bronchoalveolar lavage fluid in normal children Eur Respir J 1994 7 1865 70 10.1183/09031936.94.07101865 7828697 14. Ordonez CL Shaughnessy TE Matthay MA Fahy JV Increased neutrophil numbers and IL-8 levels in airway secretions in acute severe asthma: Clinical and biologic significance Am J Respir Crit Care Med 2000 161 1185 90 10.1164/ajrccm.161.4.9812061 10764310 15. Gibson PG Simpson JL Saltos N Heterogeneity of airway inflammation in persistent asthma : evidence of neutrophilic inflammation and increased sputum interleukin-8 Chest 2001 119 1329 36 10.1378/chest.119.5.1329 11348936 16. Jartti T Lehtinen P Vuorinen T Ruuskanen O Bronchiolitis: age and previous wheezing episodes are linked to viral etiology and atopic characteristics Pediatr Infect Dis J 2009 28 311 7 10.1097/INF.0b013e31818ee0c1 19258922 17. Bisgaard H Hermansen MN Buchvald F Loland L Halkjaer LB Bonnelykke K Childhood asthma after bacterial colonization of the airway in neonates N Engl J Med 2007 357 1487 95 10.1056/NEJMoa052632 17928596 18. Chen GD Chen AH The wheezing associated with Mycoplasma pneumoniae infection in children Zhonghua Er Ke Za Zhi 2012 50 756 8 23302565 19. Chung HL Kim SG Shin IH The relationship between serum endothelin (ET)-1 and wheezing status in the children with Mycoplasma pneumoniae pneumonia Pediatr Allergy Immunol 2006 17 285 90 10.1111/j.1399-3038.2006.00393.x 16771782 20. Stolz D Stulz A Muller B Gratwohl A Tamm M BAL neutrophils, serum procalcitonin, and C-reactive protein to predict bacterial infection in the immunocompromised host Chest 2007 132 504 14 10.1378/chest.07-0175 17573524 21. Schellhase DE Fawcett DD Schutze GE Lensing SY Tryka AF Clinical utility of flexible bronchoscopy and bronchoalveolar lavage in young children with recurrent wheezing J Pediatr 1998 132 312 8 10.1016/S0022-3476(98)70451-2 9506647 22. Everard ML Swarbrick A Wrightham M McIntyre J Dunkley C James PD Analysis of cells obtained by bronchial lavage of infants with respiratory syncytial virus infection Arch Dis Child 1994 71 428 32 10.1136/adc.71.5.428 7826113 23. Becker S Koren HS Henke DC Interleukin-8 expression in normal nasal epithelium and its modulation by infection with respiratory syncytial virus and cytokines tumor necrosis factor, interleukin-1, and interleukin-6 Am J Respir Cell Mol Biol 1993 8 20 7 10.1165/ajrcmb/8.1.20 8417753 24. Becker S Quay J Soukup J Cytokine (tumor necrosis factor, IL-6, and IL-8) production by respiratory syncytial virus-infected human alveolar macrophages J Immunol 1991 147 4307 12 1753101 25. Jarjour NN Gern JE Kelly EA Swenson CA Dick CR Busse WW The effect of an experimental rhinovirus 16 infection on bronchial lavage neutrophils J Allergy Clin Immunol 2000 105 1169 77 10.1067/mai.2000.106376 10856152 26. Huang H Zhong LL Peng L Li J Lin L Levels of interleukin-4 and interferon-gamma in the bronchoalveolar lavage fluid of children with refractory Mycoplasma pneumoniae pneumonia Zhongguo Dang Dai Er Ke Za Zhi 2012 14 581 4 22898277 27. Fayon M Just J Thien HV Chiba T Pascual L Sandouk G Bacterial flora of the lower respiratory tract in children with bronchial asthma Acta Paediatr 1999 88 1216 22 10.1111/j.1651-2227.1999.tb01020.x 10591422 28. Munthe-Kaas MC Gerritsen J Carlsen KH Undlien D Egeland T Skinningsrud B Eosinophil cationic protein (ECP) polymorphisms and association with asthma, s-ECP levels and related phenotypes Allergy 2007 62 429 36 10.1111/j.1398-9995.2007.01327.x 17362255 29. Bjornsson E Janson C Hakansson L Enander I Venge P Boman G Serum eosinophil cationic protein in relation to bronchial asthma in a young Swedish population Allergy 1994 49 730 6 10.1111/j.1398-9995.1994.tb02095.x 7695062

PMC005xxxxxx/PMC5002097.txt

==== Front TrialsTrialsTrials1745-6215BioMed Central London 152710.1186/s13063-016-1527-6Study ProtocolWalking adaptability therapy after stroke: study protocol for a randomized controlled trial http://orcid.org/0000-0001-8451-2326Timmermans Celine c.timmermans@vu.nlc.timmermans@reade.nl 12Roerdink Melvyn m.roerdink@vu.nl 1van Ooijen Marielle W. m.van.ooijen-kerste@vu.nlm.kerste@reade.nl 12Meskers Carel G. c.meskers@vumc.nl 3Janssen Thomas W. t.w.j.janssen@vu.nlt.janssen@reade.nl 12Beek Peter J. p.j.beek@vu.nl 11 MOVE Research Institute Amsterdam, Department of Human Movement Sciences, Vrije Universiteit Amsterdam, Van der Boechorststraat 9, Amsterdam, 1081 BT The Netherlands 2 Amsterdam Rehabilitation Research Center, Reade, Overtoom 283, Amsterdam, 1054 HW The Netherlands 3 VU Medical Centre, Department of Rehabilitation Medicine, De Boelelaan 1118, Amsterdam, 1081 HZ The Netherlands 26 8 2016 26 8 2016 2016 17 1 42515 3 2016 25 7 2016 © Timmermans et al. 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Background Walking in everyday life requires the ability to adapt walking to the environment. This adaptability is often impaired after stroke, and this might contribute to the increased fall risk after stroke. To improve safe community ambulation, walking adaptability training might be beneficial after stroke. This study is designed to compare the effects of two interventions for improving walking speed and walking adaptability: treadmill-based C-Mill therapy (therapy with augmented reality) and the overground FALLS program (a conventional therapy program). We hypothesize that C-Mill therapy will result in better outcomes than the FALLS program, owing to its expected greater amount of walking practice. Methods This is a single-center parallel group randomized controlled trial with pre-intervention, post-intervention, retention, and follow-up tests. Forty persons after stroke (≥3 months) with deficits in walking or balance will be included. Participants will be randomly allocated to either C-Mill therapy or the overground FALLS program for 5 weeks. Both interventions will incorporate practice of walking adaptability and will be matched in terms of frequency, duration, and therapist attention. Walking speed, as determined by the 10 Meter Walking Test, will be the primary outcome measure. Secondary outcome measures will pertain to walking adaptability (10 Meter Walking Test with context or cognitive dual-task and Interactive Walkway assessments). Furthermore, commonly used clinical measures to determine walking ability (Timed Up-and-Go test), walking independence (Functional Ambulation Category), balance (Berg Balance Scale), and balance confidence (Activities-specific Balance Confidence scale) will be used, as well as a complementary set of walking-related assessments. The amount of walking practice (the number of steps taken per session) will be registered using the treadmill’s inbuilt step counter (C-Mill therapy) and video recordings (FALLS program). This process measure will be compared between the two interventions. Discussion This study will assess the effects of treadmill-based C-Mill therapy compared with the overground FALLS program and thereby the relative importance of the amount of walking practice as a key aspect of effective intervention programs directed at improving walking speed and walking adaptability after stroke. Trial registration Netherlands Trial Register NTR4030. Registered on 11 June 2013, amendment filed on 17 June 2016. Electronic supplementary material The online version of this article (doi:10.1186/s13063-016-1527-6) contains supplementary material, which is available to authorized users. Keywords ExerciseRehabilitationStrokeTherapyWalking adaptabilityWalking speedissue-copyright-statement© The Author(s) 2016 ==== Body Background The ability to adapt walking to environmental circumstances, such as the ability to avoid obstacles and to secure safe foot placement in a cluttered environment, is a prerequisite for safe walking in everyday life circumstances. This gait adaptability or walking adaptability [1, 2] is often reduced after stroke [3, 4], which might contribute to the high fall risk in this population [5]. There is thus a clear need to improve this aspect of walking ability in people with stroke. One of the most promising exercise therapies that include practice of walking adaptability is task-specific gait training [5]. Task-specific gait training refers to the practice of associating functional tasks with walking. The benefits of task-specific training in stroke rehabilitation have been demonstrated in several studies [6–8]. Besides task-specific training, context-specific training is a well-accepted rehabilitation principle after stroke, suggesting that training should target the goals relevant for the needs of people with stroke attuned to their environmental circumstances [6, 8]. Hence, including walking adaptability exercises in training interventions aimed at improving safe community ambulation seems appropriate and potentially beneficial for people with stroke. The FALLS program [9] is one such task-specific and context-specific type of overground training intervention, which integrates the practice of complex situations of community walking, such as walking over an obstacle course (Fig. 1a). The FALLS program is based on the Nijmegen Falls Prevention Program, which was designed for community-dwelling older adults with a history of falling, and was shown to reduce the number of falls in this population [10, 11]. Although the effectiveness of the FALLS program needs to be determined in people with stroke, it has been shown to be feasible for this population [9].Fig. 1 Snapshots of the two interventions aimed at improving walking speed and walking adaptability: (a) obstacle course of the overground FALLS program; (b) targeted-stepping exercise of treadmill-based C-Mill therapy C-Mill therapy is another promising example of task-specific and context-specific training with an emphasis on walking adaptability exercises. The C-Mill (Fig. 1b) is an instrumented treadmill augmented with task-relevant visual context (e.g., obstacles, stepping targets) projected on the treadmill’s surface [12]. This context can be administered in a gait-dependent manner, owing to online monitoring of timing and location of foot placements [13]. The projected obstacles and stepping targets make C-Mill therapy well suited for task-specific and context-specific training because step adjustments are required to adapt to the projected context similar to the step adjustments required to adapt to environmental circumstances during community ambulation. A recent proof-of-concept study showed that C-Mill therapy in the chronic stage after stroke is not only well received by this population, but also beneficial [14]. C-Mill therapy resulted in training-related increments in walking speed and improvements in various other walking-related clinical scores. In addition, the ability to make step adjustments improved (i.e., higher obstacle-avoidance success rates) after 5–6 weeks of C-Mill therapy, and these adjustments required less attention (i.e., reduced dual-task interference), suggesting that the step adjustments evolved in a more automatized manner after a period of C-Mill therapy [15]. Besides task-specific and context-specific training, other key ingredients for effective rehabilitation include variability in practice, feedback of performance, and amount of movement practice [6–8, 16, 17]. Both interventions comprise variability in practice, given their wide variety of tasks and exercises. Moreover, both interventions allow for performance feedback, either by group discussions and direct feedback provided by therapists (FALLS program) or by direct feedback of walking adaptability exercise performance, e.g., visual feedback with regard to obstacle hits (C-Mill therapy). However, treadmill-based C-Mill therapy probably allows for a greater amount of walking practice (defined as the number of steps taken per session), because it incorporates treadmill walking, which has been suggested to elicit more steps per session than overground training [18–22]. In this study, we will empirically test this suggestion, using the amount of walking practice as a process measure. The study’s aim is to compare the effects of two promising interventions for improving walking speed, walking adaptability, and commonly used clinical measures of walking and balance in persons after stroke: treadmill-based C-Mill therapy [14, 15] and the overground FALLS program [9]. We expect that C-Mill therapy will result in better outcomes than the FALLS program because of the expected greater amount of walking practice per session of equal duration. Methods Participants In total, 40 persons who had a stroke will be recruited from the inpatient and outpatient population of rehabilitation center Reade (Amsterdam, The Netherlands) to participate in this study. Inclusion criteria are first-ever stroke ≥3 months ago, walking or balance deficits confirmed by a physician, clinical diagnosis of hemiparesis, age ≥18 years, general walking ability as indicated by a Functional Ambulation Category score ≥3 [23], and the ability to understand and execute simple instructions. Exclusion criteria are orthopedic and other neurological disorders that affect walking (e.g., Parkinson’s disease), other treatments that could influence the effects of the interventions (e.g., recent Botulin toxin treatment of the lower extremity), contra-indication to physical activity (e.g., heart failure, severe osteoporosis), moderate or severe cognitive impairments as indicated by a Mini-Mental State Examination [24] score below 21, or severe uncorrected visual deficits. Persons with stroke who are eligible for participation will be informed about the study by their rehabilitation specialist, both orally and in writing. All participants will provide a written informed consent. Study design The proposed study is a single-center, parallel group randomized controlled trial with pre-intervention, post-intervention, retention, and follow-up tests to determine the relative efficacy of the interventions: treadmill-based C-Mill therapy and the overground FALLS program. After giving informed consent, participants will be randomly assigned to one of the two interventions using an automated, custom-made minimization algorithm written in MATLAB. The minimization procedure is based on time after stroke, age and Functional Ambulation Category score to balance groups for these stratification factors. The research assistant will enter the data for randomization in the algorithm and the participant will subsequently be informed about the resulting group allocation before the pre-intervention tests. Subsequently, the assessor will schedule the participants for the assigned 5 week intervention program. Pre-intervention tests (T0) to characterize groups and obtain baseline values of primary and secondary outcome measures will be performed one week prior to the intervention program. Within one week after completing the intervention, post-intervention tests (T1) will be performed. The same tests will be conducted 5 weeks (retention tests, T2) and 12 months (follow-up tests, T3) after completing the intervention. All assessments will be performed at the rehabilitation center. Because of the nature of the intervention studied, therapists and participants cannot be blinded to group allocation. The assessor will also not be blinded to group allocation, because of pragmatic constraints related to the planning of assessments and therapy sessions. Figure 2 shows a flow chart of the procedures that participants will undergo at T0, T1, T2 and T3.Fig. 2 Flow chart of the procedures that participants will undergo Interventions: treadmill-based C-Mill therapy and the overground FALLS program C-Mill therapy is a structured treadmill training program with a specific emphasis on practicing walking adaptability (as detailed in Table 1 and Additional file 1), using gait-dependent augmented-reality content projected on the instrumented treadmill surface to elicit step adjustments [1, 13–15, 25–28]. Figure 3 shows various exercises of C-Mill therapy, including exercises to practice avoidance of projected visual obstacles (Fig. 3a), exercises to practice accurate foot placement on a step-to-step basis by walking to a regular or irregular sequence of visual stepping targets (Fig. 3b), exercises to practice acceleration and deceleration by maintaining position within a projected walking area that moves along the treadmill (Fig. 3c), and a functional and interactive walking adaptability game (Fig. 3d). C-Mill therapy is a patient-tailored type of training in that the therapist can adjust the difficulty of the different exercises by manipulating content parameters as the obstacle size and available response time for obstacle negotiation, the variation in the sequence of stepping targets, and the degree of acceleration and deceleration of the moving walking area. As progressive training has previously been shown to have beneficial effects [29–31], therapists are instructed to increase the difficulty of C-Mill exercises as tolerated by the participant, by either changing content parameters or increasing the belt speed, as described in the pre-defined training protocol (Table 1). To assist therapists in progressively scaling the C-Mill therapy sessions, the participant’s perceived fear and difficulty levels during the sessions will be assessed at a scale from 0 (no fear or not difficult) to 10 (much fear or very difficult), as well as their rating of perceived exertion using the Borg scale (range 6–20, [32]). Furthermore, the pre-defined protocol will guide therapists to vary C-Mill exercises, both in terms of content and the order in which the exercises will be performed, inspired by recent insights into motor learning showing superior transfer and retention effects with variability in practice [16]. C-Mill therapy will be performed in groups of two persons with stroke supervised by one therapist. Therapy sessions will last 1.5 hours each, divided in exercise blocks of 3–8 min, during which the participants alternately train and rest (Table 1).Table 1 Pre-defined protocol for treadmill-based C-Mill therapy Setting Groups of two participants for 90 min; participants will alternately train and rest. Frequency Twice weekly treadmill training program with specific emphasis on walking adaptability. Therapy In the first week, a combination of obstacle avoidance (avoiding visual obstacles projected on the treadmill), practice of accurate foot placement on a step-to-step basis (walking to a regular or irregular sequence of visual stepping targets), and a functional and interactive walking adaptability game (game with the theme ‘beach’ or ‘forest’) will be performed. In weeks 2–5, the combination of obstacle avoidance, accurate foot placement on a step-to-step basis and the functional and interactive walking adaptability game will be complemented by walking speed adaptations (acceleration and deceleration evoked by a moving walking area). Participants will start in week 1 at a comfortable walking speed; this speed will be gradually increased during the 5 week period. The weekly increase of the walking speed will be 10 %, provided that the therapy remains safe and is tolerated by the participant. Besides the walking speed, the difficulty of C-Mill exercises will be gradually increased, as tolerated by the participant. Therapist C-Mill therapy will be provided by a single therapist, an expert in C-Mill therapy. The therapists involved in the C-Mill therapy were all trained with regard to operating the C-Mill and to the specific guidelines of the intervention before the study started. Most therapists were already experienced C-Mill users before the study started. The therapists regularly meet the research assistant to ensure adherence to the protocol (Additional file 1). Fig. 3 Exercises of treadmill-based C-Mill therapy: (a) obstacle avoidance; (b) visually guided stepping to a sequence of stepping targets; (c) acceleration and deceleration evoked by a moving walking area; (d) functional and interactive walking adaptability game (adopted from Van Ooijen et al. [20]) The FALLS program [9] is an overground therapy program aimed at reducing the number of falls in people with stroke by practicing walking adaptability, among other aspects (as detailed in Table 2 and Additional file 2). Figure 4 shows various exercises of this pre-defined FALLS program, including exercises to practice obstacle avoidance (Fig. 4a), exercises to practice foot placement while walking over uneven terrain (Fig. 4b), tandem walking (Fig. 4c), and slalom walking (Fig. 4d). These exercises must also be performed while cognitive and motor dual-tasks are imposed, as well as under visual constraints. In addition, the program incorporates exercises to simulate walking in a crowded environment and to practice falling techniques (one session per week). The FALLS program was originally performed in groups of six persons with stroke, with two or three therapists per group in therapy sessions lasting 2 hours each [9]. Following design considerations for this study (as detailed in the next section), the FALLS program will be performed in groups of four to six persons with two or three therapists per group, with sessions lasting 1.5 hours, including rest.Table 2 Pre-defined protocol for the overground FALLS program Setting Groups of 4–6 participants for 90 min, participants will alternately train and rest. Frequency Twice weekly overground training program, which incorporates walking adaptability exercises. Therapy The first therapy session of the week will be devoted to an obstacle course that simulates potential challenging situations of daily life. The obstacle course facilitates practicing balance, gait, and coordination, and mimics activities of daily life with high fall risk, such as walking over obstacles, uneven terrain, slalom walking and tandem walking. The obstacle course will also be negotiated while imposing cognitive and motor dual-tasks, as well as under visual constraints. The second therapy session of the week will include walking exercises and practice of fall techniques. The walking exercises simulate walking in a crowded environment. Adjustments in walking speed and direction are required during these exercises and collisions with other people challenge balance. The practice of fall techniques is based on martial arts techniques and will include falling forwards, backwards, and laterally. The level of difficulty will be gradually enhanced by increasing fall height (from sitting on a safety mat to stance height). Therapist The therapy sessions will be provided by two or three therapists, depending on the size of the group. At least one of therapists is trained in the background, methods, and techniques of the FALLS program. All therapists involved in the FALLS program are trained and experienced with regard to the program protocol and instructed to follow the specific guidelines of the intervention for the purpose of this trial. The therapists regularly meet the research assistant to ensure adherence to the protocol. (Additional file 2). Fig. 4 Exercises of the obstacle course of the overground FALLS program: (a) obstacle avoidance; (b) walking over uneven terrain; (c) tandem walking; (d) slalom walking Both interventions are matched for therapy duration (90 min), frequency (twice weekly) and therapist attention (mean participant-to-therapist ratio, 2:1). The amount of walking practice per session (defined as the number of steps performed during therapy sessions) will be compared between the two interventions and treated as the process measure. Therefore, the number of steps taken during C-Mill therapy sessions will be registered using the treadmill’s inbuilt step counter, while an observer will count the number of steps taken during the FALLS program offline in a random selection of FALLS program sessions using video recordings of the sessions in question. Finally, after completing the last session of the intervention, participants will be asked to fill in a purpose-designed questionnaire to register perceived discomfort during and after therapy sessions, as well as their experience with the therapy, to compare the feasibility of the interventions from a participant’s perspective. Outcome measures After group allocation, pre-intervention tests will be performed to assess the baseline values of primary and secondary outcome measures and to collect participant characteristics (sex, age, height, body mass, medication use, co-morbidities, side and location of the lesion, current living situation, daily functioning and the use of assistive devices). The primary outcome measure in this study will be walking speed. Walking speed will be assessed using the 10 Meter Walking Test [33], which has been shown to be a reliable and robust means for measuring walking speed [34]. The secondary outcome measures are inspired by the targeted-stepping and obstacle-avoidance results of Hollands et al. [35] and Van Ooijen et al. [15], underscoring the importance of task-specificity and context-specificity in walking adaptability assessments. Van Ooijen et al. [15] showed enhanced obstacle-avoidance success rates at lower attentional costs after a period of C-Mill walking adaptability therapy, while Hollands et al. [35] showed that measures of targeted stepping were clinically meaningful components in the recovery of functional mobility after stroke. Therefore, the 10 Meter Walking Test will also be performed in combination with context (10 Meter Walking Test with three obstacles, a tandem walking path and three stepping targets) (Fig. 5), a cognitive dual-task (10 Meter Walking Test while counting backwards in steps of three [36]) and both context and dual-task (10 Meter Walking Test with three obstacles, a tandem walking path, and three stepping targets, and while counting backwards in steps of three). Walking adaptability will also be assessed using the Interactive Walkway (Technology4Science, Vrije Universiteit Amsterdam, The Netherlands), a walkway instrumented with multiple Microsoft Kinect for Windows sensors and a projector to present visual context, such as obstacles and stepping targets in a gait-dependent manner (Fig. 6). The walking adaptability evaluation with the Interactive Walkway includes targeted-stepping assessments, obstacle-avoidance assessments, and obstacle-avoidance assessments while counting backwards in steps of three. The obstacle-avoidance assessment of the Interactive Walkway differs from the 10 Meter Walking Test with context in that the Interactive Walkway obstacles can be suddenly presented in a gait-dependent manner, that is, the obstacle suddenly appears at the location where the participant would place his or her foot without adjusting gait. Hence, a step adjustment is always required to avoid the obstacle successfully. Moreover, this step adjustment needs to be performed under high time-pressure demands, which is especially difficult for persons after stroke [37]. The difference between the stepping targets within the 10 Meter Walking Test with context and the Interactive Walkway targeted-stepping assessment is that the Interactive Walkway targets are presented in regular and irregular sequences of visual stepping targets based on participants’ self-selected step length. In this way, it is possible to evaluate foot placement errors on a step-to-step basis for each participant. The 10 Meter Walking Test scores and Interactive Walkway assessment scores will be given in seconds required to complete each test, as well as in the number of errors made during the obstacle crossings, targeted stepping, and tandem walking. The cognitive dual-task, a serial-3 subtraction task, will be analyzed by counting the number of subtractions, as well as the number of mistakes made (dual-task performance [DTP]). Subtraction-task performance while walking will be normalized to subtraction-task performance while sitting (i.e., single-task control condition). These walking adaptability evaluation tools are expected to be sensitive and specific for finding improvements after walking adaptability interventions.Fig. 5 Walking adaptability assessment using the 10 Meter Walking Test with context: (a) obstacle avoidance; (b) targeted stepping and; (c) tandem walking Fig. 6 Walking adaptability assessments using the Interactive Walkway: (a) avoidance of suddenly appearing obstacles and (b) walking to a sequence of stepping targets, both presented on the walking surface in a gait-dependent manner Secondary outcome measures are drawn from a comprehensive set of common clinical measures to determine walking ability, balance, and other walking-related constructs, including Timed Up-and-Go test [38] and Functional Ambulation Category [23]. The obstacle-avoidance subtask of the modified Emory Functional Ambulation Profile will be performed [39], a conventional clinical test closely related to the construct of walking adaptability. The modified Emory Functional Ambulation Profile is reliable and valid for use in people with stroke [40]. Balance will be assessed using the Berg Balance Scale, which provides a psychometrically sound measure of balance impairment for use in post-stroke assessment [40, 41]. Executive function will be assessed using the valid and reliable Trail Making Test [42]. Balance confidence will be assessed with the Activities-specific Balance Confidence scale, a questionnaire measuring balance confidence in performing specific activities, which has good test-retest reliability and validity [43, 44]. Self-reported limitations in walking will be assessed using the Walking Questionnaire [45], which targets experienced limitations in indoor and outdoor walking relative to pre-stroke walking limitations. Finally, the Nottingham Extended Activities of Daily Living scale [46–48] will be used to assess activities of daily living. Table 3 provides an overview of the tests that will be performed at T0, T1, T2 and T3.Table 3 Overview of all tests performed at T0, T1, T2 and T3 Primary outcome measure  10 Meter Walking Test (m/s) Secondary outcome measures  10 Meter Walking Test with context (m/s, number of errors)  10 Meter Walking Test with a cognitive dual-task (m/s, DTP)  10 Meter Walking Test with context and a cognitive dual-task (m/s, number of errors, DTP)  Interactive Walkway targeted-stepping assessment (m/s, number of errors)  Interactive Walkway obstacle-avoidance assessment (m/s, number of errors)  Interactive Walkway obstacle-avoidance assessment with a cognitive dual-task (m/s, number of errors, DTP)  Timed Up-and-Go test (m/s)  Functional Ambulation Category (3–5)  Obstacle-avoidance subtask of the modified Emory Functional Ambulation Profile (m/s)  Berg Balance Scale (0–56)  Activities-specific Balance Confidence scale (0–100 %)  Trail Making Test (s)  Walking Questionnaire  Nottingham Extended Activities of Daily Living scale (0–66) Finally, the number of steps taken per therapy session will be recorded, since we expect that the amount of walking practice per session (defined as the number of steps performed during therapy sessions) will be higher for treadmill-based C-Mill therapy than for the overground FALLS program. This expectation will be tested by comparing this process measure between the two intervention groups. Sample size The primary outcome measure in this study will be walking speed. Previous clinical trials in people with stroke by Yang et al. [49] and Jaffe et al. [50] showed greater improvements in walking speed after treadmill training in a complex and challenging virtual reality environment than after, respectively, conventional treadmill training and overground obstacle-avoidance training [49, 50]. Unfortunately, effect sizes and required sample sizes for a controlled clinical trial with multiple comparisons cannot be estimated from the results of these studies, but both reported significant between-group differences in walking speed with small sample sizes of 9 to 10 participants in each intervention group. The study of Yang et al. [49] allows for a sample size calculation for post-hoc analyses for significant group effects on walking speed with independent t tests. Based on those results, we aim for a relative, clinically relevant, improvement in walking speed of 0.50 km/h (∆) with a common standard deviation (SD) of 0.47 km/h, which results in a sample size of 14 participants in each group to achieve 80 % power with a two-tailed α of 0.05, i.e., following N=2SD2Zα+Zβ2Δ2 [51]. Considering a drop out of 10–25 %, we chose to increase our sample to 20 participants in each intervention group to be on the safe side for establishing the relative efficacy of the two interventions in terms of improvements in walking speed. Data analysis Descriptive group statistics will be used to characterize the two intervention groups in terms of sex, age, height, body mass, Mini-Mental State Examination, Functional Ambulation Category, medication use, co-morbidities, side and location of the lesion, current living situation, daily functioning and the use of assistive devices, as well as perceived discomforts during and after therapy sessions and participant’s experience with the therapy. An independent t test will be used to compare the mean number of steps taken per session between the two interventions. Primary and secondary longitudinal outcome measures that are normally distributed will be analyzed using repeated-measures ANOVA with the between-subject factor group (two levels: C-Mill therapy and the FALLS program) and the within-subject factor time (four levels: pre-intervention [T0], post-intervention [T1], retention [T2], and follow-up [T3] tests). Post-hoc analysis using independent t tests between groups per time level will be performed in case of significant interaction effects. For ordinal or non-normal distributed variables, we will use Mann–Whitney U tests and Friedman tests to evaluate possible main effects of group and time, respectively. To analyze possible interactions between groups and times, we will apply Kruskal–Wallis tests to change scores (i.e., relative to the previous time level) at T1, T2, and T3. When significant, Mann–Whitney U post-hoc tests will be performed to identify between-group differences in change scores per time level. Significant effects are assumed for P < 0.05. Data will be analyzed as randomized. Missing data will be imputed using the data from the last available measurement. Discussion This randomized controlled trial will evaluate the relative effects of treadmill-based C-Mill therapy and the overground FALLS program on walking speed and walking adaptability in people with stroke. Although both C-Mill therapy and the FALLS program incorporate practice of walking adaptability and thereby aim at improving community ambulation, and first results are encouraging in this regard [9, 14, 15], it is hypothesized that C-Mill therapy will result in better outcomes than the FALLS program, as a result of the expected greater amount of walking practice owing to treadmill training [18, 20–22]. The results of the study of Moore et al. [19] indeed showed significant gains in daily stepping and walking efficacy after treadmill training, compared with conventional physical therapy, which appears to be related to the number of steps taken per session. In this study, we will explicitly test the anticipated greater amount of walking practice with treadmill training by comparing the registered number of steps taken per session between the two intervention groups. The expected superior outcome of C-Mill therapy relative to the FALLS program may be further mediated by the possibility of tailoring the training to the patient’s needs and progress. During C-Mill therapy, the therapist can adjust the difficulty of the different exercises by manipulating content parameters, such as the variation in the sequence of stepping targets, the obstacle size and available response time for obstacle negotiation, and the degree of acceleration and deceleration of the moving walking area. As progressive training has superior effects [29–31], this patient-tailored challenge of C-Mill therapy might be beneficial, compared with the FALLS program. Conversely, the use of real obstacles and context and the practice of falling techniques might favor outcomes of the FALLS program compared with C-Mill therapy for its superior context-specificity. A methodological strength of this study is that both interventions will be matched for therapy duration, frequency, and therapist attention. This means that if there is a superior effect on walking adaptability and walking speed of one of the interventions, this will be realized by the same investment in time and resources. Furthermore, both interventions implicitly utilize and train the direct visuolocomotor control of walking in an enriched environmental context [52, 53], allowing for a direct and natural visuolocomotor control in which the point of gaze is typically coupled to future foot placement locations. The two interventions in this study are similar with regard to visuolocomotor control of step adjustments relative to environmental context (e.g., real obstacles in the FALLS program, real visual obstacles in C-Mill therapy). The proposed trial of Hollands et al. [54] also testifies to the growing interest in the use of visual cues for task-specific gait training, thereby also implicitly training visuolocomotor control [54]. Hollands et al. intend to compare usual care without visual cues to overground visual cue training and treadmill visual cue training (using the C-Mill) in persons with stroke to examine the feasibility of task-specific locomotor practice incorporating visual cues. Therefore, our study, in combination with the study of Hollands et al. [54], might underpin the importance of visuolocomotor control in gait rehabilitation, as well as the potential surplus value of a treadmill in that regard. A limitation of this study is that it involves only one center. This might influence the generalizability of the research results to other rehabilitation centers. Another limitation of this study is the non-blinding of the assessors. To reduce potential influence of this limitation on the outcomes, instructions will be standardized and tasks will be computerized when possible. In summary, this study will shed light on the effects of treadmill-based C-Mill therapy compared with the overground FALLS program and thereby on the relative importance of the amount of walking practice as an important ingredient of effective interventions of walking speed and walking adaptability after stroke. Hence, the results of this study will be important in optimizing effective intervention programs directed at improving walking speed and walking adaptability after stroke. Trial status Recruitment commenced in 2013 and is ongoing. Results of this study are expected in 2017. Abbreviation ANOVA, analysis of variance; DTP, dual-task performance Additional files Additional file 1: C-Mill therapy treatment booklet for therapists. (PDF 166 kb) Additional file 2: FALLS program treatment booklet for therapists. (PDF 182 kb) Acknowledgements Daphne Geerse is acknowledged for her contribution to the design of the protocol. Authors’ contributions MWvO, MR, TWJ, and PJB drafted the protocol design. CT is the executive investigator and drafted this paper. MR, MWvO, CM, TWJ, and PJB critically revised the manuscript. All authors read and approved the manuscript and consider themselves accountable for all aspects of the work. Competing interests MR and PJB are inventors of rehabilitation treadmills that include visual context for foot placement [12]. Vrije Universiteit Amsterdam granted this invention exclusively to ForceLink (Culemborg, The Netherlands), an industrial partner of Vrije Universiteit Amsterdam. ForceLink is the manufacturer of the C-Mill treadmill and assignee of a patent for rehabilitation treadmills with visual context for foot placement, with MR and PJB listed as inventors. Vrije Universiteit Amsterdam received part of the patent revenues, to spend freely on their research endeavors. Vrije Universiteit Amsterdam used these revenues to finance a research project on the effectiveness of C-Mill therapy. This study is part of that research project. MR and PJB did not receive any reimbursements, fees, funding, or salary from ForceLink, nor did they benefit personally from patent revenues. This study is not funded by a major funding body. Consent for publication Written informed consent was obtained from the patients for publication of this manuscript and accompanying images. A copy of the written consent is available for review by the editor-in-chief of this journal. Ethics approval and consent to participate This study was approved by the Medical Ethical Reviewing Committee of VU Medical Centre, Amsterdam, The Netherlands (protocol number 2013/53) and the Central Committee on Research Involving Human Subjects (protocol number NL 42461.029.13). ==== Refs References 1. Houdijk H van Ooijen MW Kraal JJ Wiggerts HO Polomski W Janssen TWJ Roerdink M Assessing gait adaptability in people with a unilateral amputation on an instrumented treadmill with a projected visual context Phys Ther. 2012 92 1452 60 10.2522/ptj.20110362 22836005 2. Balasubramanian CK Clark DJ Fox EJ Walking adaptability after stroke and its assessment in clinical settings Stroke Res Treat 2014 25254140 3. Said CM Goldie PA Patla AE Sparrow WA Martin KE Obstacle crossing in subjects with stroke Arch Phys Med Rehabil. 1999 80 1054 9 10.1016/S0003-9993(99)90060-6 10489008 4. Smulders K van Swigchem R de Swart BJM Geurts ACH Weerdesteyn V Community-dwelling people with chronic stroke need disproportionate attention while walking and negotiating obstacles Gait Posture. 2012 36 127 32 10.1016/j.gaitpost.2012.02.002 22418584 5. Weerdesteyn V de Niet M van Duijnhoven HJR Geurts ACH Falls in individuals with stroke J Rehabil Res Dev. 2008 45 1195 213 10.1682/JRRD.2007.09.0145 19235120 6. Langhorne P Bernhardt J Kwakkel G Stroke rehabilitation Lancet. 2011 377 1693 702 10.1016/S0140-6736(11)60325-5 21571152 7. Stretton CM Mudge S Kayes NM McPherson KM Interventions to improve real-world walking after stroke: a systematic review and meta-analysis Clin Rehabil 2016 27056251 8. Hornby TG Straube DS Kinnaird CR Holleran CL Echauz AJ Rodriguez KS Wagner EJ Narducci EA Importance of specificity, amount, and intensity of locomotor training to improve ambulatory function in patients poststroke Top Stroke Rehabil. 2011 18 293 307 10.1310/tsr1804-293 21914594 9. Van Duijnhoven HJR de Kam D Hellebrand W Smulders E Geurts ACH Weerdesteyn V Development and process evaluation of a 5-week exercise program to prevent falls in people after stroke (the FALLS program) Stroke Res Treat 2012 22195292 10. Weerdesteyn V Rijken H Geurts ACH Smits-Engelsman BCM Mulder T Duysens J A five-week exercise program can reduce falls and improve obstacle avoidance in the elderly Gerontology. 2006 52 131 41 10.1159/000091822 16645293 11. Weerdesteyn V Smulders E Rijken H Duysens J Preserved effectiveness of a falls prevention exercise program after implementation in daily clinical practice J Am Geriatr Soc. 2009 57 2162 4 10.1111/j.1532-5415.2009.02531.x 20121968 12. Roerdink M, Beek PJ. Device for displaying target indications for foot movements to persons with a walking disorder. US patent 2009246746-A1 (1 October 2009), European patent 2106779-A1 (7 October 2009), Japanese patent 2009240775-A (22 October 2009), Dutch patent 1035236-C2 (1 October 2009). 13. Roerdink M Coolen BH Clairbois BHE Lamoth CJC Beek PJ Online gait event detection using a large force platform embedded in a treadmill J Biomech. 2008 41 2628 32 10.1016/j.jbiomech.2008.06.023 18657816 14. Heeren A van Ooijen MW Geurts ACH Day BL Janssen TWJ Beek PJ Roerdink M Weerdesteyn V Step by step: a pilot study on the effectiveness of gait adaptability training in the chronic phase after stroke using a treadmill with visual context J Rehabil Med. 2013 45 616 22 10.2340/16501977-1180 23811818 15. Van Ooijen MW Heeren A Smulders K Geurts ACH Janssen TWJ Beek PJ Weerdesteyn V Roerdink M Improved gait adjustments after gait adaptability training are associated with reduced attentional demands in persons with stroke Exp Brain Res. 2015 233 1007 18 10.1007/s00221-014-4175-7 25537466 16. Beek PJ Roerdink M Selzer M Clarke S Cohen L Kwakkel G Miller R Evolving insights into motor learning and their implications for neurorehabilitation Textbook of neural repair and rehabilitation 2014 Cambridge Cambridge University Press 95 104 17. Lang CE Macdonald JR Reisman DS Boyd L Jacobson KT Schindler-Ivens SM Hornby TG Ross SA Scheets PL Observation of amounts of movement practice provided during stroke rehabilitation Arch Phys Med Rehabil. 2009 90 1692 8 10.1016/j.apmr.2009.04.005 19801058 18. Hesse S Konrad M Uhlenbrock D Treadmill walking with partial body weight support versus floor walking in hemiparetic subjects Arch Phys Med Rehabil. 1999 80 421 7 10.1016/S0003-9993(99)90279-4 10206604 19. Moore JL Roth EJ Killian C Hornby TG Locomotor training improves daily stepping activity and gait efficiency in individuals poststroke who have reached a ‘plateau’ in recovery Stroke. 2010 41 129 35 10.1161/STROKEAHA.109.563247 19910547 20. Van Ooijen MW Roerdink M Trekop M Visschedijk J Janssen TWJ Beek PJ Functional gait rehabilitation in elderly people following a fall-related hip fracture using a treadmill with visual context: design of a randomized controlled trial BMC Geriatr. 2013 13 34 10.1186/1471-2318-13-34 23590327 21. Holleran CL Straube DD Kinnaird CR Leddy AL Hornby TG Feasibility and potential efficacy of high-intensity stepping training in variable contexts in subacute and chronic stroke Neurorehabil Neural Repair. 2014 28 643 51 10.1177/1545968314521001 24515925 22. Hornby TG Holleran CL Hennessy PW Leddy AL Connolly M Camardo J Woodward J Mahtani G Lovell L Roth EJ Variable Intensive Early Walking Poststroke (VIEWS): a randomized controlled trial Neurorehabil Neural Repair 2016 30 5 440 50 10.1177/1545968315604396 26338433 23. Holden MK Gill KM Magliozzi MR Nathan J Piehl-Baker L Clinical gait assessment in the neurologically impaired: reliability and meaningfulness Phys Ther. 1984 64 35 40 6691052 24. Folstein MF Folstein SE McHugh PR ‘Mini-Mental State’: a practical method for grading the cognitive state of patients for the clinician J Psychiatry. 1975 12 189 98 25. Bank PJM Roerdink M Peper CE Comparing the efficacy of metronome beeps and stepping stones to adjust gait: steps to follow! Exp Brain Res. 2011 209 159 69 10.1007/s00221-010-2531-9 21221956 26. Peper CE Oorthuizen JK Roerdink M Attentional demands of cued walking in healthy young and elderly adults Gait Posture. 2012 36 378 82 10.1016/j.gaitpost.2012.03.032 22608701 27. Mazaheri M Roerdink M Bood RJ Duysens J Beek PJ Peper CE Attentional costs of visually guided walking: effects of age, executive function and stepping-task demands Gait Posture. 2014 40 182 6 10.1016/j.gaitpost.2014.03.183 24767613 28. Potocanac Z Hoogkamer W Carpes FP Pijnappels M Verschueren SMP Duysens J Response inhibition during avoidance of virtual obstacles while walking Gait Posture. 2014 39 641 4 10.1016/j.gaitpost.2013.07.125 23968973 29. Pohl M Mehrholz J Ritschel C Rückriem S Speed-dependent treadmill training in ambulatory hemiparetic stroke patients: a randomized controlled trial Stroke. 2002 33 553 8 10.1161/hs0202.102365 11823669 30. Sullivan KJ Knowlton BJ Dobkin BH Step training with body weight support: effect of treadmill speed and practice paradigms on poststroke locomotor recovery Arch Phys Med Rehabil. 2002 83 683 91 10.1053/apmr.2002.32488 11994808 31. Eng JJ Tang PF Gait training strategies to optimize walking ability in people with stroke: a synthesis of the evidence Expert Rev Neurother. 2007 7 1417 36 10.1586/14737175.7.10.1417 17939776 32. Borg G Perceived exertion as an indicator of somatic stress Scand J Rehabil Med. 1970 2 92 8 5523831 33. Graham JE Ostir GV Fisher SR Ottenbacher KJ Assessing walking speed in clinical research: a systematic review J Eval Clin Pract. 2008 14 552 62 10.1111/j.1365-2753.2007.00917.x 18462283 34. Collen FM Wade DT Bradshaw CM Mobility after stroke: reliability of measures of impairment and disability Int Disabil Stud. 1990 12 6 9 10.3109/03790799009166594 2211468 35. Hollands KL Pelton TA van der Veen S Alharbi S Hollands MA A novel and simple test of gait adaptability predicts gold standard measures of functional mobility in stroke survivors Gait Posture. 2015 3 170 5 26455475 36. Yang L He C Pang MY Reliability and validity of dual-task mobility assessments in people with chronic stroke Plos One 2016 11 1 e0147833 10.1371/journal.pone.0147833 26808662 37. Den Otter AR Geurts ACH de Haart M Mulder T Duysens J Step characteristics during obstacle avoidance in hemiplegic stroke Exp Brain Res. 2005 161 180 92 10.1007/s00221-004-2057-0 15517222 38. Podsiadlo D Richardson S The timed ‘Up & Go’: a test of basic functional mobility for frail elderly persons J Am Geriatr Soc. 1991 39 142 8 10.1111/j.1532-5415.1991.tb01616.x 1991946 39. Baer HR Wolf SL Modified Emory Functional Ambulation Profile: an outcome measure for the rehabilitation of poststroke gait dysfunction Stroke. 2001 32 973 9 10.1161/01.STR.32.4.973 11283399 40. Blum L Korner-Bitensky N Usefulness of the Berg Balance Scale in stroke rehabilitation: a systematic review Phys Ther. 2008 88 559 66 10.2522/ptj.20070205 18292215 41. Pollock CL Eng JJ Garland SJ Clinical measurement of walking balance in people post stroke: a systematic review Clin Rehabil. 2011 25 693 708 10.1177/0269215510397394 21613511 42. Reitan RM Wolfson D The Halstead–Reitan neuropsychological test battery: theory and clinical interpretation 1993 2 Tucson, AZ Neuropsychology Press 43. Botner EM Miller WC Eng JJ Measurement properties of the Activities-specific Balance Confidence Scale among individuals with stroke Disabil Rehabil. 2005 27 56 163 10.1080/09638280400008982 44. Ylva N Anette F Psychometric properties of the Activities-Specific Balance Confidence Scale in persons 0–14 days and 3 months post stroke Disabil Rehabil. 2012 34 1186 91 10.3109/09638288.2011.637604 22148983 45. Roorda LD Roebroeck ME van Tilburg T Molenaar IW Lankhorst GJ Bouter LM Measuring activity limitations in walking: development of a hierarchical scale for patients with lower-extremity disorders who live at home Arch Phys Med Rehabil. 2005 86 2277 83 10.1016/j.apmr.2005.06.014 16344023 46. Nouri FM Lincoln NB An extended activities of daily living scale for stroke patients Clin Rehabil. 1987 1 301 5 10.1177/026921558700100409 47. Hsueh I Huang SL Chenoe MH Jushoe SD Hsieh CL Evaluation of stroke patients with the extended activities of daily living scale in Taiwan Disabil Rehabil. 2000 22 495 500 10.1080/096382800413989 10972353 48. Wu CY Chuang LL Lin KC Lee SD Hong WH Responsiveness, minimal detectable change, and minimal clinically important difference of the Nottingham Extended Activities of Daily Living scale in patients with improved performance after stroke rehabilitation Arch Phys Med Rehabil. 2011 92 1281 7 10.1016/j.apmr.2011.03.008 21807147 49. Yang Y Tsai M Chuang T Sung W Wang R Virtual reality-based training improves community ambulation in individuals with stroke: a randomized controlled trial Gait Posture. 2008 28 201 6 10.1016/j.gaitpost.2007.11.007 18358724 50. Jaffe DL Brown DA Pierson-Carey CD Buckley EL Lew HL Stepping over obstacles to improve walking in individuals with poststroke hemiplegia J Rehabil Res Dev. 2004 41 283 92 10.1682/JRRD.2004.03.0283 15543446 51. Vincent WJ Weir JP Statistics in kinesiology 2012 4 Leeds Human Kinetics 52. Patla AE Vickers JN How far ahead do we look when required to step on specific locations in the travel path during locomotion? Exp Brain Res. 2003 148 133 8 10.1007/s00221-002-1246-y 12478404 53. Hollands MA Marple-Horvat DE Coordination of eye and leg movements during visually guided stepping J Mot Behav. 2001 33 205 16 10.1080/00222890109603151 11404215 54. Hollands KL Pelton T Wimperis A Whitham D Jowett S Sackley C Wing A van Vliet P Visual cue training to improve walking and turning after stroke: a study protocol for a multi-centre, single blind randomised pilot trial Trials. 2013 14 276 10.1186/1745-6215-14-276 24004882

PMC005xxxxxx/PMC5002098.txt

==== Front BMC Infect DisBMC Infect. DisBMC Infectious Diseases1471-2334BioMed Central London 179010.1186/s12879-016-1790-xStudy ProtocolEvaluation of high-dose rifampin in patients with new, smear-positive tuberculosis (HIRIF): study protocol for a randomized controlled trial Milstein Meredith Meredith_Milstein@hms.harvard.edu 1Lecca Leonid Llecca_ses@pih.org 23Peloquin Charles Peloquin@cop.ufl.edu 4Mitchison Denis dmitchis@sgul.ac.uk 5Seung Kwonjune Kjseung@pih.org 26Pagano Marcello Pagano@hsph.harvard.edu 7Coleman David bofcoleman@gmail.com 5Osso Elna Elna_Osso@hms.harvard.edu 1Coit Julia Julia_Coit@hms.harvard.edu 1Vargas Vasquez Dante Elmo dantevargasv_4@hotmail.com 8Sanchez Garavito Epifanio epis53@yahoo.es 9Calderon Roger RCalderon_ses@pih.org 23Contreras Carmen CContreras_ses@pih.org 23Davies Geraint GerryDavies@doctors.org.uk 10Mitnick Carole D. Carole_Mitnick@hms.harvard.edu 121 Harvard Medical School, Boston, MA 02118 USA 2 Partners In Health, Boston, MA 02215 USA 3 Socios En Salud, Sucursal-Peru, Lima, Peru 4 University of Florida, Gainesville, FL USA 5 St. Georges University, London, UK 6 Brigham and Women’s Hospital, Boston, MA USA 7 Harvard School of Public Health, Boston, MA USA 8 Hospital Nacional Hipólito Unanue, El Augustino, Lima, Peru 9 Hospital Nacional Sergio E. Bernales, Comas, Lima, Peru 10 University of Liverpool, Liverpool, UK 27 8 2016 27 8 2016 2016 16 1 4534 5 2016 18 8 2016 © The Author(s). 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Background Evidence has existed for decades that higher doses of rifampin may be more effective, but potentially more toxic, than standard doses used in tuberculosis treatment. Whether increased doses of rifampin could safely shorten treatment remains an open question. Methods/Design The HIRIF study is a phase II randomized trial comparing rifampin doses of 20 and 15 mg/kg/day to the standard 10 mg/kg/day for the first 2 months of tuberculosis treatment. All participants receive standard doses of companion drugs and a standard continuation-phase treatment (4 months, 2 drugs). They are followed for 6 months post treatment. Study participants are adults with newly diagnosed, previously untreated, smear positive (≥2+) pulmonary tuberculosis. The primary outcome is rifampin area under the plasma concentration-time curve (AUC0–24) after at least 14 days of study treatment/minimum inhibitory concentration. 180 randomized participants affords 90 % statistical power to detect a difference of at least 14 mcg/mL*hr between the 20 mg/kg group and the 10 mg/kg group, assuming a loss to follow-up of up to 17 %. Discussion Extant evidence suggests the potential for increased doses of rifampin to shorten tuberculosis treatment duration. Early studies that explored this potential using intermittent, higher dosing were derailed by toxicity. Given the continued large, global burden of tuberculosis with nearly 10 million new cases annually, shortened regimens with existing drugs would offer an important advantage to patients and health systems. Trial registration This trial was registered with clinicaltrials.gov (registration number: NCT01408914) on 2 August 2011. Keywords TuberculosisRandomized trialPharmacokineticsRifampi(ci)nTreatment shorteninghttp://dx.doi.org/10.13039/100000060National Institute of Allergy and Infectious DiseasesU19AI091429issue-copyright-statement© The Author(s) 2016 ==== Body Background Nearly 10 million new cases of tuberculosis (TB) and 1.5 million deaths due to TB occur worldwide each year. Of those who receive treatment, 86 % experience successful outcomes [1]. Although TB treatment is currently recommended for six months, it has been argued that shortening treatment by as little as two months would accrue substantial benefits to patients and health systems [2, 3]. Consequently, studies of shortened treatment have been implemented. These generally fall into three categories. First are studies that have assessed whether shortened regimens would be adequate in patients with less severe disease (represented by absence of cavitation on radiography) [4]. The second group has investigated regimens containing a new drug, pretomanid with or without bedaquiline [5] (STAND:NCT02342886;NC-005:NCT02193776). The third, and most common type has investigated introduction or modification of doses of existing drug classes (fluoroquinolones and rifamycins) in treatment shortening (ReMoxTB, RIFAQUIN, OFLOTUB, TBTC Studies 27, 28, 29) [6–11]. Rifampin (RIF), one member of the rifamycin class, is unique among the drugs explored for shortening potential. It is the only drug that already has a TB indication from multiple stringent regulatory authorities, is globally used routinely for TB, and is cheap and produced by multiple quality-assured generic manufacturers. Additionally, there is an extensive body of in vitro, animal, and human evidence suggesting that higher-than-standard daily doses of RIF may safely and successfully shorten the 6-month TB treatment [12–16]. For TB, RIF is dosed at 600 mg daily (10 mg/kg/day). This dose was selected in the absence of studies optimizing daily dose and in the presence of perceived resource constraints [17, 18]. Although in vitro and animal data revealed concentration-dependent killing [19–25], translation of these findings to clinical studies was influenced by the perceived high cost of rifampin at its introduction. To minimize costs, intermittent (1–3 times/week), higher doses of rifampin for TB were explored in humans. Increased plasma concentrations and more rapid culture conversion were observed [26, 27]. Decroix et al. reported that increasing the rifampin dose from 600 to 900 mg resulted in a near doubling of serum concentrations during the two months when concentrations were monitored [26]. Acocella observed tripling of the maximum serum concentration (Cmax) with a doubling of the dose of rifampin, from 10 to 20 mg/kg/day. This is likely due to the significant saturation of rifampin first-pass metabolism as doses were increased [28]. Thus, significant increases in exposure are to be expected from even modest increases in dose. The presence of toxicity among patients receiving intermittent, higher doses of rifampin, however, ended efforts to explore improved activity in the 1970s [29–32]. High-dose (15–20 mg/kg) daily rifampin has now been used for other indications (leprosy [33], resistant Streptococcus pneumoniae [34], staphylococcal infections of orthopedic implants [35], Legionella jordanis [36], and cutaneous leishmaniasis [37]) without evidence of dose-dependent toxicity. This is consistent with the results of a recently published maximum dose-tolerability study, which found no serious adverse events occurring at doses of up to 35 mg/kg/day for 1- 2 weeks [38]. These observations suggest that the serious toxicities previously associated with high-dose RIF—particularly hepatotoxicity and flu-like syndrome—may be idiosyncratic rather than dose related, or may be linked to intermittency [12, 14–16]. More recent work has corroborated the findings that rifampin exposure is known to be dose-related and at least dose-proportional [39]. One study revealed a near quadrupling of area under the plasma concentration-time curve (AUC)/minimum inhibitory concentration (MIC) when rifampin dose was increased from 300 to 600 mg [40]. In an observational study of Indonesian patients with pulmonary TB by Ruslami et al., AUC increased by a factor of 1.65 when the rifampin dose was increased from 450 to 600 mg [41]. The observed increase in pharmacokinetic exposure with dose has been linked to improved response using markers of efficacy in humans. In studies of early bactericidal activity, a linear increase in the activity of rifampin has been reproducibly demonstrated up to a dose of 1200 mg [42]. This body of evidence supports the concept that rifampin doses greater than 600 mg may increase treatment response during the first two months, which could, in turn, permit treatment shortening. To address this question, we embarked on a Phase II trial, entitled “Evaluation of high-dose rifampin in patients with new, smear-positive tuberculosis” or, HIRIF. The study is being conducted under investigational new drug application (# 106635) with the US Food and Drug Administration. The study protocol is presented here. Objectives HIRIF has three primary objectives:To assess the difference in steady state pharmacokinetic exposure of rifampin and 25-desacetyl-rifampin across three, daily, oral doses of rifampin (10, 15 and 20 mg/kg/day). This is done through evaluation of AUC0–24/MIC of rifampin at steady state. To assess the difference in sputum culture sterilization during the initial 8 weeks across all three rifampin doses. To compare the incidence of grade 2 or higher adverse events related to the study drug during the 8-week intensive phase of treatment, and up to 4 weeks later. Secondary objectives refine the understanding of safety and efficacy of higher doses of rifampin and explore potential surrogate endpoints for failure and relapse. Methods Design HIRIF is a multi-site, randomized, controlled, triple-blinded clinical trial assessing the pharmacokinetics, efficacy, and safety of higher doses of rifampin. Two experimental arms, rifampin 15 and 20 mg/kg/day, are compared to the control arm, rifampin 10 mg/kg/day. The study is randomizing 180 participants to one of three treatment arms in a 1:1:1 allocation. Randomization is blocked, but not stratified. Unblinded pharmacy staff implement treatment assignment and prepare weight-based prescriptions providing only blinded information to participants and other study staff. Study participation lasts 12 months. During the first 2 months, participants receive the randomly assigned dose of rifampin 7 days/week, in combination with standard doses of companion anti-TB drugs: isoniazid (H, 5 mg/kg/day), ethambutol (E, 20 mg/kg/day), and pyrazinamide (Z, 25 mg/kg/day). All study participants then receive a 4-month continuation phase of therapy with standard treatment doses (H: 10 mg/kg/day; RIF: 10 mg/kg/day) 3 days/week. Throughout, 50 mg of pyridoxine is administered 3 times/week to prevent peripheral neuropathy, a common side effect of isoniazid. Study participants are followed for 6 months after treatment completion (see Fig. 1).Fig. 1 Flow of 12-month study participation Setting HIRIF is being implemented in two districts in Lima, Peru (Lima Este and Lima Ciudad) where the local implementing partner, Socios En Salud (SES), operates. The year prior to study initiation, approximately 2500 cases of smear-positive TB were reported in the catchment area. Less than 4 % of TB cases were co-infected with human immunodeficiency virus (HIV) and roughly 10 % had diabetes mellitus. Potential participants are identified in 43 peripheral health centers and then referred to either Hospital Nacional Hipólito Unanue (HNHU) or Hospital Nacional Sergio E. Bernales (HNSEB), according to jurisdiction. Both hospitals are research centers certified by the Peruvian National Institute of Health (INS). The National TB Program of Peru endorsed the study and provides oversight and supervision to the recruiting health centers and research centers. Pre-screening sputum microscopy is performed by health center laboratories. Study microbiology is performed by the SES research microbiology laboratory, which is quality assured by the Peruvian INS and the College of American Pathologists. A research pharmacy approved by the Peruvian national regulatory authority and a private clinical lab completed the site resources. Study population and eligibility The study population comprises adults with newly diagnosed, previously untreated, smear positive (≥2+) pulmonary tuberculosis. Patients who meet these criteria on presentation to ambulatory care facilities in the two districts are invited to participate. They are then referred to study staff at the research centers for informed consent and eligibility screening. Main inclusion criteria for participation in HIRIF are newly diagnosed, pulmonary TB with acid-fast bacilli (≥2+) in a stained sputum smear, and susceptibility to isoniazid and rifampin detected by HAIN MTBDR+ test. Eligible participants are adults (18–60 years old), ≥30 kg, who have never been treated with multidrug anti-TB therapy for more than one month and who have no known intolerance or contraindications to the study drug or companion drugs, and who are not taking any additional drugs for which there may be potential negative drug interactions, synergistic toxicities, or contraindications. Other criteria include a Karnofsky score of ≥50. The following exclusions apply: central nervous system or miliary TB; pericardial or pleural involvement; significant hemoptysis; any uncontrolled condition that may interfere with drug absorption, distribution, metabolism or excretion; uncontrolled diabetes mellitus (glycocylated hemoglobin >7.5 %); serology positive for hepatitis B virus surface antigen or hepatitis C virus antibody; pulmonary silicosis; history of liver disease or current amino alanine transferase greater than 2 times the upper limit of normal (ULN); total bilirubin concentration greater than 2.5 times the ULN, creatinine concentration greater than two times the ULN (or creatinine clearance <60 mL/min), hemoglobin concentration <7.0 g/dL, platelet count <150,000/mm3, or white blood cell count <4500 cells/μL. Women of child-bearing potential must not currently be pregnant or breastfeeding and must agree to practice an effective double-barrier method of birth control during treatment. All participants must be willing to undergo HIV testing according to the National Health Guidelines for TB control in Peru; however, patients can be included in the trial regardless of HIV status. Finally, all participants must willingly sign the informed consent form, intend to remain within the jurisdiction of the health center throughout the study to facilitate monitoring and completion of follow-up, and be assessed to be capable of adhering to the study protocol. Treatment delivery and retention All study treatment doses are directly observed. Retention is assured through a system of treatment support and enablers. All treatment is ambulatory and delivered by dedicated directly observed therapy (DOT) supporters. Treatment adherence is assessed through reviews of treatment logs throughout the intensive and continuation phases. Transport costs for study visits are covered by the study. Participants receive regular food vouchers for their participation and meals during prolonged study visits. Assessment of study endpoints The primary pharmacokinetic (PK) endpoint, area under the concentration curve (AUC0–24), is assessed through blood sampling on a single day after steady state rifampin exposure is achieved [43] and before the intervention dose is discontinued, between 15 and 56 days post randomization. Participants are randomly assigned to either a sparse (samples are collected pre-dose and at two timepoints after dosing) or intensive (samples are collected pre-dose and at six timepoints after dosing) PK sampling group in a 2:1 allocation. Because AUC0–24/MIC is thought to be the pharmacokinetic-pharmacodynamic (PK-PD) parameter best correlated with anti-TB activity [21], the MIC of each participant’s pre-treatment infecting isolate is also estimated from early morning and overnight sputum samples collected at the pre-treatment visit. The primary efficacy endpoint, change in M. tuberculosis log10 colony forming units (CFU) in sputum, is assessed by counting log10 CFUs in sputum cultures, grown in 7H11 Middlebrook medium. Samples from the same timepoints are also cultured in the BACTEC 960 system (MGIT). These are used to estimate time to culture conversion and change in time to positivity in MGIT. All participants are sampled at baseline, then at 5 time points during intensive phase. Information from sequential pooled sputum samples is used to calculate decline in log10 CFUs. Additional early-morning sputum samples are collected for smear microscopy and culture in Löwenstein Jensen medium. The occurrence of adverse events, use of concomitant medications, and risk of pregnancy and/or breastfeeding are assessed at each study visit. Laboratory screenings are performed to identify hematologic and biochemical abnormalities throughout the study period. Clinical and laboratory findings are graded by clinical investigators according to the modified Adult Toxicity Table [Draft Nov. 2007] for the Division of Microbiology and Infectious Diseases (DMID), National Institute of Allergy and Infectious Diseases (NIAID), National Institutes of Health (NIH). Analysis (including power calculation) The study population of 180 patients (60 per treatment arm) is based on the following assumptions for PK, efficacy, and safety. Existing unpublished data suggest a minimum increase in AUC0–24 of 12 mcg/mL*hr across dose groups. Assuming a standard deviation of AUC0–24 of 24 mcg/mL*hr and α = 0.05, a linear contrast test across the three treatment groups, the study affords 90 % power to detect a total effect size of 14 mcg/mL*hr between the top and bottom dose levels at a sample size of 50 evaluable subjects per arm, permitting a 17 % loss to follow up. The estimation of the sample size for the efficacy endpoint is based on computation of the population Fisher’s Information matrix derived from linearization of a non-linear mixed effects model for the data. The parameter, θ4, in this model represents the late phase decay in colony counts, a surrogate measure of sterilizing activity. Under conventional assumptions of α = 0.05 and ß = 0.20 with a coefficient of variation on θ4 of 20 %, a sample size of 48 per arm is sufficient to detect a difference between the highest and lowest dose arms of 0.025 log10 CFU/ml. This is comparable in magnitude to differences observed in previous colony counting studies. A sample size of 60 per arm allows up to 25 % patient withdrawals. Among adverse events thought to be associated with rifampin dose, hepatotoxicity is one of the most worrying. Hepatotoxicity has been observed to occur in up to 27 % of patients receiving RIF-containing regimens for TB, with the summary frequency from one meta-analysis estimated at 2.7 % [44]. Other serious toxicities, such as hematologic disorders and flu-like syndrome are estimated to occur in between 1 and 5 % of patients on standard doses of RIF [45]. In a pivotal trial of rifapentine (compared to standard doses of RIF) 5 % of subjects receiving rifampin permanently discontinued treatment [45]. We expect at least 10 % of patients in the control arm to experience a grade 2 or higher event related to RIF. We have 62 % power (1-sided α = 0.1) to detect adverse events occurring twice as frequently in the intervention arms combined, and greater than 95 % power to detect a relative risk of ≥3. Since we are comparing incidence of adverse events, a continuous variable, the statistical power afforded by the study sample size is actually slightly higher. Analysis of primary endpoints The primary PK analysis will be a two-sided linear contrast test of dose-response of AUC0–24/MIC across the three dose groups with a significance level of 5 %. Secondary analyses of the summary parameters will include similar tests of dose-response for AUCo-∞ and Cmax. Exploratory analyses of additional determinants of the exposure parameters (AUC0–24, AUCo-∞ and Cmax) and clearance will include as covariates study site, body mass index, sex, and plasma concentration of companion drugs. Population PK modeling of the parent compound and its major metabolite, 25-desacetylrifampicin, using the rich and complete datasets will be carried out. The primary efficacy analyses will be two-sided linear contrast tests of the three dose levels on: 1) the parameter θ4 (late-phase sterilizing slope) derived from non-linear mixed effects modeling and 2) the hazard ratio of culture conversion in MGIT derived from the Cox proportional hazards model at a significance level of 5 %. These two endpoints, and their performance characteristics for predicting failure/relapse, will be compared with each other, and with the binary 2-month LJ culture conversion endpoint. To evaluate the effect of dose size on adverse events (AEs) grade 2 or higher, the primary analysis will be a comparison, between the intervention groups and the control group, of time to adverse events occurring in the first 12 weeks after randomization and determined to be related to rifampin. In secondary analyses, the rate of discontinuation due to hepatotoxicity will be compared between the intervention groups and the control group. If a significant difference is detected, comparison will be made between the 15 mg/kg intervention group and the control as well as between the 20 mg/kg intervention group and the control. Dissemination of trial findings The study principal investigators hold primary responsibility for the preparation of publications. Once the trial is complete, the investigators anticipate publishing results of this study in several manuscripts in peer-reviewed scientific journals. In compliance with the policy of International Committee of Medical Journal Editors, this trial is registered in ClinicalTrials.gov: NCT01408914. Discussion The HIRIF trial is a Phase IIB design that incorporates several approaches that are novel in the tuberculosis field. Most importantly, the study is designed to fully support a rigorous pharmacokinetic-pharmacodynamic analysis, which is appropriate to the goals of an early- phase dose-ranging trial. The pharmacokinetic aspect of the study is based on a population approach which is facilitated by the intensive-sparse sampling design and enables pharmacokinetic exposure (AUC0–24) to be estimated for all participants in the study. In addition, RIF MIC will be obtained for each participant enabling more direct comparisons with existing preclinical data for the first time in a human clinical trial and accounting for this important source of variability in treatment response [46]. Finally, for increased statistical power [21], the pharmacodynamic outcomes are powered on the basis of quantitative bacteriology, specifically serial sputum colony counting, rather than on the more traditional two-month culture conversion endpoint. A balanced-blocks design of staggered sputum sampling times based on prior studies was adopted to make this approach convenient for patients and logistically feasible for the laboratory [47]. Several alternative biomarkers of treatment response are also being evaluated using these samples. HIRIF addresses an important evidence gap in the treatment of TB. Although the work that led to the implementation of the 6-month, 4-drug regimen containing daily, 600 mg doses of rifampin was highlighted as a model for medical interventions [48], the optimal dosing for rifampin was overlooked in the series of trials that led to these laudatory remarks. Cost trumped efficacy in the selection of regimens tested in the Medical Research Council trials that led to the development of the current standard of care. That tens of millions of patients may have been subjected to suboptimal doses is an inexcusable tragedy, which must not be repeated in future drug-development efforts. In the case of rifampin, cost is no longer an issue as it is currently produced by several quality-assured manufacturers and sold for pennies per dose: 3.7–4.7 cents per 150-mg rifampicin tablet/capsule, or 7.3–8.5 cents per 300-mg rifampicin tablet/capsule [49]. Results from HIRIF will complement those from other recently conducted studies. These include another Phase II study conducted by the PANACEA Consortium in Tanzania, known as HIGHRIF2 (NCT00760149). This study examined the same doses of rifampin in 180 participants with similar endpoints. Both studies were powered for the PK endpoints, which required smaller sample sizes than efficacy and safety endpoints. Data from the two studies will be pooled to provide greater power to assess dose-related efficacy and toxicity endpoints. PANACEA also has conducted a maximum-dose tolerability study, HIGHRIF1 [38]. Among participants who received up to 35 mg/kg of RIF, both monotherapy and multidrug therapy, there were no serious adverse events. In light of this information, the consortium embarked on a Phase II/III adaptive study, MAMS-TB-01 trial (NCT01785186), that examined rifampin doses of 35 mg/kg in combination therapy. Lastly, RIFATOX (ISRCTN55670677), a toxicity study of higher doses (600 mg vs. 900 and 1200 mg) of rifampin found no association between rifampin dose and toxicity. When HIRIF results are available, we will examine the whole body of recently generated evidence to determine if and what further investigation is necessary. Options include additional Phase II work with higher doses for longer duration than in HIGHRIF1. Also possible is moving forward with Phase III studies of either the 15 or 20 mg/kg dose, depending on results from HIRIF and results pooled with HIGHRIF2. Any proposed study would complement those already conducted or underway (i.e., PANACEA, the proposed RIFASHORT [NCT02581527]). And, it is possible that based on combined efficacy and safety data, there will be no justification for further investigation of higher doses of rifampin for new, smear-positive pulmonary TB. Irrespective of the final outcome, however, we can be confident that an historical wrong has been righted: the dose of rifampin used for TB in the future will reflect the complete evidence on PK, efficacy, and safety assessed using modern methods, available in the 21st century. Abbreviations AEAdverse events AUCArea under the plasma concentration-time curve CFUColony forming units CmaxMaximum serum concentration DMIDDivision of Microbiology and Infectious Diseases, National Institute of Allergy and Infectious Diseases, National Institutes of Health DOTDirectly observed therapy EEthambutol HIsoniazid HAINLine probe assay to test for resistance to isoniazid and rifampin HIRIFStudy acronym HIVHuman immunodeficiency virus HNHUHospital Nacional Hipólito Unanue HNSEBHospital Nacional Sergio E. Bernales INSPeruvian National Institute of Health IRBInstitutional Review Board MICMinimum inhibitory concentration NNumber (refers to number of subjects) PDPharmacodynamics PKPharmacokinetics RIFRifampin SESSocios En Salud TBTuberculosis ULNUpper limit of normal ZPyrazinamide Acknowledgements We would like to acknowledge the Department of Microbiology and Infectious Diseases, National Institute of Allergy and Infectious Diseases, National Institutes of Health which supports the trial and time of the co-authors on the preparation of this manuscript under Award Number U01A1091429. Funding This trial and the time of the co-authors on the preparation of this manuscript is supported by the Department of Microbiology and Infectious Diseases, National Institute of Allergy and Infectious Diseases, National Institutes of Health under Award Number U01A1091429. Availability of data and materials Not applicable. Authors’ contributions CM, GD, and MM wrote the first draft of the paper. All authors provided significant feedback. CM and MM prepared the final version for publication. All authors have approved the final version. Competing interests The authors declare that they have no competing interests. Consent for publication Not applicable. Ethics approval and consent to participate The study protocol and informed consent documents were reviewed and approved by: the Partners Healthcare Human Research Committee (approval number 2011P001577), the Liverpool School of Tropical Medicine Research Ethics Committee and the University of Liverpool Research Support Office (approval number RETH000725), Institutional Ethics Committees at HNHU and University of San Martin de Porres (under ceding agreement with the study hospital, HNSEB). The study received exemption from the University of Florida IRB (exemption number IRB201200199). Additionally, the trial was approved by the Peruvian INS (approval number 081–12) and the Division of Microbiology and Infectious Diseases, National Institutes of Allergy and Infectious Diseases, National Institutes of Health (protocol number 11–0050). ==== Refs References 1. WHO. Global Tuberculosis Report 2015. Geneva, Switzerland: World Health Organization; 2015. 2. TB Alliance: Global Alliance for TB Drug Development New TB Regimens: What Countries Want 2009 3. Owens JP Fofana MO Dowdy DW Cost-effectiveness of novel first-line treatment regimens for tuberculosis Int J Tuberc Lung Dis 2013 17 5 590 6 10.5588/ijtld.12.0776 23575322 4. Johnson JL Hadad DJ Dietze R Maciel EL Sewali B Gitta P Shortening treatment in adults with noncavitary tuberculosis and 2-month culture conversion Am J Respir Crit Care Med 2009 180 6 558 63 10.1164/rccm.200904-0536OC 19542476 5. Dawson R Diacon AH Everitt D van Niekerk C Donald PR Burger DA Efficiency and safety of the combination of moxifloxacin, pretomanid (PA-824), and pyrazinamide during the first 8 weeks of antituberculosis treatment: a phase 2b, open-label, partly randomised trial in patients with drug-susceptible or drug-resistant pulmonary tuberculosis Lancet 2015 385 9979 1738 47 10.1016/S0140-6736(14)62002-X 25795076 6. Gillespie SH Crook AM McHugh TD Mendel CM Meredith SK Murray SR Four-month moxifloxacin-based regimens for drug-sensitive tuberculosis N Engl J Med 2014 371 17 1577 87 10.1056/NEJMoa1407426 25196020 7. Jindani A Harrison TS Nunn AJ Phillips PP Churchyard GJ Charalambous S High-dose rifapentine with moxifloxacin for pulmonary tuberculosis N Engl J Med 2014 371 17 1599 608 10.1056/NEJMoa1314210 25337749 8. Merle CS Fielding K Sow OB Gninafon M Lo MB Mthiyane T A four-month gatifloxacin-containing regimen for treating tuberculosis N Engl J Med 2014 371 17 1588 98 10.1056/NEJMoa1315817 25337748 9. Burman WJ Goldberg S Johnson JL Muzanye G Engle M Mosher AW Moxifloxacin versus ethambutol in the first 2 months of treatment for pulmonary tuberculosis Am J Respir Crit Care Med 2006 174 3 331 8 10.1164/rccm.200603-360OC 16675781 10. Dorman SE Johnson JL Goldberg S Muzanye G Padayatchi N Bozeman L Substitution of moxifloxacin for isoniazid during intensive phase treatment of pulmonary tuberculosis Am J Respir Crit Care Med 2009 180 3 273 80 10.1164/rccm.200901-0078OC 19406981 11. Dorman SE Goldberg S Stout JE Muzanyi G Johnson JL Weiner M Substitution of rifapentine for rifampin during intensive phase treatment of pulmonary tuberculosis: study 29 of the tuberculosis trials consortium J Infect Dis 2012 206 7 1030 40 10.1093/infdis/jis461 22850121 12. Peloquin C What is the ‘right’ dose of rifampin? Int J Tuberc Lung Dis 2003 7 1 3 5 12701829 13. Mitchison DA Role of individual drugs in the chemotherapy of tuberculosis Int J Tuberc Lung Dis 2000 4 9 796 806 10985648 14. Mitchison DA Antimicrobial therapy of tuberculosis: justification for currently recommended treatment regimens Semin Respir Crit Care Med 2004 25 3 307 15 10.1055/s-2004-829503 16088472 15. Davies GR Nuermberger EL Pharmacokinetics and pharmacodynamics in the development of anti-tuberculosis drugs Tuberculosis 2008 88 Suppl 1 S65 74 10.1016/S1472-9792(08)70037-4 18762154 16. Mitnick CD McGee B Peloquin CA Tuberculosis pharmacotherapy: strategies to optimize patient care Expert Opin Pharmacother 2009 10 3 381 401 10.1517/14656560802694564 19191677 17. van Ingen J, Aarnoutse RE, Donald PR, Diacon AH, Dawson R. Plemper van Balen G, et al. Why do we use 600 mg of rifampicin in tuberculosis treatment? Clin Infect Dis. 2011;52(9):e194–9. 18. Rifampicin. Tubercle. 1969;50(3):318–20. doi:10.1016/0041-3879(69)90058-0. 19. Peloquin C Vernon A Yu V Edwards G McKinnon PS Peloquin C Morse GD Antimycobacterial Agents: Rifamycins for Mycobacterial Infections Antimicrobial Chemotherapy and Vaccines, 2nd edition, Volume II: Antimicrobial Agents 2005 Pittsburgh ESun Technologies 383 402 20. Dickinson JM Jackett PS Mitchison DA The effect of pulsed exposures to rifampin on the uptake of uridine- 14 C by Mycobacterium tuberculosis Am Rev Respir Dis 1972 105 4 519 27 4622857 21. Jayaram R Gaonkar S Kaur P Suresh BL Mahesh BN Jayashree R Pharmacokinetics-pharmacodynamics of rifampin in an aerosol infection model of tuberculosis Antimicrob Agents Chemother 2003 47 7 2118 24 10.1128/AAC.47.7.2118-2124.2003 12821456 22. Grumbach F Experimental “in vivo” studies of new antituberculosis drugs: capreomycin, ethambutol, rifampicin Tubercle 1969 50 Suppl 12 21 5402543 23. Kradolfer F Rifampicin, isoniazid, ethambutol, ethionamide, and streptomycin in murine tuberculosis: comparative chemotherapeutic studies Antibiotica et chemotherapia Fortschritte Advances Progres 1970 16 352 60 5523730 24. Verbist L Rifampicin activity “in vitro” and in established tuberculosis in mice Acta Tuberc Pneumol Belg 1969 60 3 397 412 4984420 25. Drusano GL Antimicrobial pharmacodynamics: critical interactions of ‘bug and drug’ Nat Rev Microbiol 2004 2 4 289 300 10.1038/nrmicro862 15031728 26. Decroix G, Kreis B, Sors C, Birembaum J, Le Lirzin M, Canetti G. Comparison between regimes of rifampicin-isoniazid administered daily and administered twice a week (initial results of a comparative study conducted in 4 medical services of the Parisian region). Rev Tuberc Pneumol. 1969;33(6):751–68.  27. Nitti V Antituberculosis activity of rifampin. Report of studies performed and in progress (1966–1971) Chest 1972 61 6 589 98 10.1378/chest.61.6_Supplement.589 4624281 28. Acocella G Pharmacokinetics and metabolism of rifampin in humans Rev Infect Dis 1983 5 Suppl 3 S428 32 10.1093/clinids/5.Supplement_3.S428 6356276 29. Poole G Stradling P Worlledge S Potentially serious side-effects of high-dose twice-weekly rifampicin Postgrad Med J 1971 47 553 727 47 10.1136/pgmj.47.553.742 5159676 30. Aquinas M Allan WG Horsfall PA Jenkins PK Hung-Yan W Girling D Adverse reactions to daily and intermittent rifampicin regimens for pulmonary tuberculosis in Hong Kong Br Med J 1972 1 5803 765 71 10.1136/bmj.1.5803.765 4259217 31. A comparative study of daily followed by twice- or once-weekly regimens of ethambutol and rifampicin in the retreatment of patients with pulmonary tuberculosis: second report. Tubercle. 1976;57(2):105–13. 32. Girling DJ Adverse reactions to rifampicin in antituberculosis regimens J Antimicrob Chemother 1977 3 2 115 32 10.1093/jac/3.2.115 324964 33. Dhar S Kaur I Sharma VK Kumar B “Flu” syndrome due to rifampin; experience with four cases Int J Lepr Other Mycobact Dis 1995 63 1 92 4 7730725 34. Reichler MR Allphin AA Breiman RF Schreiber JR Arnold JE McDougal LK The spread of multiply resistant Streptococcus pneumoniae at a day care center in Ohio J Infect Dis 1992 166 6 1346 53 10.1093/infdis/166.6.1346 1431252 35. Drancourt M Stein A Argenson JN Roiron R Groulier P Raoult D Oral treatment of Staphylococcus spp. infected orthopaedic implants with fusidic acid or ofloxacin in combination with rifampicin J Antimicrob Chemother 1997 39 2 235 40 10.1093/jac/39.2.235 9069545 36. Baty V Hoen B Schuhmacher H Amiel C Reyrolle M Garin H Legionella jordanis pneumonia unresponsive to fluoroquinolones in a non-immunocompromised host Scand J Infect Dis 1997 29 3 319 20 10.3109/00365549709019054 9255901 37. Kochar DK Aseri S Sharma BV Bumb RA Mehta RD Purohit SK The role of rifampicin in the management of cutaneous leishmaniasis QJM 2000 93 11 733 7 10.1093/qjmed/93.11.733 11077029 38. Boeree MJ Diacon AH Dawson R Narunsky K du Bois J Venter A A dose-ranging trial to optimize the dose of rifampin in the treatment of tuberculosis Am J Respir Crit Care Med 2015 191 9 1058 65 10.1164/rccm.201407-1264OC 25654354 39. McIlleron H Wash P Burger A Norman J Folb PI Smith P Determinants of rifampin, isoniazid, pyrazinamide, and ethambutol pharmacokinetics in a cohort of tuberculosis patients Antimicrob Agents Chemother 2006 50 4 1170 7 10.1128/AAC.50.4.1170-1177.2006 16569826 40. Sirgel FA Fourie PB Donald PR Padayatchi N Rustomjee R Levin J The early bactericidal activities of rifampin and rifapentine in pulmonary tuberculosis Am J Respir Crit Care Med 2005 172 1 128 35 10.1164/rccm.200411-1557OC 15805182 41. Ruslami R Nijland H Aarnoutse R Alisjahbana B Soeroto AY Ewalds S Evaluation of high- versus standard-dose rifampin in Indonesian patients with pulmonary tuberculosis Antimicrob Agents Chemother 2006 50 2 822 3 10.1128/AAC.50.2.822-823.2006 16436757 42. Diacon AH Patientia RF Venter A van Helden PD Smith PJ McIlleron H Early bactericidal activity of high-dose rifampin in patients with pulmonary tuberculosis evidenced by positive sputum smears Antimicrob Agents Chemother 2007 51 8 2994 6 10.1128/AAC.01474-06 17517849 43. Borin MT Chambers JH Carel BJ Gagnon S Freimuth WW Pharmacokinetic study of the interaction between rifampin and delavirdine mesylate Clin Pharmacol Ther 1997 61 5 544 53 10.1016/S0009-9236(97)90134-X 9164416 44. Steele MA Burk RF DesPrez RM Toxic hepatitis with isoniazid and rifampin. A meta-analysis Chest 1991 99 2 465 71 10.1378/chest.99.2.465 1824929 45. Rifapentine (Priftin) [package insert]. sanofi-aventis U.S. LLC, Bridgewater, NJ 08807. Revised December 2014. http://products.sanofi.us/priftin/priftin.pdf. 46. Burhan E Ruesen C Ruslami R Ginanjar A Mangunnegoro H Ascobat P Isoniazid, rifampin, and pyrazinamide plasma concentrations in relation to treatment response in Indonesian pulmonary tuberculosis patients Antimicrob Agents Chemother 2013 57 8 3614 9 10.1128/AAC.02468-12 23689725 47. Davies GR Khoo SH Aarons LJ Optimal sampling strategies for early pharmacodynamic measures in tuberculosis J Antimicrob Chemother 2006 58 3 594 600 10.1093/jac/dkl272 16857690 48. Cochrane AL Archie Cochrane in his own words. Selections arranged from his 1972 introduction to “Effectiveness and Efficiency: Random Reflections on the Health Services” 1972 Control Clin Trials 1989 10 4 428 33 10.1016/0197-2456(89)90008-1 2691208 49. Global Drug Facility Procurement and supply. Stop TB Partnership 2015

PMC005xxxxxx/PMC5002099.txt

==== Front BMC GenomicsBMC GenomicsBMC Genomics1471-2164BioMed Central London 302810.1186/s12864-016-3028-0SoftwareG23D: Online tool for mapping and visualization of genomic variants on 3D protein structures Solomon Oz oz.solomon@live.biu.ac.il 12Kunik Vered Vered.Kunik@sheba.health.gov.il 1Simon Amos Amos.Simon@sheba.health.gov.il 3Kol Nitzan Nitzan.cal@sheba.health.gov.il 1Barel Ortal Ortal.Barel@sheba.health.gov.il 1Lev Atar Atar.Lev@sheba.health.gov.il 3Amariglio Ninette ninette.amariglio@sheba.health.gov.il 12Somech Raz raz.somech@sheba.health.gov.il 45Rechavi Gidi gidi.rechavi@sheba.health.gov.il 15http://orcid.org/0000-0002-9505-6319Eyal Eran eran.eyal@sheba.health.gov.il 11 Cancer Research Center, Sheba Medical Center, Ramat-Gan, Israel 2 The Mina and Everard Goodman Faculty of Life Sciences, Bar-Ilan University, Ramat-Gan, Israel 3 Pediatric Immunology Service, Jeffrey Modell Foundation, Sheba Medical Center, Ramat-Gan, Israel 4 Edmond and Lily Safra Children’s Hospital, Sheba Medical Center, Ramat-Gan, Israel 5 Sackler School of Medicine, Tel Aviv University, Tel Aviv, Israel 26 8 2016 26 8 2016 2016 17 1 68124 5 2016 19 8 2016 © The Author(s). 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Background Evaluation of the possible implications of genomic variants is an increasingly important task in the current high throughput sequencing era. Structural information however is still not routinely exploited during this evaluation process. The main reasons can be attributed to the partial structural coverage of the human proteome and the lack of tools which conveniently convert genomic positions, which are the frequent output of genomic pipelines, to proteins and structure coordinates. Results We present G23D, a tool for conversion of human genomic coordinates to protein coordinates and protein structures. G23D allows mapping of genomic positions/variants on evolutionary related (and not only identical) protein three dimensional (3D) structures as well as on theoretical models. By doing so it significantly extends the space of variants for which structural insight is feasible. To facilitate interpretation of the variant consequence, pathogenic variants, functional sites and polymorphism sites are displayed on protein sequence and structure diagrams alongside the input variants. G23D also provides modeling of the mutant structure, analysis of intra-protein contacts and instant access to functional predictions and predictions of thermo-stability changes. G23D is available at http://www.sheba-cancer.org.il/G23D. Conclusions G23D extends the fraction of variants for which structural analysis is applicable and provides better and faster accessibility for structural data to biologists and geneticists who routinely work with genomic information. Keywords VariantMutationStructureProteinVisualizationissue-copyright-statement© The Author(s) 2016 ==== Body Background Understanding the consequence of protein-coding point mutations is crucial to elucidate mechanisms of function and disease. This need has become even more urgent in recent years as next generation sequencing technologies and downstream pipelines typically identify many mutations which should be rapidly and accurately prioritized to assess their relevance. Most of the recent sequencing studies specify positions of interest according to the coordinates of the reference genome. Nevertheless, the numbering system of proteins in general and protein structures in particular, is by essence very different from that of the reference genome. Thus, to bridge the gap between genomic coordinates and those of proteomics and structural biology, efficient and convenient conversion tools are needed. To date, evaluation of genomic variants is mainly based on sequence features and sequence conservation scores. Structural data are rarely applied due to limited structural coverage of the human proteome and the above mentioned technical obstacles. However when applicable, structural data can be used to accurately calculate free energy changes, locate the spatial position of the residue with respect to known critical positions in the protein and known intermolecular interfaces. This issue is an example of the major challenge we are facing in the high throughput sequencing (HTS) era, which is to integrate efficiently and conveniently large amount of data from distinct origins. The tool presented here, G23D, enables mapping and visualization of genomic variants on three dimensional (3D) structures of proteins and helps integrate genomic data (either user provided or data from public databases) and protein structural data. Tools for conversion from genomic coordinates to structural data already exist, but their functionality is often limited. MuPit [1] provides an interface to locate genomic positions onto available predetermined structures of the exact protein. Polyphen2 [2] and Mutation Assessor [3] provide structural information, albeit limited, for lists of variant. Some tools like SNPs3D [4], MutDB [5], LS-SNP/PDB [6] and Cn3D (http://www.ncbi.nlm.nih.gov/Structure/CN3D/cn3d.shtml) enable the user to explore variants which are stored at variant databases but not to upload new user defined variants. A number of other cancer related tools, including Cancer3D (http://www.cancer3d.org/) and the CBI portal (http://www.cbioportal.org/) also allow structural analyses. Tools which are not web-based exist as well [7, 8]. Ball-SNP [8] for example, allows integration of UCSF chimera with Cytoscape and other resources for integrative visualization of networks, sequences and structures. These tools enable the use of more sophisticated modelling features, but are obviously, less accessible to the community. There are also other good tools, like ELAPSIC [9] and SNPeffect [10] for analysis of functional, stability and protein-protein interactions changes upon point mutations which requires protein or structure information input, and do not support conversions between genomic and protein coordinates. The new tool we present here, G23D, further helps to extend the utilization of structural data in several different ways. The structural data which is being scanned includes also homologous proteins and theoretical models. The importance of including homologous proteins in structural analysis has recently been demonstrated, in particular for analysis of interactions [11, 12]. G23D retrieves and displays hits from the PDB [13] which are not only identical but also similar to the input protein (i. e. homologous), under the well supported assumption that related sequences adopt a similar structural fold. Moreover, G23D also utilizes theoretical models from ModBase [14]. Thus, G23D facilitates structural analyses for a much broader space of proteins. We aimed to provide flexibility in selecting the appropriate structure template for a given genomic location. Therefore, the user can choose the most appropriate reference structure in each case and has considerable control over the filtering of the structural hits according to criteria such as similarity level with the input protein and the quality of the structure. G23D provides convenient and rapid conversion of genomic coordinates (or protein positions) to structural coordinates. It then allows instant on-line visualization of the structural context of the input variants, including modeling of the mutated amino acid and simultaneous visualization of the input variant alongside known pathogenic and non-pathogenic variants. Additional structure-based analysis features are provided as discussed below. Implementation The most significant steps in the technical implementation of the site are the conversions between genomic coordinates and protein sequence coordinates and between protein sequence coordinates and protein structure coordinates (Fig. 1). G23D employs third party databases and newly developed offline pre-processing steps to obtain these conversions in advance.Fig. 1 Schematic flowchart of the G23D pipeline and the resources being used. Genomic coordinates are converted to protein coordinates using dbNSFP. The proteins are then used to retrieve models from ModBase and their sequence is used for a blast search against the PDB. S2C (Roland Dunbrack lab) is used to convert sequence position to index of coordinates within PDB files. JSmol is used to visualize the hits. Side chain modeling programs are applied to model the mutant side chain. Comparative contact analyses using contact surface areas and stability analyses are based on the structure hits. Shaded boxes indicate steps which are pre-processed in advance and not during run-time dbNSFP [15] is used to convert genomic coordinates (GRCh37 or GRCh38 human reference) to uniprot protein coordinates. dbNSFP is frequently updated and currently version 3.1 is being used. Protein blast [16] is then applied for alignment of all Uniprot entries versus the PDB. The alignment not only retrieves PDB entries of the relevant proteins, but also detects structures of related proteins, either orthologs or paralogs, which can be used for structural analyses of the subject protein to a certain degree. S2C (http://dunbrack.fccc.edu/Guoli/s2c/) is applied for conversion of sequence and residue numbers from the PDB ATOM coordinates record. The G23D site is mostly implemented in CGI combined with JavaScript and SVG. The 3D visualization is based on JSmol [17] which is a java-independent tool supported by all operating systems and major browsers without the need to install additional software or plugins. The site is composed of three main layers: an input layer, hits selection layer and an analysis layer. The requested variants are submitted to the input layer. The structures which cover these coordinates of the variants are then displayed in the hits layer, where hits can be filtered by various criteria. Selected hits can then be visualized and studied in the analysis layer which includes the JSmol molecular graphics session and additional structural information and analyses options. More detailed description of the various components is provided in the results section. G23D employs data from many distinct sources. Proteins 3D structures are extracted from the PDB [13]. Protein theoretical models are retrieved from ModBase [14]. Protein sequences were downloaded from Uniprot [18]. Variants information was taken from COSMIC [19] and ClinVar [20]. Single nucleotide polymorphisms (SNPs) were taken from the common_no_known_medical_impact.vcf file of ClinVar (variants seen in healthy human population). Protein residue conservation scores were taken from HSSP [21]. Motifs and domains were taken from Prosite [22]. Protein secondary structures and disordered regions were downloaded from DisProt [23]. As this work describes implementation of software and does not involve human materials, confidential human data or animal models, no ethics approval has been required. Results Description G23D is a web-based tool which can be accessed by providing genomic coordinates, dbSNP id or protein position. Alternatively, an official gene symbol or a protein name can be provided as an input. Following submission of the requested information in the input page (Fig. 2) the coordinates are converted to protein space using dbNSFP [15]. A search is then performed to check which structure entries cover the input site. Sequences of PDB hits and ModBase hits are aligned to the sequence of the protein spanning the input coordinates. The user can easily explore the exact regions spanned by the structures and the similarity level to the protein of interest, as indicated by the color gradients (Fig. 3a). The same page also contains more detailed 3-way alignments of the structure hits (Fig. 3b). The first sequence is that of the query protein. Several query sequences may appear in this page as the mutation might be spanned by several distinct isoforms. The second line shows the sequence of the protein of the structure hit (SEQRES information in case of a PDB entry). The third line shows the part of the protein for which a 3D structure is available (ATOM part of the PDB entry). Missing domains, disordered regions and other undetermined regions will therefore not appear in this line. The amino-acid position affected by the input mutation (if provided) is highlighted. A link to the JSmol visualization page follows the alignment.Fig. 2 G23D front page includes the forms in which genomic coordinates, protein coordinates, dbSNP id, official gene symbol or protein name can be supplied. For genomic coordinates we support versions hg19 and hg38 (GRCh37 or GRCh38, respectively) of the human genome Fig. 3 The hits’ page displays schematic and detailed alignments of the structural hits identified with the input protein (or the proteins) spanning the input genomic coordinates. a Schematic display of the alignment and the spanned regions. Horizontal colored bars, presented in a manner similar to Blast’s results page, serve as links to the detailed alignments located later in the page. The green bar indicates the query protein; red bars indicate PDB hits; while blue bars indicate ModBase models of the query proteins. Several isoforms might be found for the input protein. b Detailed alignments of the query protein (upper sequence line) and structure hits. Two sequence lines appear for the structure hit. One shows the protein sequence (PDB SEQRES record) and the second shows the sequence of the resolved coordinates (PDB ATOM record). In the first two alignments in this example (Q61 in NRAS) the loop which contains the variant is missing while in the third it is present. Description of the structure hits and indications for the alignment quality precede the alignment display. Residues which appear in dbSNP, Cosmic and Clinvar databases are indicated by various colors on the query sequence. The input variant is indicated by green arrows below and above the alignments. A link to the JSmol structure visualization page follows the alignment The selected structure hits are displayed in a JSmol session (Fig. 4). The 3D display is shown in the center (Fig. 4a) in cartoon representation. The input variant, if indicated in the input page and included in the structure, appears in black. Database variants (if present) appear in the same color scheme of the sequence alignment and are shown in stick representation. The structural context of the input variant can therefore be evaluated with respect to known variants. Currently, G23D displays data from dbSNP [24], ClinVar [20], Cosmic variants [19] as well as catalytic residues from Catalytic Site Atlas [25]. Interactive sequence alignment, similar to that found in the hits’ page, is located in the bottom of the page (Fig. 4b). Residues selected in the sequence panel are highlighted on the structure. This panel includes also information regarding conservation, secondary structure and disorder regions, if available. Prosite [22] motifs are also shown in the sequence panel and are mapped to the structure. The Control panel (Fig. 4c) allows manipulation of the structure sessions (many more control options are available in the JSmol menu). The control panel also contains links to the contact analyses and stability predictions.Fig. 4 The structure visualization page is divided to three panels. a The protein is shown in cartoon representation, and the input variant (light green) and database variants (red, yellow and green with the same color scheme of the sequence alignment) are shown in stick representation. The sequence alignment b is similar to that found in the hits’ page. Additional layers of information such as conservation, secondary structure and disorder may follow the sequences. Residues on the sequence can be marked on the structure diagram. The Control panel c provides easy access to some common graphical features, alternate visualization of the wild-type and the mutant structures and links to contact analyses and stability predictions Using the combined sequence-structure presentation, the user can easily explore the structural context of selected sequence locations and patterns. The sequence conservation values from HSSP [21] can also be displayed on both the sequence and the structure using conservation color scale. Side chain modeling and contact analysis G23D is not solely a visualization tool, as it provides additional modeling and analyses features which generally cannot be found in equivalent tools. The amino acid in the mutated position is modeled using SCcomp [26] or Scwrl [27]. Both programs are quite accurate but it should be stressed that both methods model side chains on a fixed backbone, so if backbone conformational changes are involved the model may not be accurate. Therefore caution is needed when the mutant side chain is larger than the wild type side chain in the protein core. In the 3D interactive session the user can explore the side chain conformation of the mutant, instead or alongside the wild type residue (Fig. 5a).Fig. 5 Modeling and stability prediction features in G23D. a A G23D model displaying the mutant structure instead, or together with, the wild type. The modeling is done using Sccomp, a side chain prediction program. b Contact surface areas between atoms and solvent accessible surfaces are analytically calculated using Voronoi tessellation. This procedure allocates the surface area of each atom to its neighbors (colored arcs) and the solvent (dashed arc). c A residue-level summary table lists all contacts and minimal distance to contacting residues for both wild type and mutant structures. Database residues are colored according to the general coloring of G23D. d By pressing on a residue in the left column of the residue-level table, the user can explore all atomic level interactions between the input variant and the chosen residue Given the structure of the wild-type protein and the mutation position, the user can explore the molecular contacts in which the residue participates. Contact surface areas and solvent accessible surface areas are calculated using analytic procedure, as described in Mcconkey et al. [28, 29], which applies Voronoi tessellation to allocate contact surfaces between neighboring atoms. The remaining surface of each atom, not assigned to contact with other atoms, is the solvent accessible surface (Fig. 5b). The user obtains a table (Fig. 5c) which summarizes the data at the residue level. Each line in the table provides information on a single residue which forms contact/s with the residue coded by the input variant. These contacting residues are colored according to the same coloring schemes of G23D to help the user assess their significance. The information in the table includes the contact surface area (Å2) and the minimal atomic distance (Å) between any two atoms of the two residues. The user can further explore atomic contacts of contacting residues. Pressing the residue number in the left column of the table opens a new table, which provides atomic level information regarding interactions between atoms of the mutated residue and atoms of the interacting residue (Fig. 5d), including inter atomic contact surface area and inter atomic distance. Altogether, the user can evaluate the effect of the mutation by comparing the contacts and solvent accessible surface area of both the wild type residue and the mutant residue. The user can also manually examine the possible effect of the mutations over the 3D molecular graphics session. G23D also provides links to third party programs which predict the consequences of the mutation by various considerations. If the variant is supplied by its genomic coordinates, then a link to the data of the relevant entry in dbNSFP [15] is provided. This database holds pre-compiled predictions for each protein coding variant in the genome regarding the possible functional significance of the variant. Predictions are available for eight different popular programs including SIFT, polyphen2, LRT, MutationTaster, MutationAssessor, FATHMM and PROVEAN. Evolutionary conservation is the most important individual feature in these predictions which generally do not consider structural features. Consensus predictions (i.e. meta predictions) are also provided and are in principal more accurate than the individual predictors [30]. In case structural information is available, G23D provides links to two different programs for thermostablity predictions, FoldX [31] and I-Mutant-2.0 [32]. The predictions of these tools are available from the 3D session page. The DNA methyltransferase 3B (DNMT3B) case study DNMT3B is a gene responsible for de novo cytosine-5-methyltranferase (m5C) in the human genome. Familial mutations in this gene were reported to be the cause of autosomal recessive Immunodeficiency Centromeric instability and facial anomalies (ICF) syndrome [33]. Fig. 6 demonstrates how G23D can assist in meaningful structural analysis of a variant in this gene. This example will serve as an illustrative case study as well as a short tutorial for the utility of G23D.Fig. 6 Example of G23D variant analysis in the catalytic domain of methyl-transferase (DNMT3B). Variants in this protein and specifically in position A585 are known to be related to the ICF syndrome. a DNMT3B has no resolved structures in the PDB, but by using G23D the user can rapidly detect structures of the close homolog DNMT3A (red) and several theoretical models which span the variant (blue). The color gradient indicates the sequence similarity to the PDB protein or to the template protein used for the homology modeling. b Analysis based on the structure of the catalytic domain of DNMT3A which shares 70 % identity with DNMT3B, suggests that the residue in position 585 (shown in black) is completely buried in the protein core and is located on the interface between a helix and a loop. It is positioned close to known pathogenic residues in both ClinVar (red colors) which are automatically being displayed in G23D and HGMD (here manually added in purple). Many other positions appear in Cosmic (yellow). The methyl donor (SAH), whose binding pocket is likely affected by mutations in position 585, appears in magenta. c From the sequence diagram, located in the bottom of the JSmol structure session it can be deduced that the region is very conserved (conservation score of 9) and that it is part of the known Prosite methyltransferase motif (PS51679; in pink). d Contact analysis shows that position 585 has direct contact with some of the known disease related positions shown in panel B, and that the minimal distance of the mutated residue causes steric clashes with two neighbors (marked in arrows). e Stability analysis of the A585T variant (equivalent to position 644 in PDB 4u7p ATOM section) using I-Mutant-2, directly accessed from G23D, suggests decreased stability of the mutant structure A patient at two years of age was admitted to our hospital due to severe failure to thrive (FTT), immunodeficiency and diarrhea. She was the first born child of consanguineous parents from a Palestinian descent. In order to determine the genetic cause for her syndrome we performed whole exome sequencing (WES) of DNA extracted from her peripheral blood. The analysis revealed a strong candidate variant in chromosome 20, position 31387128 (hg19) from G to A, lying inside the genomic region of DNMT3B (c.1753G > A). Providing the genomic coordinates in the input page of G23D, instantly revealed that there are no PDB structure of DNMT3B. There are, however, several structures of the closely related DNMT3A (70 % identity in the catalytic domain) which cover the C-terminal part of the protein (Fig. 6a). The input variant resides in the catalytic domain located in the C-terminal part. There are also several homology models of the DNMT3B catalytic domain in ModBase. According to isoform Q9UBC3, the amino acid position affected by the change is Ala585 and the substituted amino acid is Thr (codon GCG to ACG). Note that if genomic position is provided, the protein position might differ for other isoforms. By selecting the PDB hit bar of PDB file 4u7p [34], a detailed alignment between the query protein and the sequence in the PDB file (of DNMT3A) is presented. Position 585 is coded by nucleotides which appear to overlap cancer variants in Cosmic, as indicated by the yellow background. In fact, this position is also known to be an HGMD [35] variant, suggesting clinical importance. Several additional residues in the close vicinity are reported in cosmic and ClinVar databases. By clicking on the JSmol link, a 3D session appears, showing the structure (Fig. 6b). The input position 585 is depicted in black and the database variants are depicted in colors according to the G23D scheme. The structure of the threonine mutation residue can be seen in the 3D session by checking the “mut” button in the control panel. The list of contacts, available by a button click from the 3D session (Fig. 6c) reveals clashes with two adjacent residues (marked by arrows). Such clashes suggest significantly decreased stability of the mutant and/or changes in the local backbone conformation. Indeed prediction of I-Mutant2.0, directly accessed from the structural session, suggests a decreased stability of the mutant (Fig. 6d). Interestingly, the variant reported by Wijmenga et al. [36] in the same protein position also includes a larger mutated side chain (Val). Val has a similar shape like Thr, and is also expected to give rise to changes in the backbone conformation. Moreover, many known clinical relevant variants are located in close proximity to position 585. The quantitative contact analysis (Fig. 6c), indeed suggests that position 585 (which appears as position 644 in the ATOM records of PDB 4u7p) has direct contact with other known pathogenic variants such as the residue 622 (681 in the PDB ATOM records). The region which spans position 585 is highly conserved as indicated by the “9” scores in the HSSP conservation profile (Fig. 6e). The sequence profile indicates that this region is part of a known motif of methyltransferases (SAM_MT_C5; Prosite ID: PS51679). There is a consensus among eight different function prediction tools which are included in the dbNSFP database [15] that A585T is a deleterious change, as can be deduced by a link located in the upper part of the G23D hits’ page. DNMT3B variants described in the literature seem to disrupt the function of the enzyme by several distinct mechanisms, including changing protein stability, altering DNA binding affinity, affecting oligomerization with other methyltransferase family members and affecting binding affinity of the SAH methyl donor cofactor [37]. The variants in position 585 appear to belong to the last group. Ala 585 is located close to the SAH but does not directly contact it. Its side chain points to the opposite direction of the cofactor (shown in magenta in Fig. 6a) and is completely buried (Fig. 6c). Upon mutation, backbone changes are inevitable in order to accommodate the larger side chain. These changes are likely to affect the conformation of the nearby loop which directly interacts with the cofactor. Conclusions Structural configuration is the basis for understanding molecular stability of molecules and interactions between molecules. It is also fundamental for understanding the molecular mechanisms which drive certain variants to be pathogenic. Due to the limited availability of structural data and technical difficulties in usage, structural information is often overlooked during functional interpretation of variants. We believe that G23D will help to narrow this gap, and will allow more researchers to include protein structural aspects in their studies of human variants. Abbreviations HTSHigh throughput sequencing PDBProtein data bank WESWhole exome sequencing. We thank the ModBase team for helping us including the theoretical models within G23D. The work of O.S. was done in partial fulfillment with the requirements of the Faculty of Life-Sciences, Bar-Ilan University, Israel. Funding The work was supported by grants from the Varda and Boaz Dotan Research Center in Hemato-oncology affiliated with the CBRC of Tel Aviv, University Flight Attendant Medical Research Institute (FAMRI), Israeli Centers of Excellence (I-CORE) Program (ISF grants no. 41/11 and no. 1796/12). G.R. is a member of the Sagol Neuroscience Network and holds the Djerassi Chair for Oncology at the Sackler Faculty of Medicine, Tel-Aviv University, Israel. Availability of data and material Project name: G23D Online tool for mapping and visualization of genomic variants on 3D protein structures. Project home page: http://www.sheba-cancer.org.il/G23D. Operating systems: Windows, Linux, MacOS Programming languages: Perl, Javascript, C. License: no license is required for browsing in the site. Authors’ contributions EE, NA and GR conceived the methodology and designed the website architecture, OS, VK and EE wrote the code for the website, AJS, AL and RS characterized the clinical and genetic data of the ICF patient. NK and OB performed the exome sequencing analysis, EE wrote the manuscript and the documentation pages. All authors read and approved the final manuscript. Competing interests The authors declare that they have no competing interests. Consent for publication Patient’s parents gave written, informed consent for publication of the sequencing results. Ethics approval and consent to participate The patient mentioned in the case report was studied at the Edmond and Lily Safra Children’s Hospital, Sheba Medical Center, Israel. Patient’s parents gave written, informed consent for this study, which was approved by the local institutional review board (Sheba Medical Center IRB). Data deposition Not applicable to this study. ==== Refs References 1. Niknafs N MuPIT interactive: webserver for mapping variant positions to annotated, interactive 3D structures Hum Genet 2013 132 1235 1243 10.1007/s00439-013-1325-0 23793516 2. Adzhubei IA A method and server for predicting damaging missense mutations Nat Methods 2010 7 248 249 10.1038/nmeth0410-248 20354512 3. Reva B Antipin Y Sander C Predicting the functional impact of protein mutations: application to cancer genomics Nucleic Acids Res 2011 39 e118 10.1093/nar/gkr407 21727090 4. Wang Z Moult J SNPs, protein structure, and disease Hum Mutat 2001 17 263 270 10.1002/humu.22 11295823 5. Mooney SD Altman RB MutDB: annotating human variation with functionally relevant data Bioinformatics 2003 19 1858 1860 10.1093/bioinformatics/btg241 14512363 6. Ryan M Diekhans M Lien S Liu Y Karchin R LS-SNP/PDB: annotated non-synonymous SNPs mapped to Protein Data Bank structures Bioinformatics 2009 25 1431 1432 10.1093/bioinformatics/btp242 19369493 7. Doncheva NT Integrative visual analysis of protein sequence mutations BMC Proc 2014 8 S2 10.1186/1753-6561-8-S2-S2 25237389 8. Mueller SC BALL-SNP: combining genetic and structural information to identify candidate non-synonymous single nucleotide polymorphisms Genome Med 2015 7 65 10.1186/s13073-015-0190-y 26191084 9. Witvliet D, et al. ELASPIC web-server: proteome-wide structure based prediction of mutation effects on protein stability and binding affinity. Bioinformatics. 2016;32:1589–1591. 10. De Baets G SNPeffect 4.0: on-line prediction of molecular and structural effects of protein-coding variants Nucleic Acids Res 2012 40 D935 D939 10.1093/nar/gkr996 22075996 11. Betts MJ Mechismo: predicting the mechanistic impact of mutations and modifications on molecular interactions Nucleic Acids Res 2015 43 e10 10.1093/nar/gku1094 25392414 12. Mosca R dSysMap: exploring the edgetic role of disease mutations Nat Methods 2015 12 167 168 10.1038/nmeth.3289 25719824 13. Berman HM The Protein Data Bank Nucleic Acids Res 2000 28 235 242 10.1093/nar/28.1.235 10592235 14. Pieper U ModBase, a database of annotated comparative protein structure models and associated resources Nucleic Acids Res 2014 42 D336 D346 10.1093/nar/gkt1144 24271400 15. Liu X Wu C Li C Boerwinkle E dbNSFP v3.0: A One-Stop Database of Functional Predictions and Annotations for Human Nonsynonymous and Splice-Site SNVs Hum Mutat 2016 37 235 241 10.1002/humu.22932 26555599 16. Altschul SF Gapped BLAST and PSI-BLAST: a new generation of protein database search programs Nucleic Acids Res 1997 25 3389 3402 10.1093/nar/25.17.3389 9254694 17. Jmol: an open-source HTML5 viewer for chemical structures in 3D. http://jmol.sourceforge.net. 18. Consortium U UniProt: a hub for protein information Nucleic Acids Res 2015 43 D204 D212 10.1093/nar/gku989 25348405 19. Forbes SA COSMIC: exploring the world’s knowledge of somatic mutations in human cancer Nucleic Acids Res 2015 43 D805 D811 10.1093/nar/gku1075 25355519 20. Landrum MJ ClinVar: public archive of interpretations of clinically relevant variants Nucleic Acids Res 2016 44 D862 D868 10.1093/nar/gkv1222 26582918 21. Touw WG, et al. A series of PDB-related databanks for everyday needs. Nucleic Acids Res. 2015;43:D364–8. 22. Sigrist CJA New and continuing developments at PROSITE Nucleic Acids Res 2013 41 D344 D347 10.1093/nar/gks1067 23161676 23. Sickmeier M DisProt: the Database of Disordered Proteins Nucleic Acids Res 2007 35 D786 D793 10.1093/nar/gkl893 17145717 24. Sherry ST dbSNP: the NCBI database of genetic variation Nucleic Acids Res 2001 29 308 311 10.1093/nar/29.1.308 11125122 25. Furnham N The Catalytic Site Atlas 2.0: cataloging catalytic sites and residues identified in enzymes Nucleic Acids Res 2014 42 D485 D489 10.1093/nar/gkt1243 24319146 26. Eyal E Najmanovich R McConkey BJ Edelman M Sobolev V Importance of solvent accessibility and contact surfaces in modeling side-chain conformations in proteins J Comput Chem 2004 25 712 724 10.1002/jcc.10420 14978714 27. Krivov GG Shapovalov MV Dunbrack RL Jr Improved prediction of protein side-chain conformations with SCWRL4 Proteins 2009 77 778 795 10.1002/prot.22488 19603484 28. McConkey BJ Sobolev V Edelman M Quantification of protein surfaces, volumes and atom-atom contacts using a constrained Voronoi procedure Bioinformatics 2002 18 1365 1373 10.1093/bioinformatics/18.10.1365 12376381 29. McConkey BJ Sobolev V Edelman M Discrimination of native protein structures using atom-atom contact scoring Proc Natl Acad Sci U S A 2003 100 3215 3220 10.1073/pnas.0535768100 12631702 30. Dong C Comparison and integration of deleteriousness prediction methods for nonsynonymous SNVs in whole exome sequencing studies Hum Mol Genet 2015 24 2125 2137 10.1093/hmg/ddu733 25552646 31. Guerois R Nielsen JE Serrano L Predicting changes in the stability of proteins and protein complexes: a study of more than 1000 mutations J Mol Biol 2002 320 369 387 10.1016/S0022-2836(02)00442-4 12079393 32. Capriotti E Fariselli P Casadio R I-Mutant2.0: predicting stability changes upon mutation from the protein sequence or structure Nucleic Acids Res 2005 33 W306 W310 10.1093/nar/gki375 15980478 33. Hamidi T Singh AK Chen T Genetic alterations of DNA methylation machinery in human diseases Epigenomics 2015 7 247 265 10.2217/epi.14.80 25942534 34. Guo X, et al. Structural insight into autoinhibition and histone H3-induced activation of DNMT3A. Nature. 2015;517:640–4. 35. Stenson PD Human Gene Mutation Database (HGMD): 2003 update Hum Mutat 2003 21 577 581 10.1002/humu.10212 12754702 36. Wijmenga C Genetic variation in ICF syndrome: evidence for genetic heterogeneity Hum Mutat 2000 16 509 517 10.1002/1098-1004(200012)16:6<509::AID-HUMU8>3.0.CO;2-V 11102980 37. Cheng X Blumenthal RM Mammalian DNA methyltransferases: a structural perspective Structure 2008 16 341 350 10.1016/j.str.2008.01.004 18334209

PMC005xxxxxx/PMC5002100.txt

==== Front BMC GenetBMC GenetBMC Genetics1471-2156BioMed Central London 43010.1186/s12863-016-0430-1Research ArticleDetection of genomic signatures of recent selection in commercial broiler chickens Fu Weixuan wfu@udel.edu 1Lee William R billlee@mapleleaffarms.com 2Abasht Behnam abasht@udel.edu 11 Department of Animal and Food Sciences, University of Delaware, Newark, DE 19716 USA 2 Maple Leaf Farms, Inc, Leesburg, IN 46538 USA 26 8 2016 26 8 2016 2016 17 1 12210 3 2016 22 8 2016 © The Author(s). 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Background Identification of the genomic signatures of recent selection may help uncover causal polymorphisms controlling traits relevant to recent decades of selective breeding in livestock. In this study, we aimed at detecting signatures of recent selection in commercial broiler chickens using genotype information from single nucleotide polymorphisms (SNPs). A total of 565 chickens from five commercial purebred lines, including three broiler sire (male) lines and two broiler dam (female) lines, were genotyped using the 60K SNP Illumina iSelect chicken array. To detect genomic signatures of recent selection, we applied two methods based on population comparison, cross-population extended haplotype homozygosity (XP-EHH) and cross-population composite likelihood ratio (XP-CLR), and further analyzed the results to find genomic regions under recent selection in multiple purebred lines. Results A total of 321 candidate selection regions spanning approximately 1.45 % of the chicken genome in each line were detected by consensus of results of both XP-EHH and XP-CLR methods. To minimize false discovery due to genetic drift, only 42 of the candidate selection regions that were shared by 2 or more purebred lines were considered as high-confidence selection regions in the study. Of these 42 regions, 20 were 50 kb or less while 4 regions were larger than 0.5 Mb. In total, 91 genes could be found in the 42 regions, among which 19 regions contained only 1 or 2 genes, and 9 regions were located at gene deserts. Conclusions Our results provide a genome-wide scan of recent selection signatures in five purebred lines of commercial broiler chickens. We found several candidate genes for recent selection in multiple lines, such as SOX6 (Sex Determining Region Y-Box 6) and cTR (Thyroid hormone receptor beta). These genes may have been under recent selection due to their essential roles in growth, development and reproduction in chickens. Furthermore, our results suggest that in some candidate regions, the same or opposite alleles have been under recent selection in multiple lines. Most of the candidate genes in the selection regions are novel, and as such they should be of great interest for future research into the genetic architecture of traits relevant to modern broiler breeding. Electronic supplementary material The online version of this article (doi:10.1186/s12863-016-0430-1) contains supplementary material, which is available to authorized users. Keyword ChickensSelection signaturesBreedingCommercial broilershttp://dx.doi.org/10.13039/100006094University of Delawareissue-copyright-statement© The Author(s) 2015 ==== Body Background Artificial selection is the primary factor in the domestication and breeding history of livestock species. Modern broiler (meat-type) chickens have been under strong artificial selection, mostly for traits of economic importance for farmers, such as growth rate, feed efficiency and body composition [1]. Comparing a modern broiler chicken cross, Ross 308, with a broiler population that had not been subjected to artificial selection since 1957 [i.e., Athens-Canadian Random-bred Control (ACRBC) strain], Havenstein et al. (2003) found that the average body weight at 42 days of age increased from 539 g in 1957 as represented by the ACRBC strain to 2,672 g in 2001 as represented by the Ross 308 strain and that the feed conversion ratio decreased from 2.34 to 1.43 over the same time period [2]. The authors indicated that genetic selection contributed 85–90 % of the improvement in growth rate over the past 45 years. These dramatic phenotypic changes imply that the frequencies of the underlying causal polymorphisms themselves have been altered by selection for performance in these traits during the intervening time period. Thus, detecting the genomic footprints of artificial selection should help researchers to identify the causal polymorphisms underlying phenotypic changes and to better understand the biological and genetic mechanism controlling these traits. In broiler chicken genetic stocks, the traits of most relevance for recent decades of breeding, such as feed efficiency, growth rate and meat yield, are complex traits that are controlled by many genes. Consequently, it is highly likely that selection for these traits has worked simultaneously on multiple causal genes across the genome. Therefore, high throughput methods are required to screen the whole genome for signatures of recent selection. With the availability of high throughput genotyping tools, such as high-density SNP arrays and next-generation sequencing, it has become possible to conduct genome-wide studies for detection of genomic footprints of artificial selection. Using whole-genome re-sequencing and high-density SNP chips, respectively, Rubin et al. (2010) and Elferink et al. (2012) have investigated selection signatures in large numbers of chicken breeds using Z-transformed pooled heterozygosity (ZHp) scores. This statistic estimates local heterozygosity depression in chromosomal regions [3, 4] and has been appropriately applied for detecting alleles that have swept to fixation or near-fixation by long-term directional selection or during domestication [5]. However, modern broiler chicken breeding practices that have a more recent selection history, and have been employed to select for a suite of more sophisticated traits, such as feed efficiency and meat yield with different selection priorities in different specialized component lines, would not be expected to leave such common and distinct changes. Therefore, most signatures of more recent selection are likely yet to be uncovered in the genome of modern broiler chickens. In contrast with the ZHp method, it has been suggested that methods based on extended haplotype homozygosity (EHH) [6] or change in allele frequency spectrum can be more useful for detecting signatures of recent selection in animal breeds [5, 7]. These population genetics methods are developed to find frequent alleles and long-range haplotypes with high frequency, which are indicatives of chromosomal regions under recent selection [6]. In dairy cattle, Qanbari et al. (2010) adopted the relative EHH method to detect signatures of positive selection in Holstein–Friesian cattle using a 50K SNP array [7]. Zhang et al. (2012) applied the same method to detect selection signatures in two broiler chicken lines divergently selected for abdominal fat content and reported the PC1/PCSK1 region as the most likely candidate region to have a causal effect on abdominal fat weight [8, 9]. To detect selection signatures in Fleckvieh cattle, Qanbari et al. (2014) applied two statistical methods, the integrated haplotype score (iHS) [10] and the composite likelihood ratio (CLR) [11, 12], and found that many candidate regions were relevant to coat coloring pattern, neurobehavioral functioning and sensory perception [13]. Concerned that the EHH and iHS methods may have insufficient power to identify the alleles that have been more recently strongly selected and swept to near-fixation or fixation [14], Sabeti et al. (2007) developed the cross-population extended haplotype homozygosity (XP-EHH) test to detect signatures of recent selection by comparing EHH between two different populations regardless of whether or not the favored allele had reached fixation. Also, the single-population CLR method does not take advantage of larger differences in allele frequencies between two breeds and is very sensitive to SNP ascertainment bias. To overcome these limitations, Chen et al. (2010) developed the cross-population composite likelihood ratio (XP-CLR) test [15]. In this study, we applied both XP-EHH and XP-CLR methods to five commercial broiler purebred populations, including three male lines and two female lines, to detect the signatures of recent selection in these commercial broiler stocks. The findings here help to improve our understanding of the biological mechanisms controlling economically important traits in modern commercial broiler chickens. Methods Animals and data preparation A total of 565 chickens from five commercial purebred lines were genotyped using the 60K SNP Illumina iSelect chicken array [16]. Blood samples were collected from a wing vein and the samples were shipped on dry ice to DNA LandMarks (Saint-Jean-sur-Richelieu, Quebec, Canada) for DNA extraction and genotyping with the 60K SNP Illumina iSelect chicken array. All genotyped birds were males and were sampled from 3 male (broiler sire) lines, ML1, ML2 and ML3, and 2 female (broiler dam) lines, FL1 and FL2. In total, 318 birds were sampled from male lines: 24 ML1, 256 ML2 and 38 ML3 chickens; and 247 birds were sampled from female lines: 126 FL1 and 121 FL2 chickens. The FL1, FL2 and ML1 chickens as well as a portion of ML2 (ML2_0; n = 96) chickens were elite sires randomly sampled from three overlapping generations. Another portion of ML2 genotyped chickens (ML2_1; n = 160) was a random sample of the progeny of the ML2_0 elite sires. The ML3 genotyped chickens were random samples of male chickens from this purebred population. Male and female lines originated from different breeds, i.e. the male line from Cornish, a meat type breed, and the female lines from White Rock, a dual-purpose breed. Each of these five lines came from a different source to Heritage Breeders, and all lines, except FL1, have been reproductively isolated for more than 40 generations. A one-time crossbreeding with ML2 and then backcrossing with FL1 happened early in the history of FL1, and the resulting new FL1 population has been reproductively isolated for more than 25 generations. In each generation within each purebred line, approximately 50 to 80 male and 500 to 800 female birds have been selected for reproducing the next generation. More details about genetic diversity in ML2, FL1 and FL2 can be found in a previous study, where the three lines were labeled as B, C and D, respectively [17]. Since allele frequency of SNPs and linkage disequilibrium (LD) among SNPs were highly consistent between the two sampled generations of ML2, ML2_0 and ML2_1 (unpublished data in our laboratory) and the methods we used for detecting recent selection signatures relied on allele frequency and LD information, we combined data from ML2_0 and ML2_1 in the current study. Although male lines (ML1, ML2 and ML3) have shared similar selection objectives, which have been primarily focused on increasing growth rate, feed efficiency and breast muscle yield, the relative magnitude of selection pressure on these major traits varied among the three lines: ML1 has been more heavily selected for rapid growth, ML2 for high breast meat yield and ML3 for improved feed efficiency. Therefore, we expect artificial selection has unequally increased the frequency of alleles controlling these traits among the male lines, and some alleles may have been selected in opposite direction between male and female lines. The expected differences in the allele frequency among these purebred lines make it possible to apply the two population comparison based methods, XP-EHH and XP-CLR, in our study. The 60K SNP Illumina iSelect chicken array contains a total of 57,636 SNPs [16]. For the purpose of this study, we used only SNPs with assigned positions on the current chicken genome based on the latest reference genome (Gallus gallus 4.0 UCSC, May 2012). We excluded SNPs with a call rate < 90 % or Mendelian inconsistency > 0.001 and SNPs that were monomorphic among all the purebred lines. We also excluded SNPs on chromosomes 16 and W and two linkage groups, as there were too few SNPs in the 60K SNP Illumina iSelect chicken array for these chromosomes. After quality control, 48,950 SNPs were used in subsequent analyses of the five populations (Table 1).Table 1 Quality control of genotype data Total SNPs Quality control SNPs used NI1 MI2 UG3 MO4 LCR5 57,636 1,507 1,478 873 4,292 536 48,950 Note: 1SNPs on GGA16, W and two linkage groups (LGE22C19W28_E50C23 and LEG64) or SNPs with unknown positions on Galgal4; 2SNPs with Mendelian inconsistency; 3Ungenotyped SNPs; 4SNPs monomorphic among all the 5 purebred lines; 5SNPs with low call rate Since a linkage map was required for the XP-CLR method, we calculated the genetic positions of all the markers in the 60K SNP Illumina iSelect chicken array using a subset of markers with known genetic positions based on the male linkage map previously provided by Groenen et al. [18]), and assuming that the recombination rates between two markers were uniformly distributed. We used BEAGLE (Version 3.3.2) [19] to impute missing genotypes, phase the chromosomes and identify haplotype structure at the candidate selection regions in each purebred line. The XP-EHH test The XP-EHH test uses the integrated EHH (iHH) of a core SNP in two populations, A and B, rather than two alleles in a single population. The unstandardized XP-EHH statistic can be calculated as [14]: 1 unstandardizedXP−EHH=lniHHAiHHB where iHHA and iHHB are the integrated EHH of a given core SNP in population A and B, respectively. A large positive value of XP-EHH suggests either selection in population A or a negative value in population B. We used the software developed by Pickrell et al. [20] to estimate unstandardized XP-EHH statistics for all SNPs (after quality control) in all five purebred lines with cross-population comparison of each purebred line with the four remaining lines: for example, ML1 vs. ML2, ML3, FL1 or FL2 (four cross-population tests for each line). The unstandardized XP-EHH statistics were standardized using their means and variances in each purebred comparison. Because previous studies found that the standardized XP-EHH statistics follows the standard normal distribution [14, 21, 22], P-values of SNPs were estimated using the standard normal distribution. For each purebred comparison, we determined the candidate regions under positive selection by clustering the significant core SNPs (P-value < 0.05) with a distance of less than 200 kb. The XP-CLR test To confirm selection signatures detected by the XP-EHH analysis, we applied the XP-CLR test based on the change in the allele frequency spectrum, since it has the advantage of enlarging signals to allow the resolution of more precise regions [15]. The XP-CLR test [15] was also adopted for the five purebred lines by cross-population comparison of each line with the four remaining lines as reference populations using the XP-CLR 1.0 software available at http://genetics.med.harvard.edu/reich/Reich_Lab/Software.html (last accessed Jan. 3, 2016). The grid points at the putative selected allele positions were set along each chicken chromosome with a spacing of 2 kb, and sliding window size was set as 0.5 cM around the grid points. To reduce the contribution of SNPs in high LD to the likelihood function, the cut-off level of absolute pairwise correlation coefficient of two SNPs was set to 0.9 for estimation of the weight factor (w [15]). For each cross-population comparison, the cutoff threshold of 0.5 % XP-CLR scores was applied to determine windows with strong signals across the whole genome. We then determined the candidate selection regions by clustering these windows, such that windows with genetic distances less than 1 cM constitute a candidate selection region. The selection regions detected by both statistical methods for each purebred line were determined as candidate regions under positive selection. A Karyogram layout of candidate selection regions detected by both tests was created using the ggbio R package [23]. In comparison with human populations, modern livestock breeds generally have much smaller effective population sizes due to animal breeding programs [24–26]. To minimize false discovery due to genetic drift resulted from the small effective population size, the candidate selection regions shared by multiple purebred lines were consider as high-confidence selection regions, and these regions were chosen to identify candidate genes using the genomic database search engine BioMart (http://www.biomart.org/). Results In total, 1,079 putative selection regions were detected with P-values < 0.05 using XP-EHH test (Additional file 1: Table S1), and 1,018 putative selection regions were detected using the criterion of a 0.5 % cutoff of XP-CLR scores (Additional file 2: Table S2). Regions detected using XP-EHH overlapped 31.53 % of the regions that were identified using XP-CLR (Additional file 3: Table S3). Even though 328 overlapped regions (i.e., detected by both methods) were presented on Additional file 3: Table S3, some regions detected by either the XP-EHH or XP-CLR tests were wide enough to overlap with more than one region detected by the other test. Therefore, in total, 224 and 321 unique regions were detected using XP-EHH and XP-CLR tests, respectively. In each line, approximately 11.09 % of the chicken genome was covered by regions detected by XP-EHH methods, while approximately 2.58 % of the chicken genome was covered by regions detected by XP-CLR methods. The overlapped regions (shared by both methods) only represented approximately 1.45 % of the chicken genome in each line. Selection regions detected by XP-EHH were much wider, mainly because the EHH test is an LD-based method, and LD is expected to extend over longer distances in regions under recent selection [17]. For example, Fig. 2a and 2b represent the results of XP-EHH and XP-CLR tests on GGA5, which show candidate regions detected by XP-CLR tests were overall narrower and perhaps more accurate than those detected by the XP-EHH tests. Thus, to narrow down regions that overlapped between the two methods, we considered the 321 regions based on the XP-CLR test as the candidate selection regions. Their ranges are presented in Additional file 3: Table S3 and visualized in Fig. 1.Fig. 1 Candidate selection regions detected by XP-EHH and XP-CLR tests. For each purebred line, the overlapped regions detected by the XP-EHH and XP-CLR tests were presented based on the ranges from XP-CLR test. Each population is denoted by a different color By further examining these 321 regions, we identified 42 regions that were shared by two or more purebred lines (Additional file 4: Table S4) and considered them as high-confidence selection regions. Figure 2c and 2d represent the results of XP-EHH and XP-CLR tests in one of the 42 regions (GGA5: 31.06-31.82 Mb) shared by ML1 and ML3. To further narrow down these high-confidence selection regions, only common regions shared by two or more purebred lines were counted in the overlapped regions. Of these 42 common regions, 20 were 50 kb or less while 4 regions were larger than 0.5 Mb. Using BioMart, 91 genes could be found in the 42 regions (Additional file 4: Table S4) among which 9 regions were located at gene deserts and 19 regions only harbored 1 or 2 genes. For the 9 regions located at gene deserts, the genes closest to them (±100 kb) are listed on Additional file 4: Table S4.Fig. 2 XP-EHH and XP-CLR scores on GGA5. A and B: The results of the XP-EHH (a) and XP-CLR (b) statistics on the whole GGA5 using multiple population comparisons. c and d: The results of the XP-EHH (a) and XP-CLR (b) statistics in a candidate selection region (GGA5: 31.06-31.82 Mb) shared by ML1 and ML3. The dots in the Fig. 2b and 2d represent the XP-CLR scores of sliding windows, and the dots in the Fig. 2a and 2c represent the standardized XP-EHH scores of SNPs. Each comparison of ML1 or ML3 against the other 3 lines is denoted by a different color To gain insight into population differences in the overlapped candidate regions, we constructed haplotypes and estimated haplotype frequencies in these regions in each population (Additional file 5: Table S5). This analysis was performed only for the the high-confidence selection regions that contained at least 5 informative SNPs in our genotype data (14 out of 42 regions). Figure 3 represents the results of haplotype analysis in four selection regions containing 10 to 20 SNPs in our genotype data. As demonstrated in Fig. 3, haplotypes with high frequencies were detected in the denoted genomic regions. For example, in a selection region on GGA4 (52.15-52.47 Mb), the same haplotype showed high frequency in FL1 and FL2, although the range of this region was more than 300 kb. Another interesting example is a ~240 kb region on GGAZ (45.49-45.73 Mb). In this region, all three male lines had the same major haplotype, but each female line had a different major haplotype (Additional file 5: Table S5).Fig. 3 Haplotype frequencies of SNPs in four selection regions detected in multiple purebred lines of broiler chickens. The various colors excluding grey refer to the major haplotypes identified in the 5 purebred lines. At each denoted region, a selection signature was detected in the purebred lines marked with “*” Discussion In modern broiler breeding, the practice of selective mating is utilized to influence the expression of economically important traits in subsequent generations. Through such selection, the “beneficial” alleles tend to become more frequent in populations over time. In our study, we applied XP-EHH and XP-CLR tests to detect the genomic regions under recent selection by measuring the characteristics of extended haplotype homozygosity and changes in the allele frequency spectrum. By cross-population comparisons of five commercial broiler purebred lines, we identified the genomic regions that are most likely to harbor genes related to traits of economic importance in broiler chickens. It should be mentioned that both bottleneck events and genetic drift have potential to influence the results of selection signature studies such as this one. Based on records from Heritage Breeders, bottleneck events did not occur in the lines used in the present study for more than 40 generations. To minimize false discovery due to genetic drift, only 42 of the candidate selection regions that were shared by 2 or more purebred lines were considered as high-confidence selection regions in our study. Another potential limitation is inherent to cross-population methods, which may fail to detect a selection signature where the desirable allele has been under similar level of positive or negative selection pressure in all these purebred lines that were studied. However, this limitation should not be a major concern because the relative magnitude of selection pressure on major traits, growth rate, feed efficiency and breast muscle yield, varied among the 5 purebred lines. Also, unlike in male lines, selection for reproduction traits has been emphasized in female lines. Candidate selection regions We compared the genes in the candidate selection regions in our study with those from two previous studies on detecting selection sweeps in chickens. Among 91 genes in the 42 regions in our study, only two genes (SOX6 and GJD2) are in genomic regions detected by Rubin et al. (2010) in commercial broilers. Also, only two genes in our list (GAS7 and STXBP6; Additional file 4: Table S4) are among 366 genes (based on Ensembl gene ID) detected by Elferink et al. (2012). This low extent of overlap with previous studies is likely related to the different methods that we used for detecting selection signatures in the present study. As mentioned before, we aimed at detecting signatures of recent selection using the cross-population methods, whereas the ZHp method used in two previous studies is primarily focused on detecting older selection signatures such as those accumulated during domestication. For better comparison, we estimated ZH scores over sliding 5-marker windows on autosomes using data from our study (Additional file 6: Supplemental file). This analysis led to the detection of 41 selection regions containing 81 genes, including 31 genes detected in broilers from two previous studies (Additional file 7: Table S7 and Additional file 8: Table S8) although our resource populations were much different from those in the two previous studies. The most possible reason of high overlap using ZH scores is that some older selection signatures during chicken domestication were shared with commercial broilers used in our study as well as two previous studies. In our results from two cross-population methods (XP-EHH and XP-CLR), 42 regions were detected by both methods in multiple populations, which might indicate that gene(s) in these regions have been independently selected in multiple populations, i.e., parallel selection. Of these 42 regions, 13 regions were shared among male lines, 5 regions were shared among female lines and 24 regions were shared among both male and female lines. In addition to the significant overlap between the suites of selected traits among all lines selected for broiler performance (especially for growth rate, meat yield, and feed efficiency), the shared selection regions among male and female lines suggests that alleles with pleiotropic effects have been under recent selection in these regions, i.e., alleles that are positively correlated with growth related traits and negatively with reproduction traits, or vice versa. Alternatively, the regions shared by male and female lines may contain closely linked genes impacting both growth and reproduction traits. To further analyze the high-confidence selection regions, we examined population differences in haplotype structure and frequencies and identified the major haplotype (the haplotype with the highest frequency) within each population at each denoted region. The results showed that, in some candidate selection regions, the same major haplotype is shared by multiple lines (Fig. 3 and Additional file 5: Table S5). However, in 9 out of 14 candidate selection regions presented in Additional file 5: Table S5, such as GGA1: 54.92-55.22 Mb and GGA13: 9.38-9.44 Mb, the major haplotype varied greatly among the five purebred lines. The difference in major haplotype may represent high diversity of genetic background among these purebred lines [17]. Alternatively, it is possible that selection has acted on different alleles of a gene in these purebred lines. For example, previous studies found that fertility was reduced in chickens under strong selection for body weight due to the negative genetic correlation between reproduction and growth traits [27–29]. Overall, selection for reproduction traits has been more emphasized in female lines, whereas selection for high feed-efficiency and increased skeletal muscle growth has been the major focus in male lines. Thus, the frequency of alleles benefiting reproduction traits but adversely affecting growth traits are expected to be relatively higher in female lines as compared with the male lines. Some candidate genes with potential pleiotropic effects, such as STXBP6 and cTR, in the high-confidence selection regions are discussed below. Candidate genes in regions detected in mulitple populations In the 42 high-confidence candidate selection regions detected by both methods (XP-EHH and XP-CLR) in two or more purebred lines, we identified several genes related to growth, development, feed efficiency and reproduction in chickens (Table 2). Only a few of them are mentioned below to discuss their potential involvements in controlling these traits.Table 2 A partial list of candidate genes in or near the 42 high-confidence selection regions detected in multiple purebred lines of broiler chickens Gene name Gene symbol Function or association Thyroid hormone receptor beta cTR Growth, development and homeostasis Sex Determining Region Y-Box 6 SOX6 Development of chondrocytes and skeletal muscle Actin, alpha, cardiac muscle 1 ACTC1 Muscle development Syntaxin binding protein 6 STXBP6 Bone allocation and fecundity traits Myosin heavy chain 13 MYH13 Skeletal muscle development Calpastatin CAST* Growth and meat quality Note: *this gene is located close to a candidate selection region detected in a gene desert area on GGAZ. Information about chromosomal locations of the 42 candidate selection regions and the full list of candidate genes in or near these regions can be found in Additional file 4: Table S4 Myosin heavy chain 13 (MYH13) MYH13 is located in a candidate selection region on GGA18 (0.22-0.40 Mb), and other four genes of the myosin heavy chain (MyHC) family (MYH1A, MYH1B, MYH1C, MYH1E) are located very close to this region. Previous studies found that MyHC genes play important roles in skeletal muscle development [30–32], and the polymorphisms in MYH3 were significantly associated with growth and body composition traits in Qinchuan cattle [33, 34]. Sex determining region Y-Box 6 (SOX6) SOX6 is located in a candidate selection region (GGA5: 10.65-11.09 Mb) detected in two male lines, ML2 and ML3. This gene encodes a Sry-related transcription factor that promotes early chondroblast differentiation and plays a critical role in differentiation and proliferation of chondrocytes as well as normal fiber type differentiation of fetal skeletal muscle in mice [35–37]. Proprotein convertase subtilisin/kexin type 1 (PCSK1) and calpastatin (CAST) Although no gene was found inside a candidate selection region around 56.76 Mb on GGAZ due to its small size (8 kb), this region is consistent with findings from two previous studies [8, 9] in which a selection signature was detected using two chicken lines divergently selected for abdominal fat content for 11 generations. Of note, a previously known candidate gene for fatness in chickens, PCSK1 [8], is located close to this region. Another gene close to this region is CAST, which encodes calpastatin, a specific inhibitor of an endogenous calpain. The calpain family plays an important role in embryonic development and muscle growth [38–40]. Many studies have found that polymorphisms in CAST are significantly associated with growth traits and meat quality traits in livestock animals [41–46]. Actin, alpha, cardiac muscle 1(ACTC1) and Syntaxin binding protein 6 (STXBP6) Another muscle-related gene, ACTC1, was found in one of the selection regions on GGA5 (31.06-31.82 Mb, Fig. 3). This gene encodes cardiac muscle alpha actin in chickens and plays an important role in fetal development as well as cell survival, differentiation and development of muscle [47–50]. STXBP6 is another gene in this candidate selection region on GGA5. A previous study has indicated STXBP6 had potential pleiotropic effect on bone tissue and fecundity traits in chickens [51]. Interestingly, this selection sweep, which was detected in two male lines (ML1 and ML3), was also found in a previous study in table egg layer breeds of chickens [52]. One possible reason why this selection sweep is shared by broiler (meat-type) and layer (egg-type) chickens is that both genes, ACTC1 and STXBP6, may influence body weight in broilers and in layers. It should be mentioned that breeders improved meat production in broilers by selection on high body weight at an early age (<8 weeks of age) while they improved feed efficiency and egg production in layers by selection on low body weight at a late age (>24 weeks of age) [53–57]. Alternatively, this shared selection sweep may be explained by the pleiotropic effect of STXBP6 on both bone tissue and fecundity traits. Thyroid hormone receptor beta (cTR) cTR was found in a selection region on GGA2 (37.90-38.07 Mb). Thyroid hormone can regulate animal growth, development and homeostasis [58], and its receptor mediates thyroid hormone actions [59]. Mice with homozygous mutant cTR gene manifest low weight gain and decreased bone development compared to normal mice [60]. In a ~40 kb-length candidate selection region (GGA2: 38.03-38.07 Mb) which was detected in 4 purebred lines (FL1, FL2, ML2 and ML3), there are 3 SNPs in our dataset, which construct the same major haplotype (AAA) in two female lines and ML2, but the major haplotype (GGG) in ML3 is completely different. It should be mentioned that among the five purebred lines of chickens used in our study, ML3 is the most feed-efficient line. It is possible that in this candidate region, the same allele of cTR has been selected in two female lines and ML2 while an alternative allele has been selected in ML3. This assumption may be further supported considering diverse functions of thyroid hormone: it has been reported that thyroid hormone also plays a critical role in fertility, but excessive amounts of this hormone in hyperthyroidism has a negative effect on reproduction in humans [61, 62]. Therefore, the pleiotropic effects of thyroid hormone on reproduction and growth traits may explain why the receptor gene, cTR, may have been under selection among both female and male lines. Conclusions In this study, we identified novel candidate regions for recent selection in broiler chickens. Based on the biological function of genes in the candidate regions, several genes, such as SOX6 and cTR, have possibly made large contributions to economically important traits in chickens. Our findings suggest that recent selection in broiler breeding has had large impact on frequency of genes controlling economically important traits, such as weight gain, muscle mass, feed efficiency and reproduction. Finally, since most of the candidate genes identified in the present study are novel and have probably been under recent selection, they should be of great interest for future research into the genetic architecture of traits relevant to modern broiler breeding. Additional files Additional file 1: Table S1. Putative selection regions detected by XP-EHH test (XLSX 99 kb) Additional file 2: Table S2. Putative selection regions detected by XP-CLR test (XLSX 92 kb) Additional file 3: Table S3. Candidate selection regions detected by both XP-EHH and XP-CLR tests (XLSX 73 kb) Additional file 4: Table S4. High-confidence selection regions detected by both methods (XP-EHH and XP-CLR) in 2 or more populations (XLSX 41 kb) Additional file 5: Table S5. Major haplotypes and their frequencies detected in the purebred lines at the high-confidence selection regions that span at least 5 SNPs in our genotype data (XLSX 12 kb) Additional file 6: Methods and results of detecting selection signatures using ZH scores. (DOCX 32 kb) Additional file 7: Table S7. Candidate selection regions on autosomes detected using ZH scores (XLSX 33 kb) Additional file 8: Table S8. Genes in regions detected using ZH scores that overlap with previous studies (XLSX 34 kb) Abbreviations CLRComposite likelihood ratio EHHExtended haplotype homozygosity iHHIntegrated extended haplotype homozygosity iHSIntegrated haplotype score LDLinkage disequilibrium MAFMinor allele frequency QTLQuantitative trait loci XP-CLRCross-population composite likelihood ratio XP-EHHCross-population extended haplotype homozygosity We would like to thank Heritage Breeders for providing SNP genotype data; and USDA Chicken GWMAS Consortium, Cobb Vantress, and Hendrix Genetics for access to the developed 60K SNP Illumina iSelect chicken array. Funding The study was financially supported by the University of Delaware. Availability of data and materials The datasets supporting the conclusions of this article are included within the article and its additional files. Authors’ contributions WF and BA contributed to the preparation of the manuscript and to the scientific discussions. WF contributed to the genetic analyses of the study, BA conceived of the study, and WL and BA participated in its management, overall design and coordination. WL participated in revising the manuscript. All authors read and approved the final manuscript. Competing interests The authors declare that they have no competing interests. Consent for publication Not applicable. Ethics approval and consent to participate The University of Delaware Agricultural Animal Care and Use Committee approved the animal protocol used for this scientific study. ==== Refs References 1. Crawford RD. Poultry Breeding and Genetics. New York: Elsevier Science Publishing Company Inc.; 1990 2. Havenstein GB Ferket PR Qureshi MA Growth, livability, and feed conversion of 1957 versus 2001 broilers when fed representative 1957 and 2001 broiler diets Poult Sci 2003 82 1500 1508 10.1093/ps/82.10.1500 14601725 3. Rubin C-J Zody MC Eriksson J Meadows JRS Sherwood E Webster MT Jiang L Ingman M Sharpe T Ka S Hallböök F Besnier F Carlborg O Bed’hom B Tixier-Boichard M Jensen P Siegel P Lindblad-Toh K Andersson L Whole-genome resequencing reveals loci under selection during chicken domestication Nature 2010 464 587 591 10.1038/nature08832 20220755 4. Elferink MG Megens H-J Vereijken A Hu X Crooijmans RPMA Groenen M a M Signatures of selection in the genomes of commercial and non-commercial chicken breeds PLoS One 2012 7 10.1371/journal.pone.0032720 22384281 5. Utsunomiya YT Pérez O’Brien AM Sonstegard TS Van Tassell CP do Carmo AS Mészáros G Sölkner J Garcia JF Detecting loci under recent positive selection in dairy and beef cattle by combining different genome-wide scan methods PLoS One 2013 8 10.1371/journal.pone.0064280 23696874 6. Sabeti PC Reich DE Higgins JM Levine HZP Richter DJ Schaffner SF Gabriel SB Platko JV Patterson NJ McDonald GJ Ackerman HC Campbell SJ Altshuler D Cooper R Kwiatkowski D Ward R Lander ES Detecting recent positive selection in the human genome from haplotype structure Nature 2002 419 832 837 10.1038/nature01140 12397357 7. Qanbari S Pimentel ECG Tetens J Thaller G Lichtner P Sharifi a R Simianer H A genome-wide scan for signatures of recent selection in Holstein cattle Anim Genet 2010 41 377 389 20096028 8. Zhang H Hu X Wang Z Zhang Y Wang S Wang N Ma L Leng L Wang S Wang Q Wang Y Tang Z Li N Da Y Li H Selection signature analysis implicates the PC1/PCSK1 region for chicken abdominal fat content PLoS One 2012 7 10.1371/journal.pone.0040736 22792402 9. Zhang H Wang S-Z Wang Z-P Da Y Wang N Hu X-X Zhang Y-D Wang Y-X Leng L Tang Z-Q Li H A genome-wide scan of selective sweeps in two broiler chicken lines divergently selected for abdominal fat content BMC Genomics 2012 13 704 10.1186/1471-2164-13-704 23241142 10. Voight BF Kudaravalli S Wen X Pritchard JK A map of recent positive selection in the human genome PLoS Biol 2006 4 10.1371/journal.pbio.0040072 16494531 11. Nielsen R Williamson S Kim Y Hubisz MJ Clark AG Bustamante C Genomic scans for selective sweeps using SNP data Genome Res 2005 15 1566 1575 10.1101/gr.4252305 16251466 12. Williamson SH Hubisz MJ Clark AG Payseur BA Bustamante CD Nielsen R Localizing recent adaptive evolution in the human genome PLoS Genet 2007 3 10.1371/journal.pgen.0030090 17542651 13. Qanbari S Pausch H Jansen S Somel M Strom TM Fries R Nielsen R Simianer H Classic selective sweeps revealed by massive sequencing in cattle PLoS Genet 2014 10 10.1371/journal.pgen.1004148 24586189 14. Sabeti PC Varilly P Fry B Lohmueller J Hostetter E Cotsapas C Xie X Byrne EH McCarroll SA Gaudet R Schaffner SF Lander ES Frazer KA Ballinger DG Cox DR Hinds DA Stuve LL Gibbs RA Belmont JW Boudreau A Hardenbol P Leal SM Pasternak S Wheeler DA Willis TD Yu F Yang H Zeng C Gao Y Hu H Genome-wide detection and characterization of positive selection in human populations Nature 2007 449 913 918 10.1038/nature06250 17943131 15. Chen H Patterson N Reich D Population differentiation as a test for selective sweeps Genome Res 2010 20 393 402 10.1101/gr.100545.109 20086244 16. Groenen MAM Megens H-J Zare Y Warren WC Hillier LW Crooijmans RPMA Vereijken A Okimoto R Muir WM Cheng HH The development and characterization of a 60K SNP chip for chicken BMC Genomics 2011 12 274 10.1186/1471-2164-12-274 21627800 17. Fu W Dekkers JCM Lee WR Abasht B Linkage disequilibrium in crossbred and pure line chickens Genet Sel Evol 2015 47 1 11 10.1186/s12711-015-0098-4 25887184 18. Groenen MAM Wahlberg P Foglio M Cheng HH Megens H Crooijmans RPMA Besnier F Lathrop M Muir WM Wong GK-S Gut I Andersson L A high-density SNP-based linkage map of the chicken genome reveals sequence features correlated with recombination rate Genome Res 2009 19 510 519 10.1101/gr.086538.108 19088305 19. Browning BL Browning SR A unified approach to genotype imputation and haplotype-phase inference for large data sets of trios and unrelated individuals Am J Hum Genet 2009 84 210 223 10.1016/j.ajhg.2009.01.005 19200528 20. Pickrell JJK Coop G Novembre J Kudaravalli S Li JZ Absher D Srinivasan BS Barsh GS Myers RM Feldman MW Signals of recent positive selection in a worldwide sample of human populations Genome Res 2009 19 826 837 10.1101/gr.087577.108 19307593 21. Ma. Y, Zhang. H, Zhang Q, Ding X. Identification of selection footprints on the X Chromosome in pig. PLoS One. 2014;9:e94911. 22. Zhao FP Wei CH Zhang L Liu JS Wang GK Zeng T Du LX A genome scan of recent positive selection signatures in three sheep populations J Integr Agric 2016 15 162 174 10.1016/S2095-3119(15)61080-2 23. Yin T Cook D Lawrence M ggbio: an R package for extending the grammar of graphics for genomic data Genome Biol 2012 13 R77 10.1186/gb-2012-13-8-r77 22937822 24. Reich DE Cargill M Bolk S Ireland J Sabeti PC Richter DJ Lavery T Kouyoumjian R Farhadian SF Ward R Lander ES Linkage disequilibrium in the human genome Nature 2001 411 199 204 10.1038/35075590 11346797 25. Hayes B, Visscher P. Novel multilocus measure of linkage disequilibrium to estimate past effective population size. Genome Res. 2003;13:635-43.  26. Tenesa A, Navarro P, Hayes BJ, Duffy DL, Clarke GM, Goddard ME, Visscher PM. Recent human effective population size estimated from linkage disequilibrium. Genome Res. 2007;17:520–26. 27. Liu G Dunnington EA Siegel PB Correlated responses to long-term divergent selection for eight-week body weight in chickens: growth, sexual maturity, and egg production Poult Sci 1995 74 1259 1268 10.3382/ps.0741259 7479503 28. Anthony NB Dunnington E a Siegel PB Egg Production and Egg Composition of Parental Lines and F1 and F2 Crosses of White Rock Chickens Selected for 56-Day Body Weight Poult Sci 1989 68 27 36 10.3382/ps.0680027 29. Jaap RG Muir FV Erratic Oviposition and egg defects in broiler-type Pullets Poult Sci 1968 47 417 423 10.3382/ps.0470417 30. Lagrutta AA McCarthy JG Scherczinger CA Heywood SM Identification and developmental expression of a novel embryonic myosin heavy-chain gene in chicken DNA 1989 8 39 50 10.1089/dna.1.1989.8.39 2707122 31. Lyons G, Ontell M, Cox R. The expression of myosin genes in developing skeletal muscle in the mouse embryo. J Cell Biol. 1990;111:1465–76. 32. Tajsharghi H, Kimber E, Kroksmark AK, Jerre R, Tulinius M, Oldfors A. Embryonic myosin heavy-chain mutations cause distal arthrogryposis and developmental myosin myopathy that persists postnatally. Arch Neurol. 2008;65:1083-90. 33. Wang L, Liu X, Niu F, Wang H, He H, Gu Y. Single nucleotide polymorphisms, haplotypes and combined genotypes in MYH3 gene and their associations with growth and carcass traits in Qinchuan cattle. Mol Biol Reports. 2013;40:417-26. 34. Niu F Wang L Liu X Wang H Yang J Genetic diversity of MYH3 gene associated with growth and carcass traits in Chinese Qinchuan cattle Mol Biol Rep 2013 40 5635 5643 10.1007/s11033-013-2665-5 24065521 35. Smits P Li P Mandel J Zhang Z Deng JM Behringer RR de Crombrugghe B Lefebvre V The transcription factors L-Sox5 and Sox6 are essential for cartilage formation Dev Cell 2001 1 277 290 10.1016/S1534-5807(01)00003-X 11702786 36. Smits P Dy P Mitra S Lefebvre V Sox5 and Sox6 are needed to develop and maintain source, columnar, and hypertrophic chondrocytes in the cartilage growth plate J Cell Biol 2004 164 747 758 10.1083/jcb.200312045 14993235 37. Hagiwara N Yeh M Liu A Sox6 is required for normal fiber type differentiation of fetal skeletal muscle in mice Dev Dyn 2007 236 2062 2076 10.1002/dvdy.21223 17584907 38. Goll DE, Thompson VF, Taylor RG, Christiansen JA. Role of the calpain system in muscle growth. Biochimie. 1992;74:225-37. 39. Huang J Forsberg NE Role of calpain in skeletal-muscle protein degradation Proc Natl Acad Sci 1998 95 12100 12105 10.1073/pnas.95.21.12100 9770446 40. Arthur JS, Elce JS, Hegadorn C, Williams K, Greer PA. Disruption of the murine calpain small subunit gene, Capn4: calpain is essential for embryonic development but not for cell growth and division. Mol Cell Biol. 2000;20:4474-81. 41. Lindholm‐Perry AK Rohrer GA Holl JW Shackelford SD Wheeler TL Koohmaraie M Nonneman D Relationships among calpastatin single nucleotide polymorphisms, calpastatin expression and tenderness in pork longissimus1 Anim Genet 2009 40 713 721 10.1111/j.1365-2052.2009.01903.x 19422367 42. Hu Y-D Zhang Z-R Zhu Q Identification and Association of the Single Nucleotide Polymorphisms in Calpastatin (CAST) Gene with Carcass Traits in Chicken J Anim Vet Adv 2011 10 2968 2974 43. Allais S, Journaux L, Levéziel H, Payet-Duprat N, Raynaud P, Hocquette JF, Lepetit J, Rousset S, Denoyelle C, Bernard-Capel C, Renand G. Effects of polymorphisms in the calpastatin and μ-calpain genes on meat tenderness in 3 French beef breeds. J Anim Sci. 2011;89:1-1. 44. Byun SO Zhou H Forrest RHJ Frampton CM Hickford JGH Association of the ovine calpastatin gene with birth weight and growth rate to weaning Anim Genet 2008 39 572 573 10.1111/j.1365-2052.2008.01745.x 18513277 45. Greguła-Kania M Effect of calpastatin gene polymorphism on lamb growth and muscling Ann Anim Sci 2012 12 63 72 46. Tait RG, Shackelford SD, Wheeler TL, King DA, Casas E, Thallman RM, Smith TP, Bennett GL. μ-Calpain, calpastatin, and growth hormone receptor genetic effects on preweaning performance, carcass quality traits, and residual variance of tenderness in Angus cattle selected to increase minor haplotype and allele frequencies. J Anim Sci. 2014;92:456-66. 47. Wu T, Zhang Z, Yuan Z, Lo LJ, Chen J, Wang Y, Peng J. Distinctive Genes Determine Different Intramuscular Fat and Muscle Fiber Ratios of the longissimus dorsiMuscles in Jinhua and Landrace Pigs. PLoS One. 2013;8:e53181. 48. Gunning P O’neill G Hardeman E Tropomyosin-based regulation of the actin cytoskeleton in time and space Physiol Rev 2008 88 1 35 10.1152/physrev.00001.2007 18195081 49. Kuninger D Kuzmickas R Peng B Pintar JE Rotwein P Gene discovery by microarray: identification of novel genes induced during growth factor-mediated muscle cell survival and differentiation Genomics 2004 84 876 889 10.1016/j.ygeno.2004.07.013 15475267 50. McNally E Dellefave L Sarcomere mutations in cardiogenesis and ventricular noncompaction Trends Cardiovasc Med 2009 19 17 21 10.1016/j.tcm.2009.03.003 19467449 51. Johnsson M Rubin C Höglund A Sahlqvist A Jonsson KB Kerje S Ekwall O Kämpe O Andersson L Jensen P The role of pleiotropy and linkage in genes affecting a sexual ornament and bone allocation in the chicken Mol Ecol 2014 23 2275 2286 10.1111/mec.12723 24655072 52. Qanbari S Strom TM Haberer G Weigend S Gheyas AA Turner F Burt DW Preisinger R Gianola D Simianer H A High Resolution Genome-Wide Scan for Significant Selective Sweeps: An Application to Pooled Sequence Data in Laying Chickens PLoS One 2012 7 e49525 10.1371/journal.pone.0049525 23209582 53. Muir WM, Aggrey SE. Poultry Genetics, Breeding, and Biotechnology. Cambridge: CABI Publishing; 2003. 54. Thiruvenkadan AK, Panneerselvam S, Prabakaran R. Layer breeding strategies: an overview. Worlds Poult Sci J. 2010;66:477-502. 55. Thiruvenkadan a K Prabakaran R Panneerselvam S Broiler breeding strategies over the decades: an overview Worlds Poult Sci J 2011 67 309 336 10.1017/S0043933911000328 56. Wolc A Stricker C Arango J Settar P Fulton JE O’Sullivan NP Preisinger R Habier D Fernando R Garrick DJ Lamont SJ Dekkers JCM Breeding value prediction for production traits in layer chickens using pedigree or genomic relationships in a reduced animal model Genet Sel Evol 2011 43 5 10.1186/1297-9686-43-5 21255418 57. Lacin E Yildiz A Esenbuga N Macit M Effects of differences in the initial body weight of groups on laying performance and egg quality parameters of Lohmann laying hens Czech J Anim Sci 2008 53 466 471 58. Zhou H Mitchell AD McMurtry JP Ashwell CM Lamont SJ Insulin-like growth factor-I gene polymorphism associations with growth, body composition, skeleton integrity, and metabolic traits in chickens Poult Sci 2005 84 212 219 10.1093/ps/84.2.212 15742956 59. Lazar MA Thyroid hormone receptors: multiple forms, multiple possibilities Endocr Rev 1993 14 184 193 8325251 60. Kaneshige M, Kaneshige K, Zhu XG, Dace A, Garrett L, Carter TA, Kazlauskaite R, Pankratz DG, Wynshaw-Boris A, Refetoff S, Weintraub B. Mice with a targeted mutation in the thyroid hormone β receptor gene exhibit impaired growth and resistance to thyroid hormone. Proc Natl Acad Sci. 2000;97:13209-14. 61. Dittrich R Beckmann MW Oppelt PG Hoffmann I Lotz L Kuwert T Mueller A Thyroid hormone receptors and reproduction J Reprod Immunol 2011 90 58 66 10.1016/j.jri.2011.02.009 21641659 62. Krassas GE Poppe K Glinoer D Thyroid function and human reproductive health Endocr Rev 2010 31 702 755 10.1210/er.2009-0041 20573783

PMC005xxxxxx/PMC5002101.txt

==== Front Malar JMalar. JMalaria Journal1475-2875BioMed Central London 143010.1186/s12936-016-1430-3ResearchAdding a single low-dose of primaquine (0.25 mg/kg) to artemether-lumefantrine did not compromise treatment outcome of uncomplicated Plasmodium falciparum malaria in Tanzania: a randomized, single-blinded clinical trial Mwaiswelo Richard richiemwai@yahoo.com 1Ngasala Billy bngasala@muhas.ac.tz 1Jovel Irina irina.jovel@ki.se 2Aydin-Schmidt Berit berit.schmidt@ki.se 2Gosling Roland roly.gosling@ucsf.edu 34Premji Zul premjizulfiqarali@gmail.com 5Mmbando Bruno b.mmbando@yahoo.com 6Björkman Anders Anders.Bjorkman@ki.se 2Mårtensson Andreas andreas.martensson@kbh.uu.se 71 Department of Parasitology and Medical Entomology, Muhimbili University of Health and Allied Sciences, Dar Es Salaam, Tanzania 2 Department of Microbiology, Tumor and Cell Biology, Karolinska Institutet, Stockholm, Sweden 3 Department of Epidemiology and Biostatistics, University of California San Francisco, San Francisco, CA USA 4 Global Health Group, University of California San Francisco, San Francisco, CA USA 5 Aga Khan University Hospital, Nairobi, Kenya 6 Tanga Centre, National Institute for Medical Research, Tanga, Tanzania 7 Department of Women’s and Children’s Health, International Maternal and Child Health (IMCH), Uppsala University, Uppsala, Sweden 26 8 2016 26 8 2016 2016 15 1 43514 4 2016 5 7 2016 © The Author(s) 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Background The World Health Organization (WHO) recently recommended the addition of a single low-dose of the gametocytocidal drug primaquine (PQ) to artemisinin-based combination therapy (ACT) in low transmission settings as a component of pre-elimination or elimination programmes. However, it is unclear whether that influences the ACT cure rate. The study assessed treatment outcome of artemether-lumefantrine (AL) plus a single PQ dose (0.25 mg/kg) versus standard AL regimen for treatment of acute uncomplicated Plasmodium falciparum malaria in Tanzania. Methods A randomized, single-blinded, clinical trial was conducted in Yombo, Bagamoyo district, Tanzania. Acute uncomplicated P. falciparum malaria patients aged ≥1 year, with the exception of pregnant and lactating women, were enrolled and treated with AL plus a single PQ dose (0.25 mg/kg) or AL alone under supervision. PQ was administered together with the first AL dose. Clinical and laboratory assessments were performed at 0, 8, 24, 36, 48, 60, and 72 h and on days 7, 14, 21, and 28. The primary end-point was a polymerase chain reaction (PCR)-adjusted adequate clinical and parasitological response (ACPR) on day 28. Secondary outcomes included: fever and asexual parasitaemia clearance, proportion of patients with PCR-determined parasitaemia on day 3, and proportion of patients with Pfmdr1 N86Y and Pfcrt K76T on days 0, 3 and day of recurrent infection. Results Overall 220 patients were enrolled, 110 were allocated AL + PQ and AL, respectively. Parasite clearance by microscopy was fast, but PCR detectable parasitaemia on day 3 was 31/109 (28.4 %) and 29/108 (26.9 %) in patients treated with AL + PQ and AL, respectively (p = 0.79). Day 28 PCR-adjusted ACPR and re-infection rate was 105/105 (100 %) and 101/102 (99 %) (p = 0.31), and 5/107 (4.7 %) and 5/8 (4.8 %) (p = 0.95), in AL + PQ and AL arm, respectively. There was neither any statistically significant difference in the proportion of Pfmdr1 N86Y or Pfcrt K76T between treatment arms on days 0, 3 and day of recurrent infection, nor within treatment arms between days 0 and 3 or day 0 and day of recurrent infection. Conclusion The new WHO recommendation of adding a single low-dose of PQ to AL did not compromise treatment outcome of uncomplicated P. falciparum malaria in Tanzania. Trial registration number NCT02090036 Electronic supplementary material The online version of this article (doi:10.1186/s12936-016-1430-3) contains supplementary material, which is available to authorized users. Keywords Plasmodium falciparum malariaArtemether-lumefantrinePrimaquineCure rateSidaBil-Tz 16/9875007059Ngasala Billy issue-copyright-statement© The Author(s) 2016 ==== Body Background Artemisinin-based combination therapy (ACT) is generally recommended as first-line treatment for uncomplicated Plasmodium falciparum malaria globally [1]. Recently, the World Health Organization (WHO) has recommended the addition of a 0.25 mg/kg single-dose of the gametocytocidal drug primaquine (PQ) to standard ACT regimen as a component of pre-elimination or elimination of malaria in low-intensity transmission settings and for containment in areas threatened by artemisinin resistance [2, 3]. Most concerns with this new recommendation have been on safety, due to the potential risk of PQ-induced haemolysis in glucose-6-phosphate dehydrogenase (G6PD)-deficient patients. Equally important is to ensure that the treatment outcome of ACT is not compromised by the addition of a single-low dose of PQ. This is of particular concern since the individual patient with uncomplicated malaria does not personally benefit from PQ intake, the potential benefits, i.e., of reduced transmission, are rather on community level. However, no study has reported on the cure rate of ACT in addition to 0.25 mg/kg single-dose PQ. Anti-malarial drug efficacy depends upon appropriate drug levels being reached and maintained for a long enough time for the drug to act [4]. Insufficient exposure is associated with increased risk of treatment failure. Inhibition of drug metabolism through drug–drug interaction may lead to insufficient exposure and consequently reduced efficacy [4, 5]. However, comprehensive data on potential interactions between artemether-lumefantrine (AL) and PQ are currently lacking, and therefore it remains unclear whether the addition of this single low-dose (0.25 mg/kg) PQ may compromise the efficacy of AL [6]. The aim of this study was to assess treatment outcome of the recent WHO recommendation of adding a single PQ dose (0.25 mg/kg) to AL versus standard AL regimen for treatment of acute uncomplicated P. falciparum malaria in Tanzania. Methods Study area This trial was conducted at Yombo primary health facility, Bagamoyo district, Tanzania, between July and November, 2014. The health facility is located southwest, about 20 km, from Bagamoyo town. It serves approximately 7000 people and has capability to carry out routine malaria microscopy and rapid diagnostic test. Malaria transmission is high and occurs throughout the year with peaks related to the long rain season from May to July and short rain season from November to December. In the study area, P. falciparum is the major malaria species and Anopheles gambiae sensu stricto the main vector [7–9]. AL has been used as the first-line treatment for uncomplicated malaria since 2006. Sulfadoxine-pyrimethamine is used for intermittent preventive treatment in pregnant women. Long-lasting, insecticide-treated mosquito nets is the major vector control method [10]. G6PD deficiency prevalence in the study area has previously been estimated to 13.6 % in hemizygous males and 4.5 % in homozygous females [11]. Study design This was a randomized, single-blinded, clinical trial comparing treatment outcome and safety of AL plus a single low-dose PQ (AL + PQ) versus standard AL regimen. Safety outcomes have been presented in a separate publication [11]. Patients with uncomplicated microscopically confirmed P. falciparum mono-infection were enrolled, randomly assigned to either AL + PQ or AL treatment, admitted during the first 3 days of treatment and thereafter followed up until day 28 after treatment initiation. Treatment outcome was based on polymerase chain reaction (PCR)-adjusted parasitological cure. Therapeutic failures were classified as; early treatment failure (ETF), late clinical failure (LCF), or late parasitological failure (LPF) [12]. Study population Patients presenting at the study site with suspected acute uncomplicated malaria were screened for eligibility. Inclusion criteria were age ≥1 year, weight ≥10 kg, body temperature ≥37.5 °C or history of fever in the last 24 h, microscopy confirmed P.falciparum mono-infection, any parasitaemia level, ability to swallow oral medication, ability and willingness to abide by the study protocol and the stipulated follow-up visits, and a written informed consent (in case of children a proxy consent from a parent/guardian). Exclusion criteria were evidence of severe malaria or danger signs, known allergy to trial medicines, reported anti-malarial intake ≤2 weeks, haemoglobin (Hb) <8 g/dL, blood transfusion within last 90 days, febrile condition other than malaria, known underlying chronic or severe disease (including severe malnutrition), pregnancy and breastfeeding. Randomization and blinding Treatment allocation was done using sex-stratified, block randomization with four blocks, two per treatment arm using RESEARCH RANDOMIZER (version 4) (computer software) (Wesleyan University, Connecticut, USA) [13]. Opaque envelopes containing the pre-determined treatment codes were kept serially in a male and female strata. The envelopes were opened by a study nurse just before the first treatment dose was administered. Patients were blinded to the assigned treatment. Treatment A standard, weight-based, three-day course of AL (Coartem®, Novartis) was administered to all patients according to Tanzania national treatment guideline for uncomplicated P. falciparum malaria [14]. AL dispersible tablets suspended in water were administered to children who were not able to swallow tablets. A single 0.25 mg/kg PQ dose (Primaquine phosphate, Sanofi) was administered together with AL first dose to patients assigned AL + PQ treatment. The accuracy of PQ dose among patients weighing less than 60 kg body weight was ensured by administering the drug in aqueous solution, whereas for adults weighing 60 kg and above the drug was administered as tablet. To make aqueous solution, a 15 mg PQ tablet was suspended in 15 mL of water, and the dose was measured using a sterile syringe based on body weight. To achieve the single blinding, the PQ dose was prepared in the absence of patients, and a glucose-based syrup was added to mask the PQ taste, whereas for patients allocated AL alone, the same glucose-based syrup was administered with AL first dose. Adult patients allocated to the AL + PQ arm, during consenting, never knew in which treatment arm they were allocated, and during drug administration, PQ tablet was mixed with AL tablets, however, in case patients asked about the additional tablet, they were told that the tablet was given to prevent treatment side effects. In order to optimize AL absorption and minimize PQ gastro-intestinal side effects, biscuits were administered prior to all drug doses [15]. A study nurse administered/supervised intake of all drug doses. Patients were monitored for 30 min after each drug dose. Treatment was re-administered in case of vomiting within this period. Lost follow-up and patient withdrawal Study withdrawal criteria were: vomiting the trial medicine >three times, withdrawal of consent, intake of any medicine with anti-malarial properties outside the study protocol, or any protocol violation. Subjects were considered lost to follow-up and consequently withdrawn if they missed a scheduled follow-up visit and did not attend on the successive days despite efforts to trace them at their homes [12]. A subject who returned before the last day of follow-up was not considered a lost follow-up. Subjects with symptoms/signs of severe disease (including repetitive vomiting of trial medicine) were treated according to the national guideline and were then followed up until recovery. Procedures Clinical assessment was performed at 0, 8, 24, 36, 48, 60, and 72 h and on days 7, 10, 14, 21, 28 or on any day of recurrent illness. The assessment included history of clinical symptoms, possible adverse events, concomitant drug consumption and clinical examination including measurement of axillary temperature. Fever was defined as body temperature ≥37.5 °C. A case record form was used to record all clinical and laboratory data. Laboratory assessment involved collection of finger-prick blood samples for Hb concentration, thick smears for microscopy-determined asexual and sexual parasitaemia, thin smears for species determination (at enrolment) and filter-paper blood samples for parasite detection and genotyping by PCR. Hb assessment was done on days 0, 1, 2, 3, 7, 10, 14, 21, 28 or on any day of recurrent illness, whereas, filter paper and blood slide samples were collected at 0, 8, 24, 36, 48, 60, and 72 h and on days 7, 10, 14, 21, 28 or on any day of recurrent illness. Hb concentration was measured using a portable spectrophotometer, HemoCue Hb 201+ (HemoCue AB, Ängelholm Sweden), with a precision of ±0.3 g/dL. A control cuvette at 16.0 ± 0.3 g/dL was used for daily calibration according to manufacturer’s instruction. A 10 % Giemsa solution was used to stain thick and thin blood smears. Thin smears were prepared once, i.e., at enrolment, whereas thick smears were prepared at all sampling time points. Asexual parasites were counted against 200 white blood cells (WBC). The obtained number was multiplied by 40, assuming 8000 leukocytes per μL of blood, to gain an approximate parasite count. In case no parasites were observed after examining 100 fields, the blood slide was considered negative. Each slide was read by two independent microscopists. In case they disagreed on presence of parasitaemia or if density differed by more than 25 %, a third independent reading was performed. In case of positive versus negative results, a third independent reading was used to confirm the reading of the first two readers. The filter paper (3MM Whatman) blood samples were labelled, air-dried at room temperature for 3–4 h and then packed in individual plastic bags and stored. After study completion, all filter papers were transported to Karolinska Institutet, Sweden, for molecular analysis. Molecular analysis A 10 % Chelex-100® method was used to extract genomic DNA from dried blood spots [16]. Paired blood samples (pre-treatment and day of recurrent parasitaemia) from patients classified as LCF or LPF were genotyped to differentiate recrudescence from re-infection by stepwise genotyping of P. falciparum block 3 of merozoite surface protein (msp) 2, block 2 of msp 1 and region II (RII) of glutamate-rich protein (glurp) [17]. The respective initial amplifications were followed by individual nested PCR reactions using family specific primers for msp1 (K1, MAD20 and RO33) and msp2 (FC27 and IC) and semi-nested for RII of glurp [17]. The amplicons were loaded on a GelRed™ (Biotium, Hayward, CA, USA) stained agarose gel, separated by electrophoresis and then visualized under ultraviolet transillumination (Gel Doc™, Bio-Rad, Hercules, CA, USA), and sized by Image Lab™ software (Bio-Rad, Hercules, CA, USA). Alleles in each family were considered the same if fragments size were within 20 base pair interval. Subjects with recurrent parasitaemia by microscopy, which could not be confirmed by PCR were considered to have uncertain-PCR-adjusted outcome. Recrudescence was defined as presence of at least one matching allelic band, and re-infection was defined as absence of any matching allelic band at baseline and on the day of parasite recurrence [17]. In addition, P. falciparum multi-drug resistant gene 1 (Pfmdr1) asparagine (N)-86-tyrosine (Y) and chloroquine resistance transporter gene (Pfcrt) lysine (K)-76-threonine (T) were genotyped using nested PCR followed by restriction fragment length polymorphism using ApoI restriction enzyme as previously described [18]. Study end-points The primary outcome was the proportion of patients with PCR-adjusted adequate clinical and parasitological response (ACPR) by day 28. Secondary outcomes included: fever and asexual parasite clearance, proportion of patients with PCR-determined parasitaemia on day 3, PCR- determined re-infection rate, and proportion of patients with Pfmdr1 N86Y and Pfcrt K76T on days 0, 3 and day of recurrent infection. Statistical analysis Sample size calculation was based on equivalence, defined as margin of 10 % PCR-adjusted ACPR, between the treatments. Allowing for 10 % attrition, 80 % power at 0.05 significance, a sample size of 110 per treatment arm was required. Data were double-entered in an electronic database and analysed using SPSS software version 16 (SPSS Inc, Chicago, USA) as per protocol. Independent sample t test was used to compare means. Non-parametric data were compared using Chi square, Fisher’s or McNemar tests as appropriate. Cure rate end points were analysed by survival analysis, and the survival curves of the two treatment arms were plotted and compared using log-rank test. Data were censored at the time of withdrawal for patients lost to follow up, withdrew consent and PCR determined re-infection or uncertain PCR outcome. A p ≤ 0.05 was considered significant. Results Patients’ characteristics and trial profile The flow of patients through the trial is presented in Fig. 1. In summary, a total of 1065 subjects were screened for eligibility, of whom 845 were not included. The remaining 220 were enrolled, 110 per treatment arm. Pre-treatment characteristics of the study participants are presented in Table 1.Fig. 1 Flow of patients through the trial Table 1 Pre-treatment characteristics of the study subjects Characteristic Treatment arm P-value AL + PQ AL n = 110 n = 110 Female, n (%) 55 (50) 55 (50) 1.0 Age (years), median (range) 9 (5.6–15) 10 (5–23) 0.392‡ Body weight (kg), mean (SD) 30.7 (17.1) 34.3 (19.6) 0.145† Body temperature (°C), mean (SD) 38.3 (1.1) 38.3 (1.3) 0.977† Fever (≥37.5 °C), n (%) 88 (80) 80 (73) 0.204¥ Geometric mean asexual parasite density/µl (95 % CI) 8327 (5451–12,717) 8384 (5437–12,927) 0.982† ‡Mann–Whitney test, † Student t-test, ¥ Chi square test Treatment outcomes Treatment outcomes are presented in Table 2 and Fig. 2. There was no statistically significant difference in the cure (p = 0.31) or re-infection rates (p = 0.95) by day 28 between the treatment arms. The survival curves for the cure rate were not statistically significant different (p = 0.94).Table 2 Treatment outcome Outcome Treatment arm P-value AL + PQ (n = 107) AL (n = 104) Early treatment failure 0 0 Late clinical failure 2 (1.8 %) 6 (5.8 %) 0.16 Late parasitological failure 5 (4.7 %) 2 (1.9 %) 0.45 Crude cure rate by day 28 100 (93.5 %) 96 (92.3 %) 0.73 Uncertain PCR outcome by day 28 2 (1.9 %) 2 (1.9 %)  1.0 PCR-determined re-infection rate by day 28 5 (4.7 %) 5 (4.8 %) 0.95 Recrudescence 0 1 (1.0 %) 0.49 PCR-adjusted ACPR by day 28 105 (100.0 %) 101 (99.0 %) 0.31 ACPR adequate clinical and parasitological response Fig. 2 Kaplan-Meier survival curve for cure rate of subjects treated with AL + PQ and AL Gametocyte carriage Microscopic gametocytaemia was only detected in one patient at enrolment and in two patients at 24 h, both treated with AL + PQ. Thereafter, no gametocyte carriage was detected. Parasite and fever clearance Microscopy-determined asexual parasite and fever clearance was rapid and similar across treatment arms (Figs. 3 and 4). No parasite carriage was observed after 36 and 48 h in AL + PQ and AL arm, respectively.Fig. 3 Microscopy-determined parasite clearance following treatment with AL + PQ and AL Fig. 4 Fever clearance following treatment with AL + PQ and AL PCR detectable parasitaemia on day 3 On day 3, PCR detectable parasitaemia was found in 31/109 (28.4 %) and 29/108 (26.9 %) patients treated with AL + PQ and AL, respectively (p = 0.79). Prevalence of Pfmdr1 N86Y and Pfcrt K76T The distribution of Pfmdr1 N86Y and Pfcrt K76T on days 0, 3 and day of recurrent infection across treatments is presented in Table 3. There was neither any statistically significant difference in the proportion of Pfmdr1 N86Y or Pfcrt K76T between treatment arms on days 0, 3 and day of recurrent infection, nor within treatment arms between days 0 and 3 or day 0 and day of recurrent infection. Pooling of all 220 patients did not result in a statistically significant selection in Pfmdr1 N86 (p = 0.07, McNemar test) or Pfcrt K76 (p = 0.25, McNemar test) between days 0 and 3, or between day 0 and day of recurrent infection [Pfmdr1 N86 (p = 1.0, McNemar test), Pfcrt K76 (p = 1.0, McNemar test)].Table 3 Pfmdr1 N86Y and Pfcrt K76T distribution on day 0, 3 and day of recurrent infection Treatment arm AL+PQ AL P-value Day 0  Pfmdr1 N86 91 (85.8 %) 92 (85.9 %) 0.49 86Y 11 (10.4 %) 8 (7.5 %) 0.45 N86Y 4 (3.8 %) 7 (6.5 %) 0.54  Pfcrt K76 97 (95.1 %) 101 (98.1 %) 0.49 76T 1 (0.9 %) 0 (0 %) 0.71 K76T 4 (3.9 %) 2 (1.9 %) 0.44 Day 3  Pfmdr1 N86 15 (65.2 %) 20 (76.9 %) 0.59 86Y 2 (8.6 %) 2 (7.7 %) 0.54 N86Y 6 (26.1 %) 4 (15.4 %) 0.48  Pfcrt K76 24 (85.7 %) 27 (96.4 %) 0.24 76T 2 (7.1 %) 0 (0 %) 0.6 K76T 2 (7.1 %) 1 (3.6 %) 0.6 Day of recurrent parasitaemia  Pfmdr1 N86 2 (66.7 %) 6 (85.7 %) 0.28 86Y 0 (0 %) 0 (0 %) – N86Y 1 (33.3 %) 1 (14.3 %) 1.0  Pfcrt K76 2 (100 %) 4 (80 %) 0.68 76T 0 (0 %) 0 (0 %) – K76T 0 (0 %) 1 (20 %) 1.0 Discussion This study provides much needed data on treatment outcome of the new WHO recommendation of adding a single low-dose (0.25 mg/kg) of PQ to AL compared with standard AL treatment alone. Both regimens provided high cure rates for the treatment of acute uncomplicated P. falciparum malaria in Tanzania, with similar PCR-ACPR as previously has been reported from the same study area of AL alone [19–21], and that of other previously reported studies on AL plus 0.75, 0.4 or 0.1 mg/kg single-dose PQ [15, 22]. Similar crude cure rates and PCR-determined re-infection rates were also found between the treatment arms. Moreover, asexual parasite clearance was rapid in both arms. However, a relatively large, but equal, proportion of patients across treatment arms had PCR-detectable parasitaemia on day 3. These findings therefore suggest that the addition of a single low-dose PQ (0.25 mg/kg) does neither compromise the treatment outcome nor interfere with parasite clearance of AL. Previous studies have reported selection of Pfmdr1 N86 and Pfcrt K76 among recurrent infections during follow-up after AL treatment [23–25], and also between enrolment and day 3 [26]. A similar selection was not observed in the present study. This may be due to that there has been a temporal selection of Pfmdr1 N86 and Pfcrt K76 in the study area ever since AL was introduced as first-line treatment of uncomplicated malaria in 2006 [27], which has resulted in an overall higher baseline prevalence of the two alleles in recent years, which in turn may mask the selection potential of lumefantrine on these drug tolerance/resistance markers. Importantly, adding a single 0.25 mg/kg dose of PQ to AL did not apparently increase the selection potential of AL on Pfmdr1 N86 and Pfcrt K76, neither during the early treatment phase nor among recurrent infections during follow-up after treatment. Thus, the findings suggest that addition of this single low-dose PQ (0.25 mg/kg) does not apparently spur the selection potential of drug resistance markers previously associated with lumefantrine tolerance/resistance. The study has some potential limitations including that it involved patients of all ages with all levels of parasitaemia despite the WHO recommendation to conduct efficacy studies in children under the age of 5 years with parasitaemia level of 2000–200,000/µl of blood in highly malaria-endemic countries such as Tanzania, to minimize the influence of immunity on treatment outcome in older individuals. However, this study was done to assess whether adding PQ to AL would potentially compromise the efficacy of latter drug probably due to drug–drug interaction, that can be done at all ages and levels of parasitaemia. Conclusion The new WHO recommendation of adding a single low-dose (0.25 mg/kg) of PQ to AL did not compromise treatment outcome of uncomplicated P. falciparum malaria in Tanzania. The findings support the use of a single low-dose of PQ together with AL as a potential component of ongoing P. falciparum malaria elimination efforts. Additional file 10.1186/s12936-016-1430-3 Artemether-lumefantrine + primaquine versus artemether-lumefantrine dataset. Abbreviations ACTartemisinin-based combination therapy ACPRadequate clinical and parasitological response ALartemether-lumefantrine CIconfidence interval crtchloroquine resistance CYPcytochrome P DNAdeoxyribonucleic acid ETFearly treatment failure GCPgood clinical practices G6PDglucose-6-phosphate dehydrogenase Hbhaemoglobin IMCHInternational maternal and child health Klysine LCFlate clinical failure LPFlate parasitological failure mdrmultidrug resistance Nasparagine nsample size NMCPNational malaria control programme PCRpolymerase chain reaction PQprimaquine RFLPrestriction fragment length polymorphism SDstandard deviation SPSSstatistical package for social sciences Tthreonine WBCwhite blood cell WHOWorld Health Organization Ytyrosine Authors’ contributions AB, AM, BN, and RM provided the conception and design of the study. RM collected data in the field. BA, IJ and RM performed molecular analysis. BM and RM performed data analysis and interpretation. RM drafted the manuscript together with AM and BN. AB, BM and ZP helped to draft the manuscript. All authors revised the manuscript critically for intellectual content. All authors read and approved the final manuscript. Acknowledgements The authors thank all the patients and parents/guardians for participating in the study. Sanofi provided the primaquine tablets. Competing interests The authors declare that they have no competing interests. Availability of data and material The dataset supporting the conclusions of this article is included within the article (and its Additional file 1). Consent for publication Not applicable. Ethics approval and consent to participate The trial adhered to good clinical practice (GCP), the Declaration of Helsinki, and applicable regulatory requirements in Tanzania. The ethics committees of the Muhimbili University of Health and Allied Sciences and the National Institute for Medical Research approved the trial. The Tanzania Food and Drug Authority ethics committee approved the importation and use of PQ. During the trial, an independent data monitoring and safety board conducted an interim safety analysis. All patients and parents/guardians of patients aged <18 years provided written informed consent, prior to enrolment. This study is registered at clinicaltrials.gov (NCT02090036). Funding Financial support was obtained from the bilateral sida grant: The Swedish Development Cooperation Agency (Bil-Tz 16/9875007059) and Swedish Research Council (VR) [2013–6594]. ==== Refs References 1. WHO. World Malaria Report 2010. Geneva, World Health Organization, 2010. 2. WHO. Evidence Review Group: The safety and effectiveness of single dose primaquine as a P. falciparum gametocytocide. Geneva, World Health Organization, 2012. 3. WHO. Guidelines for the treatment of malaria. 3rd ed. Geneva, World Health Organization, 2015. 4. Barnes KI Lindegardh N Ogundahunsi O Olliaro P Plowe CV Randrianarivelojosia M World Antimalarial Resistance Network (WARN) IV: clinical pharmacology Malar J. 2007 6 122 10.1186/1475-2875-6-122 17822537 5. Dooley KE Flexner C Andrade AS Drug interactions involving combination antiretroviral therapy and other anti-infective agents: repercussions for resource-limited countries J Infect Dis 2008 198 1 14 10.1086/591459 18544010 6. Eziefula AC Gosling R Hwang J Hsiang MS Bousema T von Seidlein L Rationale for short course primaquine in Africa to interrupt malaria transmission Malar J. 2012 11 360 10.1186/1475-2875-11-360 23130957 7. Mnzava AEP Kilama WL Observations on the ecological distribution of Anopheles gambiae complex in Tanzania Acta Trop 1986 43 277 282 2877554 8. Temu EA Minjas JN Coetzee M Hunt RH Shiff CJ The role of four anopheline species (Diptera: Culicidae) in malaria transmission in coastal Tanzania Trans R Soc Trop Med Hyg 1998 92 152 158 10.1016/S0035-9203(98)90724-6 9764318 9. Kabula B Derua Y Tungu P Massue DJ Sambu E Stanley G Malaria entomological profile in Tanzania from 1950 to 2010: A review of mosquito distribution, vectorial capacity and insecticide resistance Tanzan J Health Res. 2011 13 1 14 10.4314/thrb.v13i3.66915 24409640 10. NMCP. National malaria stratergic plan 2014–2020. National Malaria Control Prograamme. Ministry of Health and Social Welfare. The United Republic of Tanzania. http://ihi.eprints.org/3314/1/Malaria_Strategic_Plan_Full_Version_02_27_14.pdf (2014). Accessed 20 Feb 2016. 11. Mwaiswelo R Ngasala BE Jovel I Gosling R Premji Z Poirot E Safety of a single low-dose of primaquine in addition to standard artemether-lumefantrine regimen for treatment of acute uncomplicated Plasmodium falciparum malaria in Tanzania Malar J. 2016 15 316 10.1186/s12936-016-1341-3 27287612 12. WHO. Methods for surveillance of antimalarial drug efficacy. Geneva, World Health Organization, 2009. 13. Urbaniak GC, Plous S. Research Randomizer (version 4.0) (Computer software). http://www.randomizer.org/ (2013). Accessed 22 June 2013. 14. NMCP. Tanzania Malaria Programme Review 2010. Programme Review Proposal. National Malaria Control Programme. Ministry of Health and Social Welfare. Dar es Salaam, Tanzania. http://www.rollbackmalaria.org/files/files/countries/Tanzania-The-malaria-program-performance-review-2010.pdf. Accessed 12 Jan 2016. 15. Eziefula AC Bousema T Yeung S Kamya M Owaraganise A Gabagaya G Single dose primaquine for clearance of Plasmodium falciparum gametocytes in children with uncomplicated malaria in Uganda: a randomised, controlled, double-blind, dose-ranging trial Lancet Infect Dis. 2014 14 130 139 10.1016/S1473-3099(13)70268-8 24239324 16. Hsiang MS Lin M Dokomajilar C Kemere J Pilcher CD Dorsey G Greenhouse B PCR-based pooling of dried blood spots for detection of malaria parasites: optimization and application to a cohort J Clin Microbiol 2010 48 3539 3543 10.1128/JCM.00522-10 20686079 17. Snounou G Zhu X Siripoon N Jarra W Thaithong S Brown K Biased distribution of mspl populations in Thailand and msp2 allelic variants in Plasmodium falciparum Trans R Soc Trop Med Hyg 1999 2 369 374 10.1016/S0035-9203(99)90120-7 10674079 18. Djimdé A Doumba O Cortese J Kayentao K Doumbo S Diourte Y A molecular marker for chloroquine-resistant falciparum malaria N Engl J Med 2001 344 257 263 10.1056/NEJM200101253440403 11172152 19. Ngasala BE Malmberg M Carlsson AM Ferreira PE Petzold MG Blessborn D Efficacy and effectiveness of artemether-lumefantrine after initial and repeated treatment in children <5 years of age with acute uncomplicated Plasmodium falciparum malaria in rural Tanzania: a randomized trial Clin Infect Dis 2011 52 873 882 10.1093/cid/cir066 21427394 20. Kamugisha E Jing S Minde M Kataraihya J Kongola G Kironde F Efficacy of artemether-lumefantrine in treatment of malaria among under-fives and prevalence of drug resistance markers in Igombe-Mwanza Malar J. 2012 11 58 10.1186/1475-2875-11-58 22369089 21. Shayo A Mandara CI Shahada F Buza J Lemnge MM Ishengoma DS Therapeutic efficacy and safety of artemether-lumefantrine for the treatment of uncomplicated falciparum malaria in North-Eastern Tanzania Malar J. 2014 13 376 10.1186/1475-2875-13-376 25240962 22. Shekalaghe S Drakeley C Gosling R Ndaro A van Meegeren M Enevold A Primaquine clears submicroscopic Plasmodium falciparum gametocytes that persist after treatment with sulphadoxine-pyrimethamine and artesunate PLoS One 2007 2 1023 10.1371/journal.pone.0001023 23. Malmberg M Ferreira PE Tarning J Ursing J Ngasala B Björkman A Plasmodium falciparum drug resistance phenotype as assessed by patient antimalarial drug levels and its association with pfmdr1 polymorphisms J Infect Dis 2013 207 842 847 10.1093/infdis/jis747 23225895 24. Sisowath C Stromberg J Martensson A Msellem M Obondo C Bjorkman A In vivo selection of Plasmodium falciparum pfmdr1 86 N coding alleles by artemether-lumefantrine (Coartem) J Infect Dis 2005 191 1 4 10.1086/427997 15592995 25. Sisowath C Petersen I Veiga MI Mårtensson A Premji Z Björkman A In vivo selection of Plasmodium falciparum parasites carrying the chloroquine-susceptible pfcrt K76 allele after treatment with artemether-lumefantrine in Africa J Infect Dis 2009 199 750 757 10.1086/596738 19210165 26. Henriques G Hallett RL Beshir KB Gadalla NB Johnson RE Burrow R Directional selection at the pfmdr1, pfcrt, pfubp1, and pfap2mu loci of Plasmodium falciparum in Kenyan children treated with ACT J Infect Dis 2014 210 2001 2008 10.1093/infdis/jiu358 24994911 27. Malmberg M Ngasala B Ferreira PE Larsson E Jovel I Hjalmarsson A Temporal trends of molecular markers associated with artemether-lumefantrine tolerance/resistance in Bagamoyo district, Tanzania Malar J. 2013 12 103 10.1186/1475-2875-12-103 23506218

PMC005xxxxxx/PMC5002102.txt

==== Front Arthritis Res TherArthritis Res. TherArthritis Research & Therapy1478-63541478-6362BioMed Central London 108810.1186/s13075-016-1088-9Research ArticleSerum levels of interleukin-17 and adiponectin are associated with infrapatellar fat pad volume and signal intensity alteration in patients with knee osteoarthritis Wang Kang wangkang96@yahoo.cn 12Xu Jianhua xujianhua86@yahoo.cn 1Cai Jingyu cjynigao@163.com 1Zheng Shuang shuangzhengl@qq.com 1Han Weiyu weiyu.han@utas.edu.au 2Antony Benny benny.eathakkattuantony@utas.edu.au 2Ding Changhai 61-3-62267730changhai.ding@utas.edu.au 121 Department of Rheumatology and Immunology, Arthritis Research Institute, the First Affiliated Hospital of Anhui Medical University, 218 Jixi Street, Hefei, China 2 Menzies Institute for Medical Research, University of Tasmania, Private Bag 23, Hobart, Tasmania 7000 Australia 26 8 2016 26 8 2016 2016 18 1 19313 5 2016 3 8 2016 © The Author(s). 2016 Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Background In the present study, we sought to generate hypotheses regarding the associations of serum levels of interleukin (IL)-17, adiponectin, and resistin with magnetic resonance imaging-measured infrapatellar fat pad (IPFP) size and signal intensity alterations in patients with knee osteoarthritis (OA). Methods A total of 170 subjects with symptomatic knee OA (mean age 55.4 years, range 34–74, 88.2 % females) were included. IPFP volume was measured on T1-weighted spoiled gradient-recalled acquisition in the steady state images and was computed by using a software program. IPFP high signal intensity (grades 0–3) was assessed on T2-weighted fast spin echo images. Serum IL-17, adiponectin, and resistin levels were measured using an enzyme-linked immunosorbent assay. Results In multivariable analyses, serum IL-17 was negatively associated with IPFP volume (β = −0.185, 95 % CI −0.337 to −0.034) but positively associated with the severity of IPFP signal intensity alteration (OR 1.23, 95 % CI 1.06–1.42) after adjustment for age, sex, weight, and height. Serum adiponectin was positively associated with IPFP volume (β = 0.016, 95 % CI 0.001–0.032) but negatively associated with IPFP signal intensity alteration (OR 0.99, 95 % CI 0.98–1.00) after adjustment for covariates. Resistin was positively associated with IPFP signal intensity alteration (OR 1.13, 95 % CI 1.04–1.23) but not with IPFP volume. The significant associations of adiponectin or resistin disappeared after further adjustment for IL-17; in contrast, the significant associations of IL-17 remained after further adjustment for adiponectin. Conclusions While serum IL-17 and resistin were associated with reduced IPFP volume and/or increased abnormal signal intensity alteration, serum adiponectin had opposite associations that were largely through IL-17. These findings suggest that serum adipocytokines may have a role to play in IPFP changes of knee OA. Keywords OsteoarthritisInfrapatellar fat padAdipokinesCytokinesCartilageEpidemiologyhttp://dx.doi.org/10.13039/501100001809National Natural Science Foundation of China81172865Ding Changhai issue-copyright-statement© The Author(s) 2016 ==== Body Background Osteoarthritis (OA) is a common joint disease with multiple pathogenetic mechanisms, affecting the whole joint. Obesity is considered one of the potent risk factors for developing OA [1]. Although the pathogenesis of OA is not well known, recent studies indicate that obesity-related proinflammatory and metabolic factors contribute to OA progression [2]. Infrapatellar fat pad (IPFP), a local adipose tissue in knee joints with an abundance of adipocytes, immune cells, vessels, and nerve fibers, may be an active joint tissue involved in the initiation and progression of knee OA [3]. IPFP, also known as Hoffa’s fat pad, is an intracapsular and extrasynovial structure [4]. It is thought to have primarily a biomechanical function of absorbing forces generated in the knee [5], and thus to have a protective effect physiologically. Indeed, in two recent studies, researchers reported that larger IPFP size was associated with reduced knee abnormal structural changes and symptoms [5, 6]. On the contrary, IPFP signal intensity alteration assessed using magnetic resonance imaging (MRI) was positively associated with the prevalence and/or incidence of knee pain, cartilage defects, bone marrow lesions, and radiographic osteoarthritis (ROA) in older adults [7], and thus it may have a detrimental effect. IPFP has been considered an important source of cytokines and adipokines (i.e., adipocytokines) in OA [8]. IPFP tissues acquired from patients with knee OA can produce cytokines such as interleukin (IL)-1β, tumor necrosis factor (TNF)-α, IL-6, IL-8, IL-17, basic fibroblast growth factor, and vascular endothelial growth factor (VEGF) [8–10]. It can also produce various adipokines such as leptin, resistin, and adiponectin [11]. Researchers in an in vitro study reported that conditioned media from cultured white adipose tissue from OA IPFP that contained leptin could induce cartilage collagen release and increase matrix metalloproteinase (MMP)-1 and MMP-13 expression in chondrocytes, and thus had catabolic effects on cartilage [12]. Although the local inflammatory profile of IPFP tissue has been investigated in vitro, the associations between serum adipocytokines and IPFP changes have not been examined in epidemiological or clinical studies. Osteoarthritic IPFP explants can release high protein levels of IL-17, resistin, and adiponectin [10]. While serum IL-17 and resistin (proinflammatory adipocytokines) concentrations were significantly higher in patients with knee OA than in control subjects [13], serum levels of adiponectin (an anti-inflammatory adipocytokine) were associated with decreased disease severity of knee OA [14]. It is unknown if IPFP measures such as size and signal intensity alteration are associated with increased or decreased release of systemic IL-17, resistin, and adiponectin in knee OA. We expected that IPFP size would be associated with decreased levels of serum IL-17 and resistin and an increased level of serum adiponectin, but IPFP high signal intensity would be associated with increased levels of serum IL-17 and resistin and a decreased level of serum adiponectin in patients with knee OA. The aim of this study was therefore to generate hypotheses regarding the associations between serum levels of IL-17, adiponectin, and resistin and the volume and signal intensity alteration of IPFP measured using MRI in patients with symptomatic knee OA. Methods Study design and patients We consecutively enrolled into the Anhui Osteoarthritis Study 205 patients with OA who fulfilled the American College of Rheumatology criteria for the classification of clinical knee OA at the Outpatient Clinics, Department of Rheumatology, First Affiliated Hospital of Anhui Medical University, from January 2012 to November 2013. We excluded institutionalized persons, patients with rheumatoid arthritis or other inflammatory diseases, patients with severe knee OA who were planning to have knee arthroplasty in 2 years (this study was ongoing with 2 years of follow-up), and patients with contraindications to MRI (including metal sutures, presence of shrapnel, iron filings in the eye, and claustrophobia). The study was approved by the First Affiliated Hospital of Anhui Medical University Ethics Committee, and all participants signed the informed consent forms. Anthropometrics Weight was measured to the nearest 0.1 kg (with shoes, socks, and bulky clothing removed) using a single pair of electronic scales (RGZ-120; Jiangsu Province, China) that were calibrated using a known weight at the beginning of each clinic. Height was measured to the nearest 0.1 cm by using a stadiometer with shoes, socks, and headgear removed. Body mass index (BMI) (weight [kg]/height [m2]) was calculated. Radiographic OA assessment All patients underwent knee radiography. The 15-degree flexion, standing, anteroposterior view image was taken in the symptomatic knees (the severer one if both knees were affected; the right one if both knees were equally painful). Radiographic assessment was performed by a radiology specialist using the Kellgren-Lawrence (K-L) grading system (grades 0–4): grade 0, normal; grade 1, no joint space narrowing (JSN), suspicious osteophytes; grade 2, suspicious JSN, mild osteophytes; grade 3, definite JSN, moderate osteophytes, and/or subchondral bone sclerosis; and grade 4, marked JSN, large osteophytes, and/or severe subchondral bone sclerosis [15]. Radiographic OA was defined as a K-L grade ≥2. Assessments of IPFP volume and signal intensity changes MRI of the selected knee was performed with a 3.0-T whole-body magnetic resonance imaging unit (Signa HDxT 3.0T; GE Healthcare, Little Chalfont, UK), using a commercial transmit/receive extremity coil. The following sequence and parameters were used:A T1-weighted fat saturation three-dimensional spoiled gradient-recalled acquisition in the steady state (SPGR), flip angle 30 degrees; repetition time 31 milliseconds; echo time 6.71 milliseconds; field of view 16 cm; 60 partitions; 512 × 512-pixel matrix; acquisition time 11 minutes, 56 milliseconds; and 1 acquisition (Sagittal images were obtained at a partition thickness of 1.5 mm and an in-plane resolution of 0.31 × 0.31 [512 × 512 pixels].) A T2-weighted fat saturation two-dimensional fast spin echo, flip angle 90 degrees, repetition time 3067 milliseconds, echo time 112 milliseconds, field of view 16 cm, 15 partitions, 256 × 256-pixel matrix (Sagittal images were obtained at a slice thickness of 4 mm with an interslice gap of 1.0 mm. Images were checked for image noise and structural abnormalities interfering with segmentation.) IPFP area was measured by manually drawing disarticulation contours around the IPFP boundaries on T1-weighted SPGR MRI scans using OsiriX software (Pixmeo Sàrl, Bernex, Switzerland). IPFP volume was computed by using the software program. IPFP signal intensity alteration on T2-weighted MRI studies was recorded if hyperintense signal alterations were observed within the IPFP. Signal intensity alteration, defined as discrete areas of increased signal within the IPFP, was graded as follows: grade 0, none; grade 1, <10 % of the region; grade 2, 10–20 % of the region; and grade 3, >20 % of the region [16]. All images were grouped together and read in randomized order by two readers, with the reader blinded to subjects’ information. Intraobserver reliability was measured in 30 subjects, with an intraclass correlation coefficient of 0.95 (95 % CI 0.89–0.97). Interobserver reliability was 0.94 (95 % CI 0.85–0.97). Serum adipocytokine measurements Morning fasting blood samples were collected from patients. Serum was separated, aliquoted into plastic storage tubes, and stored at −80 °C until analysis. Serum levels of IL-17, adiponectin, and resistin were measured by enzyme-linked immunosorbent assay (ELISA) (eBioscience, San Diego, CA, USA) kits according to the manufacturer’s instructions. The optical density was measured at 450 nm using an automatic ELISA reader (Sunrise; Tecan, Männedorf, Switzerland). The limits of detection for IL-17, adiponectin, and resistin were 0.5 pg/ml, 0.01 ng/ml, and 3.1 pg/ml, respectively, and the coefficients of variation were 7.1 %, 4.2 %, and 5.1 %, respectively. Data analysis Student’s t test, the χ2 test, and the Mann-Whitney U test was used to compare means, proportions, and medians, respectively. Pearson’s correlations or Spearman’s analyses were used to analyze the correlations of serum levels of IL-17 (log-transformed) with serum levels of adiponectin and resistin (log-transformed) or K-L grading. Linear regression analyses were used to examine the associations between IL-17, adiponectin, or resistin and IPFP volume (the dependent variable that was normally distributed) before and after adjustment for covariates including age, sex, weight, and height. Ordinal regression analyses were used to examine the associations between IL-17, adiponectin, or resistin and IPFP signal intensity alteration (the dependent variable). Standard diagnostic checks of model fit and residuals were routinely done, and data points with large residuals and/or high influence were investigated for data errors. A p value <0.05 (two-tailed) or a 95 % CI not including the null point was regarded as statistically significant. All statistical analyses were performed using SPSS 13.0 for Windows (SPSS, Chicago, IL, USA). Results A total of 170 subjects (88.2 % females) aged between 34 and 74 years (mean 55.4 years) were included in the analyses, and another 35 subjects were excluded from the study because of incomplete data. There were no significant differences in demographic factors between those included and excluded (data not shown). The median IL-17 level was 2.00 pg/ml, the mean IPFP volume was 20.46 ml (range 12.39–42.90 ml), and the IPFP signal intensity alteration mean value was 0.78 (range 0–3). There were 73 % patients who had established ROA. Characteristics of the subjects based on the median value of IL-17 are presented in Table 1. Patients with higher and lower levels of IL-17 were similar in terms of sex, BMI, height, weight, knee OA, IPFP signal intensity alteration, IPFP volume, and resistin level. However, patients with higher levels of IL-17 were older and had lower adiponectin levels. Serum levels of IL-17 were significantly associated with serum levels of adiponectin (r = −0.234, p = 0.002) and resistin (r = 0.165, p = 0.032). Serum IL-17 and resistin were not associated with K-L grade, but adiponectin was negatively correlated with K-L grade (ρ = −0.211, p = 0.006).Table 1 Characteristics of participants (split by median level of interleukin 17) Total (n = 170) IL-17 ≤ median (n = 85) IL-17 > median (n = 85) p Values Age, yearsa 55.45 (8.25) 54.07 (8.17) 56.66 (8.39) 0.044 Female sex, %b 88.2 89 87 0.634 Height, cma 159.00 (6.89) 159.09 (7.01) 158.58 (6.88) 0.638 Weight, kga 64.83 (10.56) 64.23 (9.62) 65.09 (10.70) 0.591 BMI, kg/m2a 25.64 (3.85) 25.35 (3.20) 25.87 (3.88) 0.341 K-L grade, %b 0.268  1 27.9 25.7 28.6  2 41.9 50.0 35.1  3 25.7 20.3 31.2  4 4.5 4.1 5.2 Knee ROA, %b 72.1 74.3 71.4 0.689 IPFP signal intensity alteration, %b 43.1 44.2 40 0.298 IPFP volume, mm3a 20.46 (5.01) 20.57 (5.53) 20.42 (4.88) 0.855 Adiponectin, pg/mlc 18.67 (3.94–53.07) 35.08 (11.36–71.26) 12.82 (3.00–46.94) 0.001 Resistin, ng/mlc 2.37 (1.40–5.57) 2.18 (1.45–4.86) 2.81 (1.32–5.78) 0.532 Abbreviations: BMI body mass index, K-L Kellgren-Lawrence, IL interleukin, ROA radiographic osteoarthritis, IPFP infrapatellar fat pad ROA was defined as a K-L grade ≥2. IL-17 median level 2.00 pg/ml (interquartile range 1.56–2.82). Data in bold denote statistically significant results a t tests were used for mean (SD) bχ2 tests were used for the proportions cMann-Whitney U tests were used for median (interquartile range) Serum level of IL-17 was significantly and negatively associated with IPFP volume after adjustment for age, sex, weight, and height (Table 2, Fig. 1), and this association remained unchanged after further adjustment for adiponectin (Table 2). Serum level of adiponectin was significantly and positively associated with IPFP volume after adjustment for age, sex, weight, and height, but this became nonsignificant after further adjustment for IL-17 (Table 2). There was no significant association between resistin and IPFP volume before or after adjustment for potential confounders.Table 2 Association between adipocytokines and infrapatellar fat pad volume Adipocytokines Multivariablea β (95 % CI) p Value Multivariable β (95 % CI) p Value IL-17 −0.19 (−0.34 to −0.03) 0.017 −0.16 (−0.32 to −0.01) b 0.047 Adiponectin 0.02 (0.01 to 0.03) 0.039 0.01 (−0.01 to 0.03)c 0.190 Resistin −0.07 (−0.20 to 0.06) 0.303 0.02 (−0.01 to 0.18)c 0.773 IL interleukin Dependent variable: infrapatellar fat pad volume. Independent variable: cytokines. Data in bold denote statistically significant results aAdjusted for age, sex, weight and height bFurther adjustment for adiponectin cFurther adjustment for IL-17 Fig. 1 Association between serum IL-17 and infrapatellar fat pad volume in patients with knee osteoarthritis. r = −0.199, p = 0.015 after adjustment for age, sex, height and weight. IL interleukin, IPFP infrapatellar fat pad Serum levels of IL-17 were significantly and positively associated with severity of IPFP signal intensity alteration before and after adjustment for age, sex, weight, and height, and this association remained significant after further adjustment for adiponectin (Fig. 2, Table 3). Adiponectin had a negatively significant and resistin a positively significant association with IPFP signal intensity alteration before and after adjustment for age, sex, weight, and height, but these associations became nonsignificant after further adjustment for IL-17 (Table 3).Fig. 2 Association between serum IL-17 and infrapatellar fat pad signal intensity alteration in patients with knee osteoarthritis. Serum IL-17 levels were higher in those with signal intensity alteration (p = 0.001) Table 3 Association between adipocytokines and infrapatellar fat pad signal intensity alteration Adipocytokines Multivariablea OR (95 % CI) p Value Multivariable OR (95 % CI) p Value IL-17 1.23 (1.06–1.42) 0.007 1.20 (1.03–1.38) b 0.017 Adiponectin 0.99 (0.98–1.00) 0.042 0.99 (0.98–1.00)c 0.177 Resistin 1.13 (1.04–1.23) 0.004 1.11 (1.00–1.23)c 0.054 IL interleukin Dependent variables: IPFP signal intensity alteration (0–3). Independent variable: adipocytokines. Data in bold denote statistically significant results aAdjusted for age, sex, weight, and height bFurther adjustment for adiponectin cFurther adjustment for IL-17 The significant associations of IL-17 and resistin remained significant, but associations of adiponectin became nonsignificant, after adjustment for K-L grade (data not shown). Discussion To our knowledge, this is the first epidemiological study to illustrate the relationship between serum levels of adipocytokines and IPFP volume and signal intensity alteration in patients with knee OA. After adjustment for age, sex, weight, and height, we found that higher serum IL-17 level was associated with lower IPFP volume and increased IPFP signal intensity alteration (indicating poorer IPFP quality). Higher adiponectin level was associated with higher IPFP volume and better IPFP quality, but higher serum resistin level was associated only with poorer IPFP quality. These suggest that serum levels of adipocytokines are associated with IPFP pathophysiology in knee OA. The role of IPFP in OA progression is inconclusive. In our previous studies, we have reported that IPFP size may have a protective role in knee OA structural changes: Greater IPFP maximal area is associated with greater cartilage volume and lesser cartilage defects, bone marrow lesions, and R OA cross-sectionally and/or longitudinally in older adults [5, 6]. Although researchers in one study reported that IPFP volume was correlated only with age in patients with knee OA and might not be related to the progression of OA [17], we recently reported that, in patients with knee OA, IPFP volume was associated with greater cartilage volume and fewer cartilage defects, bone marrow lesions, and osteophytes [18]. The signal intensity changes in IPFP were positively associated with the prevalence and/or incidence of knee pain, cartilage defects, bone marrow lesions, and R OA in older adults [7], suggesting that abnormal quality of IPFP is detrimental for knee symptoms and structures. This is supported by a recent study demonstrating that synovitis measured using IPFP signal intensity change (Hoffa’s synovitis) is associated with the development of radiographic knee OA [19]. The severity of inflammation in IPFP measured by dynamic contrast-enhanced MRI was associated with the severity of pain in knee OA [20]. Animal or in vitro studies have indicated a potentially important role of IPFP in mediating knee intra-articular inflammation. IPFP produced an elevated level of inflammatory cytokines, growth factors, and adipokines in a high-fat diet-induced murine OA model, and the expression levels of the adipokines were significantly correlated with expression of TNF-α, VEGF, and transforming growth factor β [21]. Rates of IL-6 expression and secretion were higher in IPFP tissue than in subcutaneous adipose tissue, indicating that the IPFP cytokine profile could play a role in paracrine inflammation via the local production of IL-6 and contribute to damage in adjacent cartilage in patients with OA [8]. IPFP explants from patients with OA could produce a variety of cytokines, and the production was increased by local cytokine stimulation [9]. Preliminary evidence has shown that adipocytokines such as IL-17 may have roles to play in knee OA. A significant proportion (up to 20 %) of chondrocytes in cartilage samples from patients with OA expressed IL-17R [22]. An in vitro study suggested that IL-17 induced the release of chemokines by chondrocytes and synovial fibroblasts, contributing to cartilage breakdown and synovial infiltration in OA [23]. Researchers in a cross-sectional study reported that serum IL-17 concentrations were significantly higher in patients with knee OA than in control subjects, and synovial IL-17 concentrations were positively correlated with K-L grade and Lequesne index in patients with knee OA [24]. So far, there have been no studies illustrating the relationship between serum IL-17 levels and IPFP changes in patients with knee OA. We found that serum level of IL-17 was negatively associated with IPFP volume and positively with IPFP high signal intensity in patients with knee OA. According to the biopsies from the infrapatellar region in knee OA, chronic low-grade inflammation was detected as signal changes on fluid-sensitive MRI sequences [25]. Our results suggest that serum IL-17 is associated with IPFP pathophysiology, but it is unknown if IL-17 contributes to IPFP inflammation or if abnormal IPFP may produce IL-17, which could play a role in OA pathogenesis. The causal relationship needs to be confirmed in future longitudinal studies. The role of adiponectin in OA is largely unclear. In vitro findings suggested that adiponectin increased nitric oxide and MMP production in human OA chondrocytes mainly via the AMP-activated protein kinase/c-Jun N-terminal kinase pathway, which would lead to accelerated degradation of OA cartilage matrix ex vivo [26]. Adiponectin can also induce IL-6 secretion in a cultured chondrogenic cell line [27]. In contrast, adiponectin levels in both plasma and synovial fluid decreased significantly as the severity of OA (evaluated by K-L grading) increased in humans [28]. Serum adiponectin levels were negatively associated with radiographic progression in patients with hand OA [29, 30]. Furthermore, adiponectin expression in IPFP was higher than in subcutaneous adipose tissue in patients with knee OA [8] and was higher in patients with OA in end stage than in early stage [31]. In our present study, the serum levels of adiponectin were positively associated with IPFP volume and negatively with IPFP signal intensity alteration, indicating that adiponectin may have a role in IPFP of knee OA. This association was dependent on IL-17 and K-L grade, suggesting that this may be mediated largely by IL-17 or ROA. The biological effects of resistin in OA remain controversial. Recently, researchers in one study reported that resistin could upregulate the expression of multiple chemokines, cytokines, and matrix-degrading genes that are involved in OA pathobiology [32]. Researchers in a cross-sectional study reported that serum resistin levels in patients with hand ROA were higher than in patients with nonradiographic hand OA and control subjects, and were associated with radiographic bone erosion in hand OA [33]. In the present study, we did not find a significant association between resistin and IPFP volume, but we found a significantly positive association between resistin and IPFP signal intensity alteration, suggesting a potential role of resistin in abnormal IPFP changes of knee OA. The present study has some limitations. First, it was a cross-sectional study designed to generate hypotheses, and the causal relationship cannot be interpreted. Second, the subjects were recruited from the clinics consecutively rather than from the community randomly, so the results may not generalizable to patients with general knee OA. Third, the signal intensity alteration of IPFP was assessed by unenhanced MRI, which was nonspecific, and its pathology is largely unclear [34]. Last, the levels of adipocytokines were measured in serum rather than in synovial fluid, so their local effects are unknown. Conclusions While serum IL-17 and resistin were associated with reduced IPFP volume and/or increased abnormal signal intensity alteration, serum adiponectin had opposite associations, which were largely through IL-17. These results suggest that serum levels of adipocytokines may have a role to play in IPFP changes of knee OA. Abbreviations BMIBody Mass Index ILInterleukin ELISAEnzyme-linked Immunosorbent Assay IPFPInfrapatellar Fat Pad JSNJoint Space Narrowing K-LKellgren-Lawrence MMPMatrix Metalloproteinase MRIMagnetic Resonance Imaging OAOsteoarthritis ROARadiographic Osteoarthritis SPGRSpoiled Gradient-recalled Acquisition in the Steady State TNFTumor Necrosis Factor VEGFVascular Endothelial Growth Factor Acknowledgements Special thanks go to the subjects who made this study possible. We also thank Hui Zhang for technical support. Funding This study was supported by the National Natural Science Foundation of China (81172865). Authors’ contributions KW had full access to all the study and took responsibility for the integrity of the data and accuracy of the statistical analyses. CD, JX, and KW designed the study. KW, JX, JC, SZ, WH, and BA acquired data. KW and CD interpretation data. KW, JX, JC, SZ, WH, BA, and CD prepared and approved the manuscript. Authors’ information CD’s research interests center on epidemiological and clinical investigation of osteoarthritis using modern techniques such as magnetic resonance imaging. He is looking at the inflammatory and metabolic mechanisms of osteoarthritis and osteoporosis, and he is interested in evaluating new therapies for osteoarthritis and other inflammatory diseases. He has received grants from competitive sources totaling more than $4 million, and he has published over 150 manuscripts in international peer-reviewed journals with an h-index of 35. Competing interests The authors declare that they have no competing interests. Ethics approval and consent to participate The study was approved by Anhui Medical University Medical Ethics Committee, and written informed consent was obtained from all participants. ==== Refs References 1. Grotle M Hagen KB Natvig B Dahl FA Kvien TK Obesity and osteoarthritis in knee, hip and/or hand: an epidemiological study in the general population with 10 years follow-up BMC Musculoskel Dis 2008 9 132 10.1186/1471-2474-9-132 2. Wang X Hunter D Xu J Ding C Metabolic triggered inflammation in osteoarthritis Osteoarthritis Cartilage 2015 23 22 30 10.1016/j.joca.2014.10.002 25452156 3. Clockaerts S Bastiaansen-Jenniskens YM Runhaar J Van Osch GJ Van Offel JF Verhaar JA The infrapatellar fat pad should be considered as an active osteoarthritic joint tissue: a narrative review Osteoarthritis Cartilage 2010 18 876 82 10.1016/j.joca.2010.03.014 20417297 4. Saddik D McNally EG Richardson M MRI of Hoffa’s fat pad Skeletal Radiol 2004 33 433 44 10.1007/s00256-003-0724-z 15221217 5. Han W Cai S Liu Z Jin X Wang X Antony B Infrapatellar fat pad in the knee: is local fat good or bad for knee osteoarthritis? Arthritis Res Ther 2014 16 R145 10.1186/ar4607 25008048 6. Pan F Han W Wang X Liu Z Jin X Antony B A longitudinal study of the association between infrapatellar fat pad maximal area and changes in knee symptoms and structure in older adults Ann Rheum Dis 2015 74 1818 24 10.1136/annrheumdis-2013-205108 24833783 7. Han W Aitken D Zhu Z Halliday A Wang X Antony B Signal intensity alteration in the infrapatellar fat pad at baseline for the prediction of knee symptoms and structure in older adults: a cohort study Ann Rheum Dis 2015 8. Klein-Wieringa IR Kloppenburg M Bastiaansen-Jenniskens YM Yusuf E Kwekkeboom JC El-Bannoudi H The infrapatellar fat pad of patients with osteoarthritis has an inflammatory phenotype Ann Rheum Dis 2011 70 851 7 10.1136/ard.2010.140046 21242232 9. Distel E Cadoudal T Durant S Poignard A Chevalier X Benelli C The infrapatellar fat pad in knee osteoarthritis: an important source of interleukin-6 and its soluble receptor Arthritis Rheum 2009 60 3374 7 10.1002/art.24881 19877065 10. Clockaerts S Bastiaansen-Jenniskens YM Feijt C De Clerck L Verhaar JA Zuurmond AM Cytokine production by infrapatellar fat pad can be stimulated by interleukin 1β and inhibited by peroxisome proliferator activated receptor α agonist Ann Rheum Dis 2012 71 1012 8 10.1136/annrheumdis-2011-200688 22307941 11. Eymard F Pigenet A Citadelle D Flouzat-Lachaniette CH Poignard A Benelli C Induction of an inflammatory and prodegradative phenotype in autologous fibroblast-like synoviocytes by the infrapatellar fat pad from patients with knee osteoarthritis Arthritis Rheumatol 2014 66 2165 74 10.1002/art.38657 24719336 12. Hui W Litherland GJ Elias MS Kitson GI Cawston TE Rowan AD Leptin produced by joint white adipose tissue induces cartilage degradation via upregulation and activation of matrix metalloproteinases Ann Rheumatic Dis 2012 71 455 62 10.1136/annrheumdis-2011-200372 13. de Boer TN van Spil WE Huisman AM Polak AA Bijlsma JW Lafeber FP Serum adipokines in osteoarthritis; comparison with controls and relationship with local parameters of synovial inflammation and cartilage damage Osteoarthritis Cartilage 2012 20 846 53 10.1016/j.joca.2012.05.002 22595228 14. Zheng S Xu J Xu S Zhang M Huang S He F Association between circulating adipokines, radiographic changes, and knee cartilage volume in patients with knee osteoarthritis Scand J Rheumatol 2016 45 224 9 10.3109/03009742.2015.1083053 26505548 15. Kellgren JH Lawrence JS Radiological assessment of osteoarthrosis Ann Rheum Dis 1957 16 494 502 10.1136/ard.16.4.494 13498604 16. Hunter DJ Guermazi A Lo GH Grainger AJ Conaghan PG Boudreau RM Evolution of semi-quantitative whole joint assessment of knee OA: MOAKS (MRI Osteoarthritis Knee Score) Osteoarthritis Cartilage 2011 19 990 1002 10.1016/j.joca.2011.05.004 21645627 17. Chuckpaiwong B Charles HC Kraus VB Guilak F Nunley JA Age-associated increases in the size of the infrapatellar fat pad in knee osteoarthritis as measured by 3 T MRI J Orthop Res 2010 28 1149 54 10.1002/jor.21125 20225314 18. Cai J Xu J Wang K Zheng S He F Huan S Association between infrapatellar fat pad volume and knee structural changes in patients with knee osteoarthritis J Rheumatol 2015 42 1878 84 10.3899/jrheum.150175 26276969 19. Roemer FW Kwoh CK Hannon MJ Hunter DJ Eckstein F Fujii T What comes first? Multitissue involvement leading to radiographic osteoarthritis: magnetic resonance imaging–based trajectory analysis over four years in the Osteoarthritis Initiative Arthritis Rheumatol 2015 67 2085 96 10.1002/art.39176 25940308 20. Ballegaard C Riis RG Bliddal H Christensen R Henriksen M Bartels EM Knee pain and inflammation in the infrapatellar fat pad estimated by conventional and dynamic contrast-enhanced magnetic resonance imaging in obese patients with osteoarthritis: a cross-sectional study Osteoarthritis Cartilage 2014 22 933 40 10.1016/j.joca.2014.04.018 24821663 21. Iwata M Ochi H Hara Y Tagawa M Koga D Okawa A Initial responses of articular tissues in a murine high-fat diet-induced osteoarthritis model: pivotal role of the IPFP as a cytokine fountain PLoS One 2013 8 e60706 10.1371/journal.pone.0060706 23593289 22. Honorati MC Meliconi R Pulsatelli L Canè S Frizziero L Facchini A High in vivo expression of interleukin-17 receptor in synovial endothelial cells and chondrocytes from arthritis patients Rheumatology 2001 40 522 7 10.1093/rheumatology/40.5.522 11371660 23. Honorati MC Bovara M Cattini L Piacentini A Facchini A Contribution of interleukin 17 to human cartilage degradation and synovial inflammation in osteoarthritis Osteoarthritis Cartilage 2002 10 799 807 10.1053/joca.2002.0829 12359166 24. Chen B Deng Y Tan Y Qin J Chen LB Association between severity of knee osteoarthritis and serum and synovial fluid interleukin 17 concentrations J Int Med Res 2014 42 138 44 10.1177/0300060513501751 24319050 25. Fernandez-Madrid F Karvonen RL Teitge RA Miller PR An T Negendank WG Synovial thickening detected by MR imaging in osteoarthritis of the knee confirmed by biopsy as synovitis Magn Reson Imaging 1995 13 177 83 10.1016/0730-725X(94)00119-N 7739358 26. Kang EH Lee YJ Kim TK Chang CB Chung JH Shin K Adiponectin is a potential catabolic mediator in osteoarthritis cartilage Arthritis Res Ther 2010 12 R231 10.1186/ar3218 21194467 27. Lago R Gomez R Otero M Lago F Gallego R Dieguez C A new player in cartilage homeostasis: adiponectin induces nitric oxide synthase type II and pro-inflammatory cytokines in chondrocytes Osteoarthritis Cartilage 2008 16 1101 9 10.1016/j.joca.2007.12.008 18261936 28. Honsawek S Chayanupatkul M Correlation of plasma and synovial fluid adiponectin with knee osteoarthritis severity Arch Med Res 2010 41 593 8 10.1016/j.arcmed.2010.11.007 21199727 29. Klein-Wieringa IR Andersen SN Herb-van Toorn L Kwekkeboom JC van der Helm-van Mil AHM Meulenbelt I Are baseline high molecular weight adiponectin levels associated with radiographic progression in rheumatoid arthritis and osteoarthritis? J Rheumatol 2014 41 853 7 10.3899/jrheum.130888 24692529 30. Yusuf E Ioan-Facsinay A Bijsterbosch J Klein-Wieringa I Kwekkeboom J Slagboom PE Association between leptin, adiponectin and resistin and long-term progression of hand osteoarthritis Ann Rheum Dis 2011 70 1282 4 10.1136/ard.2010.146282 21470970 31. Gandhi R Takahashi M Virtanen C Syed K Davey JR Mahomed NN Microarray analysis of the infrapatellar fat pad in knee osteoarthritis: relationship with joint inflammation J Rheumatol 2011 38 1966 72 10.3899/jrheum.101302 21765106 32. Zhang Z Xing X Hensley G Chang LW Liao W Abu-Amer Y Resistin induces expression of proinflammatory cytokines and chemokines in human articular chondrocytes via transcription and messenger RNA stabilization Arthritis Rheum 2010 62 1993 2003 10.1002/art.27510 20506172 33. Choe JY Bae J Jung HY Park SH Lee HJ Kim SK Serum resistin level is associated with radiographic changes in hand osteoarthritis: cross-sectional study Joint Bone Spine 2012 79 160 5 10.1016/j.jbspin.2011.04.009 21664853 34. Roemer FW Guermazi A Zhang Y Yang M Hunter DJ Crema MD Hoffa’s fat pad: evaluation on unenhanced MR images as a measure of patellofemoral synovitis in osteoarthritis AJR Am J Roentgenol 2009 192 1696 700 10.2214/AJR.08.2038 19457837